Characterization of Claudin-Dependent Morphogenetic Events During Neural Tube Closure and the Impact of CLDN Variants

Characterization of Claudin-dependent morphogenetic events during neural tube closure and the impact of CLDN variants

Amanda Baumholtz

Department of Human Genetics McGill University, Montreal, Canada February 2018

A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Doctor of Philosophy

ABSTRACT

The claudin family of tight junction proteins regulates paracellular permeability, apical- basal cell polarity, and cell adhesion and their cytoplasmic C-termini interact with the actin cytoskeleton. Through these activities, claudins have the potential to coordinate cell and tissue behaviors during epithelial morphogenesis. We previously showed that a subset of claudins is differentially expressed between the neural and non-neural epithelium during neural tube closure. This led to my hypothesis that these domains of claudin expression correlate to claudin function during neural tube morphogenesis and, if true, I predicted that deleterious missense mutations in CLDN genes would contribute to increased susceptibility to human neural tube defects (NTDs). I showed that selective removal of two of the eleven claudins expressed in the neural ectoderm of chick embryos, Cldn4 and -8, caused open NTDs due to defective convergent extension and failure of apical constriction at the neural plate midline. The failure of these morphogenetic events appears to be due to aberrant protein localization to the apical surface. In contrast, removing only Cldn3 from the non-neural ectoderm affected the epithelial remodeling events required for fusion of the dorsal tips of the neural folds to form the closed neural tube and continuous overlying layer of non-neural ectoderm. Claudin-depleted mouse embryos also exhibited NTDs. Sequence analysis of 125 patients with open spinal NTDs identified nine rare and five novel missense variants in nine CLDN genes. Functional validation studies revealed that overexpression of CLDN19 I22T and E209G, but not wild-type CLDN19, caused open NTDs in chick embryos due to defects in neural fold fusion and convergent extension, respectively. My data indicate that claudins play an evolutionary conserved role in vertebrate neural tube closure and that deleterious missense mutations in CLDN genes may contribute to human NTDs by impeding critical phases of neural tube closure. Furthermore, my data indicate that the combination of claudins expressed in an epithelial cell layer creates distinct compartments that regulate intracellular signaling events at the apical surface of epithelial cells to influence epithelial morphogenesis.

RÉSUMÉ

Les claudines sont des composantes intégrales des jonctions serrées et qui régulent la perméabilité paracellulaire, la polarité cellulaire apicale-basale, et l’adhésion cellulaire. De plus, le domaine C-terminale des claudines se lie au cytosquelette d’actine. Par l’entremise de ces activités, les claudines ont le potentiel de coordonner les comportements des cellules et des tissus pendant la morphogenèse épithéliale. Précédemment, nous avons démontré que l’expression des membres de la famille des claudines est différente dans le neuroépithélium comparé à l’épithélium non-neural pendant la fermeture du tube neural. Cette observation nous a mené à formuler l`hypothèse suivante: les patrons d’expression des claudines dans l’épithélium correspondent aux fonctions des claudines pendant la formation du tube neural. Si tel est le cas, des mutations pathogéniques dans les CLDN contribueraient aux facteurs génétiques augmentent le risque d’anomalies de fermeture du tube neural. Nous avons démontré que la suppression sélective de deux des onze claudines exprimées dans les cellules neuroectodermiques de l’embryon de poulet cause des anomalies de fermeture du tube neural dues aux défauts de l’extension convergente et à la constriction apicale des cellules au centre de la plaque neurale. L’échec de mouvements morphogéniques semble être causé par la relocalisation des protéines à la surface apicale. La suppression de Cldn3 dans les cellules non-neuroectodermiques cause des anomalies de fermeture du tube neural dues au fait que les plis neuraux, ainsi que l`ectoderme non-neurale, ne fusionnent pas. La suppression de certaines claudines dans les embryons de souris cause un défaut de fermeture du tube neural. Le séquençage de 125 patients avec des anomalies de fermeture du tube neural à l’extrémité caudale a identifié neuf changements nucléiques rares et cinq changements non-rapportés dans neuf CLDN. Les études fonctionnelles ont montré que la surexpression du changement protéique I22T et E209G dans CLDN19, mais pas dans CLDN19 sauvage, ne permettait pas la fermeture du tube neural dans les embryons de poulet. Ceci est dû au défaut de fusionnement des plis neuraux et à l’extension convergente, respectivement. Ces résultats suggèrent que les claudines jouent un rôle essentiel dans la fermeture du tube neural chez les vertébrés et que les changements faux-sense pathogéniques dans ces gènes peuvent contribuer aux anomalies de fermeture du tube neural chez les humains en bloquant certaines phases critiques lors de la fermeture du tube neural.

De plus, la combinaison de claudines exprimée par les tissus épithéliaux crée des microenvironnements qui servent à coordonner les évènements intracellulaires au niveau du pôle apical.

iii

TABLE OF CONTENTS

ABSTRACT ...... i RÉSUMÉ ...... ii LIST OF FIGURES ...... xiii CHAPTER I: INTRODUCTION AND LITERATURE REVIEW ...... 1 1.1 OVERVIEW AND RATIONALE FOR STUDY ...... 2 1.2 TIGHT JUNCTIONS ...... 3 1.2.1 Identification of tight junctions ...... 3 1.2.2 Evolutionary conservation of tight junctions ...... 4 1.2.3 Tight junction transmembrane proteins ...... 4 1.2.3.1 Claudins ...... 5 1.2.3.2 Occludin ...... 5 1.2.3.3 Tricellulin ...... 6 1.2.3.4 MarvelD3 ...... 7 1.2.3.5 Junctional adhesion molecules ...... 7 1.2.4 Tight junction cytoplasmic proteins ...... 8 1.2.4.1 ZO proteins ...... 8 1.2.4.2 Cingulin/paracingulin ...... 10 1.2.4.3 MUPP1 ...... 10 1.2.5 Formation of tight junctions...... 10 1.2.5.1 Crumbs polarity complex ...... 11 1.2.5.2 Par polarity complex ...... 11 1.2.6 Regulation of tight junction assembly/disassembly ...... 12 1.2.7 Functions of tight junctions ...... 12 1.2.7.1 The gate functions of tight junctions ...... 13 1.2.7.2 The fence function of tight junctions ...... 14 1.2.7.3 Tight junctions and adhesion ...... 14 1.2.7.4 Tight junctions and cell proliferation and gene expression ...... 15 1.3 THE CLAUDIN FAMILY OF TIGHT JUNCTION PROTEINS...... 16 1.3.1 Identification of claudins ...... 16

1.3.2 Structure of claudins ...... 16 1.3.3 Claudin-claudin interactions ...... 17 1.3.4 Post-translational modifications ...... 18 1.3.4.1 Phosphorylation ...... 18 1.3.4.2 Palmitoylation ...... 19 1.3.5 Genomic organization and evolutionary conservation of claudins ...... 19 1.3.6 Physiological and developmental roles for claudins ...... 21 1.3.6.1 Claudin gene deletion models in mouse and mutations in human disease ...... 21 1.3.6.1.1 Cldn1 and -6 create a water barrier in the skin ...... 21 1.3.6.1.2 Cldn5 is essential for the blood-brain barrier ...... 22 1.3.6.1.3 Cldn9, -11, and -14 in the inner ear prevent deafness ...... 22 1.3.6.1.4 Cldn11 and -19 are required in the nervous system for nerve conduction ... 22 1.3.6.1.5 Cldn2, -4, -7, -8, -10, -16, and -19 regulate paracellular permeability in the kidney ...... 23 1.3.6.1.6 Cldn2, -7, -15 and -18 regulate paracellular permeability in the gastrointestinal tract ...... 24 1.3.6.1.7 Cldn2 plays a role in the hepatobiliary system ...... 24 1.3.6.1.8 Cldn4 and 18 regulate the lung epithelial barrier ...... 25 1.3.6.1.9 Cldn11 is required for spermatogenesis ...... 25 1.3.6.2 Claudin function in adhesion and cell shape changes during embryonic morphogenesis ...... 26 1.3.7 Clostridium perfringens enterotoxin: a tool to study claudins ...... 27 1.3.7.1 The interaction between CPE and claudins ...... 27 1.3.7.2 The C-terminal claudin-binding domain of CPE ...... 29 1.4 NEURAL TUBE FORMATION ...... 30 1.4.1 Phases of neural tube closure ...... 30 1.4.1.1 Phase I: Formation of the neural plate ...... 30 1.4.1.2 Phase II: Shaping the neural plate ...... 31 1.4.1.2.1 Apical-basal thickening of neural plate cells contributes to neural plate shaping ...... 31 1.4.1.2.2 Convergent extension regulates neural plate shaping ...... 31

1.4.1.2.3 Oriented cell division is critical for neural plate shaping ...... 32 1.4.1.3 Phase III: Bending of the neural plate and elevation of the neural folds ...... 33 1.4.1.3.1 Median hinge point formation causes neural plate bending and neural fold elevation ...... 34 1.4.1.3.2 Dorsolateral hinge points contribute to neural fold convergence ...... 35 1.4.1.3.3 Expansion of the non-neural ectoderm generates extrinsic forces that contribute to neural fold elevation ...... 36 1.4.1.4 Phase IV: Fusion of the neural folds ...... 37 1.4.1.4.1 Cellular protrusions make the initial contact between the apposed neural folds ...... 37 1.4.1.4.2 Epithelial adhesion stabilizes the initial contact between the neural folds .. 39 1.4.1.4.3 Tissue remodeling separates the neural and non-neural ectoderm into distinct cell layers ...... 41 1.4.2 Discontinuous neural tube closure in mammals and birds ...... 41 1.4.3 Neural tube defects ...... 42 1.4.3.1 Types of open NTDs ...... 43 1.4.3.2 Causes of NTDs ...... 44 1.4.4 Using the chick as an animal model to study the role of claudins in neural tube closure ...... 45 1.4.4.1 A brief history of the chick as an experimental model ...... 45 1.4.4.2 The chick as an experimental model of neural tube closure ...... 46 1.4.4.3 Experimental manipulations in the chick ...... 47 1.4.4.4 Evolutionary conservation of claudins in the chicken genome ...... 47 1.4.4.5 Expression pattern of claudins during chick neural tube closure ...... 48 1.5 HYPOTHESIS AND OBJECTIVES ...... 49 2 CHAPTER II: CLAUDINS ARE ESSENTIAL FOR CELL SHAPE CHANGES AND CONVERGENT EXTENSION ...... 62 2.1 ABSTRACT ...... 63 2.2 INTRODUCTION ...... 64 2.3 RESULTS ...... 65 2.3.1 C-CPE-sensitive claudins are required for neural tube closure in chick embryos .. 65

2.3.2 Evolutionarily-conserved requirement for claudins in neural tube closure in mice 67 2.3.3 Depletion of C-CPE-sensitive claudins does not affect ectoderm differentiation .. 67 2.3.4 Claudins are required during neural plate shaping and median hinge point formation...... 68 2.3.5 C-CPE-treated embryos have convergent extension and apical constriction defects ...... 69 2.3.6 Claudins function upstream of PCP and RhoA/ROCK signaling ...... 70 2.3.7 C-CPE treatment affects cell morphology and protein localization at the apical surface...... 71 2.4 DISCUSSION ...... 72 2.4.1 Claudins in convergent extension movements ...... 72 2.4.2 Claudins function upstream of RhoGTPase signaling during neural tube closure . 74 2.5 MATERIALS AND METHODS ...... 76 2.5.1 Production of GST and GST-C-CPE fusion protein ...... 76 2.5.2 Embryo culture ...... 77 2.5.3 Immunofluorescence staining ...... 77 2.5.4 Whole mount in situ hybridization ...... 79 2.5.5 Transmission electron microscopy ...... 79 2.5.6 Morphometric assessment of axial length, length-to-width ratios, apical constriction and oriented cell division...... 79 2.5.7 Whole mount TUNEL assay ...... 80 2.5.8 Statistical Analyses ...... 80 2.6 ACKNOWLEDGEMENTS ...... 80 2.7 COMPETING INTERESTS ...... 80 2.8 CONNECTING TEXT BETWEEN CHAPTER II AND III ...... 97 3 CHAPTER III: CLDN3 IS REQUIRED IN THE NON-NEURAL ECTODERM TO REGULATE NEURAL FOLD FUSION ...... 98 3.1 ABSTRACT ...... 99 3.2 INTRODUCTION ...... 100 3.3 RESULTS ...... 102 3.3.1 Cldn3 is required for neural tube closure in chick embryos ...... 102

vii

3.3.2 Cldn3 is not required for bending and elevation of the neural folds ...... 103 3.3.3 Cldn3 is required for neural fold fusion ...... 104 3.3.4 C-CPELDR-treatment affects protein localization at the apical surface, but not apical-basal polarity, of non-neural ectoderm cells ...... 104 3.3.5 Cldn3 is required to form a meshwork of fibrils that connects the apposed neural folds ...... 105 3.4 DISCUSSION ...... 105 3.4.1 Cldn3 mediates the initial contact between the apposed neural folds ...... 106 3.4.2 Cldn3 regulates protein localization at the apical surface of non-neural ectoderm cells ...... 107 3.5 MATERIALS AND METHODS ...... 108 3.5.1 Production of GST and GST-C-CPE variants ...... 108 3.5.2 Embryo culture ...... 108 3.5.3 Immunofluorescence staining ...... 109 3.5.4 Whole mount in situ hybridization ...... 109 3.5.5 Scanning electron microscopy ...... 109 3.5.6 Transmission electron microscopy ...... 109 3.5.7 Morphometric analysis of length-to-width ratios and oriented cell division ...... 110 3.6 ACKNOWLEDGEMENTS ...... 110 3.7 CONNECTING TEXT BETWEEN CHAPTER III AND IV ...... 118 4 CHAPTER IV: MUTATIONS IN THE CLAUDIN FAMILY OF TIGHT JUNCTION PROTEINS CONTRIBUTE TO THE ETIOLOGY OF HUMAN NEURAL TUBE DEFECTS ...... 119 4.1 ABSTRACT ...... 120 4.2 INTRODUCTION ...... 121 4.3 MATERIALS AND METHODS ...... 123 4.3.1 Human subjects ...... 123 4.3.2 Next-generation DNA sequencing ...... 123 4.3.3 Bioinformatics...... 124 4.3.4 Sanger sequencing ...... 124 4.3.5 Generation of claudin wild-type and mutant constructs ...... 124

viii

4.3.6 Immunolocalization of ectopically expressed claudins ...... 125 4.3.7 Overexpression assay in the chick embryo ...... 125 4.4 RESULTS ...... 126 4.4.1 Variants identified in CLDN genes among myelomeningocele patients ...... 126 4.4.2 Rare and novel variants in CLDN3 ...... 126 4.4.3 Rare variant in CLDN4 ...... 127 4.4.4 Rare variant in CLDN6 ...... 127 4.4.5 Rare variant in CLDN9 ...... 128 4.4.6 Rare variant in CLDN16 ...... 128 4.4.7 Rare variant in CLDN18 ...... 128 4.4.8 Rare variants in CLDN19 ...... 128 4.4.9 Rare variant in CLDN23 ...... 129 4.4.10 Rare and novel variants in CLDN24 ...... 129 4.4.11 Patients with a mutation in a CLDN gene and a PCP gene ...... 130 4.4.12 The substitution N223S in CLDN16 impedes formation of tight junction strands ...... 130 4.4.13 Overexpressing CLDN19 variants causes open NTDs in chick embryos ...... 131 4.4.14 Variants in the Cldn8 cytoplasmic C-terminus disrupt neural tube closure ...... 131 4.4.15 Residues in the Cldn8 cytoplasmic C-terminus regulate protein localization at the apical surface of neural ectoderm cells ...... 132 4.5 DISCUSSION ...... 133 4.6 ACKNOWLEDGMENTS ...... 136 5 CHAPTER V: DISCUSSION ...... 154 5.1 OVERVIEW AND MAJOR FINDINGS ...... 155 5.2 GENERAL DISCUSSION ...... 159 5.2.1 The role of claudins in the neural ectoderm during neural tube closure ...... 159 5.2.1.1 Claudin function in convergent extension ...... 160 5.2.1.2 Claudin function in apical constriction ...... 161 5.2.2 The role of Cldn3 in the non-neural ectoderm during neural tube closure ...... 164 5.2.2.1 Cldn3 functions in epithelial fusion ...... 164

5.2.3 The role of CLDN rare and novel missense variants in the etiology of human NTDs ...... 166 5.2.3.1 Rare mutations in CLDN19 confer increased susceptibility to NTDs ...... 166 5.2.3.2 CLDN-gene interaction ...... 168 5.2.3.3 CLDN-environment interactions and folic acid ...... 169 5.2.4 Individual claudins have unique roles in regulating intracellular signaling events that affect epithelial morphogenesis ...... 170 5.3 FUTURE DIRECTIONS ...... 170 5.3.1 Defining the role of Cldn8 in neural tube closure ...... 171 5.3.1.1 Determining if Cldn8 mutant mice develop NTDs ...... 171 5.3.1.2 Determining if Cldn8 genetically interacts with PCP proteins to regulate neural tube closure ...... 172 5.3.1.3 Determining how Cldn8 regulates epithelial morphogenesis during neural tube closure ...... 172 5.3.2 Characterization of Cldn3-dependent epithelial remodeling events required for neural fold fusion ...... 173 5.3.2.1 Characterization of the meshwork that links the apposed neural folds ...... 173 5.3.2.2 Defining the interaction between Cldn3 and Eph receptors and ephrin ligands during neural fold adhesion ...... 173 5.3.3 Determining the contribution of CLDN mutations to the etiology of human NTDs ...... 174 5.3.3.1 Identifying how CLDN19 mutations cause NTDs ...... 174 5.3.3.2 Determining if CLDN variants interact with other genes to increase susceptibility to NTDs ...... 175 5.4 CONCLUDING SUMMARY ...... 176 5.5 ORIGINAL CONTRIBUTIONS ...... 177 REFERENCES ...... 179

LIST OF ABBREVIATIONS

Å: Ångström Amot: Angiomotin ATP: adenosine tri-phosphate BBS: Bardet-Biedl syndrome BMP: Bone morphogenetic protein Ca2+: Calcium ions Cl-: Chloride ions cDNA: Complementary deoxyribonucleic acid Cldn/CLDN: Claudin CPE: Clostridium perfringens enterotoxin CPE-R: Clostridium perfringens enterotoxin receptor CRB: Crumbs CRIB: Cdc42/Rac interactive binding DLHP: Dorsolateral hinge point DNA: Deoxyribonucleic acid DVL: Dishevelled E: Embryonic day ECL: extracellular loop ExAC: Exome Aggregation Consortium EVS: Exome Variant Server FZD: frizzled GEF: Guanine nucleotide exchange factors Grhl: Grainyhead-like GST: Glutathione S-transferase GTP: Guanosine triphosphate GAP: GTPase activating protein H&E: hematoxylin and eosin HH: Hamilton and Hamburger IBD: Inflammatory bowel disease IPTG: Isopropyl β-D-galactopyranoside JACOB: Junction-associated coiled-coil protein JAM: Junctional adhesion molecule kDa: Kilodalton MAF: Minor allele frequency MAGUK: Membrane associated guanylate kinase MAL: myelin and lymphocyte MARVEL: MAL related proteins for vesicle trafficking and membrane link MDCK: Manin-Darby canine kidney cell line Mg2+: Magnesium ions MHP: Median hinge point mIMCD3: mouse inner medullary collecting duct cells MLC: Myosin light chain MLCK: Myosin light chain kinase

mRNA: messenger RNA MTHFR: 5,10-methylene-tetrahydrofolate reductase MO: morpholino MUPP1: Multi-PDZ domain protein 1 Na+: sodium ions NES: Nuclear export signal NLS: Nuclear localization signal nm: Nanometer NTD: Neural tube defect OSP: Oligodendrocyte-specific protein PALS1: Protein associated with Line Seven 1 PAR: Partitioning defective PATJ: PALS1-associated tight junction protein Pax: Paired box PCNA: Proliferating cell nuclear antigen PCP: Planar cell polarity PDZ: Post-synaptic density 95/Discs large/zonula occludens-1 PKC: Protein kinase C pMLC: phosphorylated MLC RCAS: Replication competent avian retrovirus RNA: ribonucleic acid ROCK: Rho kinase SH3: Src homology 3 Shh: Sonic hedgehog siRNA: short interfering ribonucleic acid TAMP: Tight junction associated Marvel proteins TER: transepithelial electrical resistance WNT: Wingless integration site ZO: Zonula occludens ZONAB: ZO-1 associated nucleic acid binding protein C: Degrees celcius µg: Microgram ⁰µM: Micromolar

xii

LIST OF FIGURES

Figure 1.1 Schematic of the apical junctional complex in vertebrate epithelial cells ...... 51 Figure 1.2 Morphology of the tight junction ...... 52 Figure 1.3 Schematic of components of the tight junction and polarity complexes...... 53 Figure 1.4 Schematic of tight junction fence and gate functions...... 54 Figure 1.5 Schematic of a claudin molecule...... 55 Figure 1.6 Phases of neural tube closure in the chick embryo...... 56 Figure 1.7 The vertebrate non-canonical Wnt/planar cell polarity pathway...... 57 Figure 1.8 Schematic of key mechanisms regulating hinge point formation...... 58 Figure 1.9 Schematic of mechanisms involved in neural fold fusion...... 59

Figure 2.1 C-CPE-treated embryos exhibit dose-dependent, folic acid resistant neural tube defects...... 81 Figure 2.2 C-CPE prevents closure of the posterior neuropore in mouse embryos...... 82 Figure 2.3 C-CPE-sensitive claudins are required for apposition of neural folds...... 83 Figure 2.4 C-CPE-treated embryos exhibit convergent extension defects...... 84 Figure 2.5 Neural plate midline cells are not apically constricted in C-CPE-treated embryos. ... 85 Figure 2.6 PCP and RhoGTPase signaling proteins are reduced at tight junctions in C-CPE- treated embryos...... 87 Figure 2.7 C-CPE-sensitive claudins are required for Par3 and Cdc42 localization to the lateral membranes at the apical cell surface...... 88

Movie 2.1 The heart of C-CPE-treated embryos is beating...... 89

Supplemental Figure 2.1 Expression of claudins at neurulation in chick embryos...... 90 Supplemental Figure 2.2 Embryos treated with C-CPE are viable...... 91 Supplemental Figure 2.3 Differentiation and patterning of the neural and non-neural ectoderm were not affected by C-CPE treatment...... 92 Supplemental Figure 2.4 Embryos treated with C-CPE are viable and do not exhibit increased cell death...... 93 Supplemental Figure 2.5 Cldn4-specific C-CPE variant, C-CPELSID, does not cause NTDs...... 94 Supplemental Figure 2.6 Localization of Cldn14, RhoA, and Cdc42 is not affected in embryos treated with C-CPEYL...... 95 Supplemental Figure 2.7 Inhibitors of ROCK signaling did not affect Claudin expression or localization...... 96

Figure 3.1 Cldn3-specific C-CPE variant, C-CPELDR, causes NTDs...... 111 Figure 3.2. Formation of the hinge points is not affected in C-CPELDR-treated embryos...... 112 Figure 3.3. Oriented cell division, convergent extension and changes in cell shape of non-neural ectoderm cells are not affected in C-CPELDR-treated embryos...... 113 Figure 3.4. Aberrant neural crest migration in C-CPELDR-treated embryos...... 114 Figure 3.5. Non-neural ectoderm cells of C-CPELDR-treated embryos exhibit loss of epithelial integrity and an altered apical surface...... 115 Figure 3.6. A meshwork of thin fibrils connects the non-neural ectoderm of apposing neural folds during spinal neurulation...... 116

xiii

Figure 3.7. The apposed neural folds of C-CPELDR-treated embryos are not connected by fibrils...... 117

Figure 4.1. Chromatograms of rare and novel nonsynonymous CLDN mutations...... 137 Figure 4.2. Rare and novel CLDN mutations in NTD patients...... 138 Figure 4.3. Subcellular localization of CLDN variants...... 139 Figure 4 4. Open NTDs and convergent extension defects in chick embryos overexpressing CLDN variants...... 140 Figure 4.5. Embryos overexpressing CLDN19 E209G, but not CLDN19 I22T, exhibit convergent extension defects but apical constriction is not affected by CLDN19 variants...... 141 Figure 4.6. S198 and S216, but not the PDZ-binding domain, are critical residues for Cldn8 function...... 142

xiv

LIST OF TABLES

Table 1.1 Phenotypes of mouse claudin knockout and knockdown models and human diseases caused by claudin mutations ...... 60 Table 1.2 CPE sensitivity of the second extracellular loop of claudins ...... 61

Table 4.1. Common nonsynonymous variants in the coding sequence of CLDN1-25 ...... 143 Table 4.2. Novel and rare nonsynonymous variants in the coding sequence of CLDN1-25 ...... 144 Table 4. 3. Patients carrying mutations in a CLDN gene and core PCP gene ...... 145

Supplemental Table 4.1. Fluidigm primers ...... 146 Supplemental Table 4.2. Accession numbers of claudin amino acid sequences used for ortholog alignment...... 149 Supplemental Table 4.3. Primer sequences (5’->3’) used for PCR ...... 151 Supplemental Table 4.4. Primer sequences (5’->3’) used for RT-PCR ...... 152 Supplemental Table 4.5. Primer sequences used for mutagenesis ...... 153

ACKNOWDLEGMENTS

I would like to begin my thanking my supervisor Dr. Aimee Ryan for her support and guidance throughout my PhD. I am extremely thankful for the countless hours she has spent training me at the bench, discussing data, editing my manuscripts and committee meeting reports and helping me prepare for platform presentations. She has provided me with numerous opportunities to attend and present at scientific meetings which has helped foster my development as a scientist. Her encouragement and patience has made my time in the lab a delight and I am glad to have had such a great mentor and role model. Her continued passion for science is evident in our day-to-day interactions and is a source of inspiration. I would like to thank the members of my thesis supervisory committee, Dr. Peter Siegel and Dr. Yojiro Yamanaka, for their guidance and suggestions throughout the years. I would also like to thank Dr. Indra Gupta and Dr. Loydie Jerome-Majewska for their helpful discussions and suggestions during our weekly lab meetings. Furthermore, I would like to thank the many people at the McGill University core services including Dr. Min Fu at the RI-MUHC Imaging Platform and Dr. Jeannie Miu, Lee Ann Monaghan and Dr. David Lui at the Facility for Electron Microscropy Research for their technical assistance without which the data presented in this thesis would not have been possible. In addition, I would like to thank Dr. Pierre Lepage at Genome Quebec for his technical expertise in sequencing the genomic DNA of the neural tube defect patients and for his helpful discussions on analyzing the data. I have been fortunate to have worked with some amazing people during my time in the lab. I owe a huge thank you to Dr. Michelle Collins who supervised me when I was an undergraduate student in the lab and continued to play an important role in my growth as a scientist during my graduate studies. I am extremely thankful to her for training me at the bench, for her claudin discussions, and for her sense of humour which made my day-to-day experience in the lab more enjoyable. I would also like to thank other past and current lab members for their scientific discussion and support and help preparing for oral presentations. I would especially like to thank Enrique Gamero Estevez for being my lab partner in purifying the C-CPE protein. I have been lucky to be surrounded by a wonderful group of people during my time at Place Toulon and the Glen. I have made so many great friends over the years and I am thankful to them for making those tough days in the lab just a little bit easier. In particular, I would like

xvi to thank Dora Siontas, Marie-Lyne Fillion, Dominic Hou, Swati Gupta, Maria Laverde, and Fatima Tokhmafshan without whom I don’t think that I would have survived my PhD. I would also like to thank Dr. Karen Christensen for her folate talks and Kether Guerrero for his helpful discussions on how to analyze the patient mutations. I would also like to thank my family for their support throughout my degree. My parents, Steve and Esther, have supported me throughout my education and were understanding of the long hours and weekends that I spent in the lab. Lastly, I would like to thank our many collaborators including Dr. Nicholas Greene, Dr. Jörg Piontek, and Dr. Valeria Capra without whom the work presented in this thesis would not have been possible. This thesis is dedicated to all the patients and their families who have been affected by neural tube defects and I hope that the data presented in this thesis will, one day, contribute to reducing the worldwide incidence of neural tube defects

xvii

PREFACE TO THE THESIS

This thesis is written in accordance with the guidelines of the McGill University Graduate and Postdoctoral Studies Office. This thesis is comprised of five chapters. Three of these chapters are scientific data chapters written in manuscript format. Chapter II has been published and Chapter III and IV are manuscripts that are currently in preparation. Chapter I is a review of the current background literature relevant to this thesis. This chapter reviews tight junctions and claudins, neural tube closure, and the use of the chick embryo as a model system. This chapter also includes the rationale and objectives of the thesis. Chapter II to Chapter IV contain experimental data in manuscript format. Chapter II describes the role of Cldn4 and -8 in regulating morphogenesis of the neural ectoderm during neural tube closure and has been published in Developmental Biology and is reproduced in this thesis with permission. Chapter III is a manuscript in preparation describing the role of Cldn3 in regulating morphogenesis of the non-neural ectoderm during neural tube closure. Chapter IV is a manuscript in preparation describing the contribution of rare and novel missense mutations in CLDN1-25 to increased susceptibility to human neural tube defects. Connecting texts between Chapter II and III, and II and IV provide a link between the manuscripts. Chapter V is a summary and discussion of the experimental data in Chapters II to IV and describes future studies

xviii

CONTRIBUTION OF AUTHORS

The experimental research presented in this thesis was carried out by Amanda Baumholtz under the supervision of Dr. Aimee Ryan. Amanda Baumholtz carried out the majorty of experiments, participated in experimental design, analyzed data, wrote the first draft of each manuscript, and actively participated in the revision of manuscripts. Dr. Aimee Ryan supervised the work, participated in experimental design, analyzed data, and was involved in manuscript editing. The contribution of the remaining authors to each chapter are described below.

For Chapter II, Annie Simard assisted with setting up the ex ovo embryo culture system and processing and imaging embryos for transmission electron microscopy. Dr. Evanthia Nikolopoulou and Dr. Nicholas D.E. Greene carried out the mouse experiments. Marcus Oosenbrug performed whole mount in situ hybridization to analyze expression of markers in the somites. Dr. Michelle Collins participated in experimental design and participated in editing of the manuscript. Dr. Anna Piontek, Dr. Gerd Krause and Dr. Jörg Piontek provided the C-CPE variants C-CPEYL and C-CPELSID.

For Chapter III, Dr. Anna Piontek, Dr. Gerd Krause and Dr. Jörg Piontek provided the C-CPE variants C-CPEYL and C-CPELDR.

For Chapter IV, Dr. Valeria Capra, Patrizia DeMarco and Elisa Merello provided the genomic DNA of the patients with neural tube defects. .

xix

CHAPTER I: INTRODUCTION AND LITERATURE REVIEW

1.1 OVERVIEW AND RATIONALE FOR STUDY Epithelial morphogenesis describes the process by which epithelial cell layers assemble into complex tissues and organs. As tissues and organs forms, epithelial cells layers dynamically change their morphology via several processes including cell division, cell shape changes and cell rearrangements that are coordinated in time and space. A good model to study epithelial morphogenesis is neural tube closure. It involves formation of the neural plate, bending and elevation of the neural plate to create neural folds, which fuse at the midline to generate a closed neural tube. These phases of neural tube closure are regulated by changes in the behavior of neuroepithelial cells, which form the neural tube itself. During neural tube closure in avian and mouse embryos, neuroepithelial cell divisions are preferentially oriented along the anterior- posterior axis compared to the medial-lateral axis playing a role in lengthening of the neural plate. Intercalation of neuroepithelial cells contributes to the narrowing (convergence) and anterior-posterior lengthening (extension) of the neural plate. Cells at the midline of the neural plate change from columnar to wedge-shaped through contraction of the apical actin-myosin cytoskeleton causing the lateral edges of the neural plate to bend and elevate. Changes in the behavior of cells in the surrounding non-neural epithelium also contribute to neural tube closure. In particular, changes in the shape of non-neuroepithelial cells from cuboidal to squamous, cell rearrangements, and non-randomly oriented mediolateral cell divisions generate extrinsic forces that cause neural fold elevation and convergence.

Epithelial cells are polarized cells that are connected to each other by junctional complexes and are attached to a basal extracellular matrix. Throughout morphogenesis, epithelial cells remain connected through these intercellular junctions. The most apical junctional complex in epithelial cells is the tight junction. Claudins are integral tight junction proteins that regulate paracellular permeability, apical-basal cell polarity, cell adhesion, and interact with the actin cytoskeleton via their intracellular C-terminal tails. Our previous analysis of claudin expression patterns in chick embryos revealed that claudins exhibit distinct expression boundaries in the ectoderm prior to differentiation of this epithelial cell layer into neural and non-neural progenitors. I therefore hypothesized that the different combinations of claudins expressed in the ectoderm create microenvironments that regulate differentiation and morphogenesis of the neural and non-neural ectoderm during neural tube closure and that, if true, deleterious missense mutations in CLDN genes contribute to increased susceptibility to human neural tube defects.

The aims designed to address this hypothesis were to investigate the role of claudins in differentiation and morphogenesis of the neural and non-neural ectoderm during neural tube closure and to determine if mutations in CLDN genes are responsible for a subset of neural tube defects in human patients.

1.2 TIGHT JUNCTIONS Epithelial layers play a structural role to separate organs into functional biological compartments and a barrier function to separate the external and internal environments. These functions are mediated by intercellular junctions that attach epithelial cells to each other. The most apical and functionally diverse of these junctional complexes is the tight junction. Tight junctions have two canonical functions. First, they act as gates that restrict the paracellular movement of ions and small molecules based on charge and size. Second, they function as a fence to block the movement of membrane proteins and lipids between the apical and basolateral domains. In addition to their canonical functions, tight junctions act as signaling platforms to regulate gene expression, cell proliferation, and differentiation. Below is an overview of the structure and function of tight junctions.

1.2.1 Identification of tight junctions The observation that epithelial cells are connected by junctional complexes was first reported following transmission electron microscopy studies of epithelial tissues in the early 1960s (Farquhar and Palade, 1963). The junctional complexes identified in these studies were the zonula occludens, the zonula adherens, and then the more basal macula adherens, which are now commonly referred to as the tight junction, adherens junction and desmosome, respectively (Fig. 1.1). By transmission electron microscopy, the tight junction appears as a series of fusion or kissing points of adjacent cell membranes that seal the intercellular space (Fig. 1.2A). Freeze fracture electron microscopy subsequently revealed that tight junctions form a continuous network of strands that encircles the cell below the apical surface (Staehelin et al., 1969) (Fig. 1.2B). The tight junction structure varies across epithelial tissues due to variation in the number of tight junction strands and how they are organized. Each tight junction strand within the plasma membrane associates laterally with a tight junction strand in the membrane of an adjacent cell. These paired tight junction strands form a series of semipermeable barriers across the paracellular space. Electron microscopy studies also revealed a dense band of cytoplasmic

material associated with the tight junctions, which was subsequently named the cytoplasmic plaque (Farquhar and Palade, 1963). The cytoplasmic plaque consists of a complex of cytosolic proteins that are connected to the membrane proteins of the tight junction. In summary, these ultrastructural studies led to our understanding of the basic three-dimensional structure of the tight junction and suggest that the tight junction consists of a complex network of membrane and cytoplasmic proteins (Fig. 1.3).

1.2.2 Evolutionary conservation of tight junctions Although tight junctions are a unique feature of vertebrates, many of their protein components and functions are evolutionarily conserved. In invertebrates, septate junctions are thought to be the functional counterparts of the vertebrate tight junction (Jonusaite et al., 2016). Septate junctions are functionally similar to tight junctions in their ability to act as a barrier to regulate the paracellular movement of ions and small molecules. However, septate and tight junctions differ in their relative position with respect to other junctions in the membrane: tight junctions are the most apical junctions in vertebrates, while septate junctions are located basally to the adherens junctions. In ultrathin sections, septate junctions appear to form a ladder-like structure with steps or ‘septa’ connecting the plasma membrane of adjacent cells (Flower and Filshie, 1975; Gilula et al., 1970). Unlike the tight junction which seals the intercellular space, septate junctions are characterized by a 15 nm-wide intercellular space. Despite the positional and morphological differences, both tight junctions and septate junctions form a network of strands that circumscribes epithelial cells and both are composed of transmembrane and cytoplasmic proteins, supporting the idea that tight junctions and septate junctions have a common molecular basis that regulates their barrier function and that they share a common origin in evolution.

1.2.3 Tight junction transmembrane proteins The tight junction is composed of a complex of transmembrane and cytoplasmic proteins that interact to mediate the function of tight junctions in paracellular barriers, cell polarity, cell adhesion, cytoskeletal attachment, and signaling (Fig 1.3) (Van Itallie and Anderson, 2014). The main transmembrane proteins of tight junctions are claudins, occludin, tricellulin, MarvelD3, and junctional adhesion molecules (JAMs). Occludin, tricellulin and MarvelD3 are collectively called tight junction-associated Marvel proteins (TAMPs) because they contain a

four transmembrane myelin and lymphocyte (MAL) and related proteins for vesicle trafficking and membrane link (MARVEL) domain, a structural motif involved in membrane apposition and fusion. Together, the transmembrane proteins form the structural backbone of tight junction strands. Their extracellular and intracellular functions are described below.

1.2.3.1 Claudins The word claudin derives from the Latin word “claudere” that means “to close” and refers to the fact that claudins are the principle barrier-forming proteins of the tight junction. Claudins comprise a multigene family with 24 vertebrate family members. All claudins encode 20-27 kDa proteins with four transmembrane domains, two extracellular loops and cytoplasmic N- and C- termini (Furuse et al., 1998a). Claudins are the building blocks of tight junctions. Exogenous expression of individual claudins in tight junction-free fibroblasts results in formation of tight junction strands similar to those of epithelial cells (Furuse et al., 1998b). The claudin component of the tight junction determines the size and ion specificity of the tight junction. The claudin family of tight junction proteins are the principal focus of this thesis and are discussed in greater detail in later sections.

1.2.3.2 Occludin Occludin was the first transmembrane component of the tight junction identified and derives its name from the Latin word “occludere” that means “to occlude”. It is a ~65kDa protein with four transmembrane domains, two extracellular loops and cytoplasmic N- and C- termini (Furuse et al., 1993). Despite having a similar structure to claudins, occludin shares no sequence similarity with claudins (Furuse et al., 1998a). Occludin binds to the tight junction scaffolding proteins ZO-1, -2 and -3 through its C-terminal domain, which links occludin to the actin cytoskeleton (Furuse et al., 1994; Haskins et al., 1998; Itoh et al., 1999b). In addition, the C-terminal tail of occludin directly binds to F-actin, a property that is not shared by other tight junction transmembrane proteins that require the mediation of scaffolding proteins for actin association (Wittchen et al., 1999). The role of occludin as an integral tight junction protein remains unclear. When occludin is transfected into fibroblasts cells, which do not have endogenous tight junction strands, only a small number of short strands are formed (Furuse et al., 1998b). This is in contrast to claudins, which form a well-developed network of strands when exogenously expressed in fibroblasts

(Furuse et al., 1998b). Occludin-deficient embryonic stem cells differentiate into polarized embryoid bodies with well-developed tight junction strands that are morphologically indistinguishable from embryoid bodies derived from wild-type embryonic stem cells (Saitou et al., 1998). Occludin-null mice do not exhibit abnormalities in the structure or barrier function of their tight junctions (Saitou et al., 2000). These data suggest that occludin is an accessory protein and is not the main component of tight junction strands. However, several lines of evidence indicate that occludin has an important role at tight junctions. Deleting the C-terminal domain of occludin (Balda et al., 1996; Chen et al., 1997) or treating epithelial cells with synthetic peptides homologous to the occludin extracellular loops (Everett et al., 2006; Wong and Gumbiner, 1997) disrupts the barrier properties of the tight junction. Overexpression studies using C-terminally truncated occludin mutants in epithelial cells (Balda et al., 1996) and Xenopus embryos (Chen et al., 1997) show that occludin also modulates the fence properties of the tight junction as the polarized distribution of lipids is disrupted in tight junctions containing the mutant proteins. Furthermore, occludin mediates cell-cell adhesion since treating epithelial cells with a synthetic peptide homologous to the first extracellular loop decreases cell adhesion (Van Itallie and Anderson, 1997). Taken together, these data suggest that occludin is not required for the formation of tight junctions but has a supporting role in modulating the fence, barrier and cell adhesion properties of tight junctions, functions which can partially be replaced by other tight junction transmembrane proteins.

1.2.3.3 Tricellulin Tricellulin, also called MARVELD2, is a ~63kDa protein with a similar structure to occludin containing four transmembrane domains, two extracellular loops and cytoplasmic N- and C-termini. It preferentially localizes to tricellular contacts, where three cells join together (Ikenouchi et al., 2005). Like occludin, tricellulin can interact with ZO-1 via its C-terminal domain (Riazuddin et al., 2006). Tricellulin can form homodimers within a cell via interactions of its transmembrane domains (Westphal et al., 2010). Whether or not tricellulin can also form heterodimers with occludin is unclear (Raleigh et al., 2010; Westphal et al., 2010). Tricellulin plays a critical function in forming the tight junction barrier and in organizing tight junction strands. Knockdown of tricellulin in cultured monolayers compromises the tight junction barrier and results in breaks in tight junction strands (Ikenouchi et al., 2005). In a complementary

experiment, knockdown of occludin leads to the redistribution of tricellulin from tricellular to bicellular junctions (Ikenouchi et al., 2008). These observations suggest that occludin and tricellulin have unique and redundant roles and that tricellulin may compensate for some of the functions of occludin in the occludin null mouse.

1.2.3.4 MarvelD3 MarvelD3 is a 40kDa testraspan membrane protein with a longer N-terminal and shorter C- terminal domain relative to occludin (Sanchez-Pulido et al., 2002). The shorter C-terminal domain does not contain the ZO-1 binding site found in occludin and tricellulin. MarvelD3 can form heterodimers with occludin and tricellulin. Knockdown of MarvelD3 in cultured monolayers revealed that it not required for tight junction formation, but plays a role in modulating the paracellular barrier function of tight junctions (Steed et al., 2009).

1.2.3.5 Junctional adhesion molecules Junctional adhesion molecules (JAMs) comprise a group of three ~40 kDa proteins, JAM- A, -B and -C, that belong to the immunoglobulin superfamily (Martin-Padura et al., 1998). In contrast to TAMPs and claudins, they contain a single transmembrane domain and two immunoglobulin-like extracellular loops. The extracellular loops participate in homophilic and heterophilic interactions. All JAMs can form homodimers and heterophilic interactions can occur between JAM-B and JAM-C (Arrate et al., 2001; Cunningham et al., 2000). Additionally, the JAM extracellular domains interact with integrins. All JAMs have a Postsynaptic density 95/Discs Large/Zonula occludens-1 (PDZ)-binding domain in their C-terminus that interacts with PDZ adaptor proteins at the tight junction cytoplasmic plaque (Bazzoni et al., 2000; Ebnet et al., 2003; Monteiro et al., 2013). JAMs are not required for tight junction formation but they contribute to several tight junction functions including barrier formation, adhesion through homophilic and heterophilic interactions of their extracellular domain, signaling by associating with intracellular binding partners at their cytoplasmic C-terminus, and establishing apical-basal polarity through interactions with members of the cell polarity complex (Ebnet et al., 2003; Laukoetter et al., 2007; Rehder et al., 2006; Severson and Parkos, 2009).

1.2.4 Tight junction cytoplasmic proteins Work by several groups over the last 50 years has revealed that the electron dense tight junction cytoplasmic plaque observed by Farquahar and Palade (1963) is composed of a cluster of proteins including adaptor proteins, kinases, phosphatases, G proteins, and transcription factors (Zihni et al., 2016). The adaptor proteins, also referred to as scaffolding proteins, are the main structural components of the tight junction cytoplasmic plaque and include zonula occludens (ZO) proteins, cingulin/paracingulin, and multi-PDZ domain protein (MUPP1). These cytoplasmic proteins serve as a bridge between the transmembrane proteins of the tight junction and the actin cytoskeleton. They also have multiple protein-protein binding domains that allow them to function as scaffolds for the attachment of signaling proteins.

1.2.4.1 ZO proteins Zonula occludens 1 (ZO-1), a 225 kDa protein, was the first identified tight junction protein (Stevenson et al., 1986). Subsequent studies identified ZO-2, a 160kDa protein, and ZO- 3, a 130kDa protein, as binding partners of ZO-1 (Gumbiner et al., 1991; Haskins et al., 1998). ZO proteins belong to the membrane-associated guanylate kinase (MAGUK) family that has a core structure of three N-terminal PDZ domains, a central Src homology 3 (SH3) domain and a guanylate kinase (GUK) homology domain. ZO proteins act as molecular scaffolds within the tight junction cytoplasmic plaque connecting the transmembrane proteins of the tight junction to the actin cytoskeleton and signaling proteins. ZO proteins bind to each other through their PDZ domains and interact with the PDZ-binding motif found in the C-terminal tails of the transmembrane proteins claudins, occludin and JAMs (Monteiro et al., 2013). ZO-1 also associates with tricellulin (Riazuddin et al., 2006). All three ZO proteins directly bind to F-actin (Wittchen et al., 1999) and bind to actin binding proteins including cingulin (D'Atri et al., 2002) and afadin (Ooshio et al., 2010). ZO-1 also interacts with components of the Wnt-signaling and planar cell polarity pathways (Van Itallie et al., 2013). The sequence of ZO proteins contains nuclear localization and export signals that allow ZO proteins to shuttle between the nucleus and cytoplasm (Gonzalez-Mariscal et al., 1999). Nuclear ZO proteins interact with proteins that regulate gene expression and cell proliferation. For example, ZO-1 binds to the Y-box transcription factor ZO-1 associated nucleic acid binding protein (ZONAB) via its SH3 domain (Balda and Matter, 2000) and ZO-2 binds to the cell cycle regulator cyclin D1 (Huerta et al.,

2007). These interactions establish a crosstalk between the nucleus and tight junction that balances epithelial cell proliferation and differentiation. The ZO proteins are evolutionarily conserved across the metazoan kingdom, localizing to tight junctions in vertebrates and septate junctions in invertebrates. ZO proteins have unique and redundant roles in promoting tight junction assembly and barrier formation. Epithelial cells in which ZO-1 is knocked out and ZO-2 is knocked down are polarized but fail to form tight junctions suggesting that ZO proteins play a key role in marking the location of tight junctions and in the subsequent polymerization of claudins to form tight junction strands (Umeda et al., 2006). Exogenous expression of ZO-1 or ZO-2 into these cells rescues tight junction formation. In contrast, expression of exogenous ZO-3, which is not endogenously expressed in this cell line, does not rescue the phenotype (Umeda et al., 2006). These data suggest that ZO-1 and ZO-2, but not ZO-3, act functionally redundantly to promote tight junction assembly and that ZO proteins are not required for the establishment of apical- basal polarity. Studies in animal models revealed that ZO proteins are required for embryonic development. ZO-1 deficiency causes an embryonic lethal phenotype in mice associated with impaired yolk sac angiogenesis and extensive cell death in epithelial tissues including the neural tube (Katsuno et al., 2008). The Drosophila homologue of ZO-1, polychaetoid, regulates cell fate determination in sensory organs, plays a role in cell specification and rearrangement during tracheal morphogenesis, and is required for coordinated cell shape changes in the epidermis during dorsal closure (Chen et al., 1996; Choi et al., 2011; Jung et al., 2006). ZO-2 null mice die shortly after implantation as they fail to undergo normal gastrulation and exhibit decreased proliferation and increased apoptosis (Xu et al., 2008). Apical-basal polarity is maintained, but the structure and tight junction barrier are disrupted in ZO-2 knockout mice. In contrast to ZO-1 and ZO-2, ZO-3 is dispensable for mammalian embryonic development as mice lacking ZO-3 show no apparent phenotype with normal establishment of epithelial tight junctions (Adachi et al., 2006; Xu et al., 2008). However, ZO-3 is essential for epidermal barrier function in zebrafish embryos (Kiener et al., 2008). These data suggest that ZO proteins play a critical role in embryogenesis by regulating the structural integrity and barrier properties of tight junctions, cell proliferation and differentiation, and cell shape changes and rearrangement during morphogenesis.

1.2.4.2 Cingulin/paracingulin Cingulin and paracingulin/junction-associated coiled-coil protein (JACOB) are vertebrate- specific tight junction proteins characterized by a globular head domain, a coiled-coil rod domain, and a small globular tail (Guillemot et al., 2008b). Both cingulin and paracingulin interact with the F-actin, myosin, ZO proteins and JAMs. (Aijaz et al., 2005; Bazzoni et al., 2000; Cordenonsi et al., 1999; D'Atri and Citi, 2001; D'Atri et al., 2002; Guillemot et al., 2008b; Pulimeno et al., 2011). They also interact with RhoGTPases to regulate tight junction remodeling, gene expression, and cell proliferation (Guillemot et al., 2004; Guillemot et al., 2008a; Terry et al., 2011). Interestingly, cingulin regulates the expression levels of claudins through the RhoGTPase RhoA (Guillemot and Citi, 2006). In embryoid bodies lacking both copies of cingulin, tight junctions form normally but Cldn2 mRNA levels are upregulated. The effects on Cldn2 expression are reversed by expressing a dominant-negative form of RhoA or by inhibiting RhoA activity. While cingulin is restricted to tight junctions, paracingulin localizes to tight and adherens junctions and, therefore, may play a role in crosstalk at the apical junctional complex.

1.2.4.3 MUPP1 Multi-PDZ domain protein 1 (MUPP1) is a scaffolding protein localized at the tight junction cytoplasmic plaque. MUPP1 was originally identified as a binding partner for serotonin 5-hydroxytryptamine type 2 receptor (Ullmer et al., 1998) and then as an interaction partner of the C-terminus of Claudin-1 (Hamazaki et al., 2002). MUPP1 is a ~220 kDa protein with thirteen PDZ domains and a Lin 27 domain. While MUPP1 is not required for tight junction formation, it is an important scaffolding protein that links transmembrane and adaptor proteins at the tight junction cytoplasmic plaque (Adachi et al., 2009). MUPP1 interacts with tight junction transmembrane and adaptor proteins through its different PDZ domains such as claudins, JAM, and ZO-3 (Hamazaki et al., 2002; Jeansonne et al., 2003). In addition, MUPP1 interacts with components of adherens junctions (Adachi et al., 2009) and, thus, may play a role in crosstalk at the junctional complex.

1.2.5 Formation of tight junctions The transmembrane and cytoplasmic adaptor proteins of the tight junction regulate the structure and function of mature tight junctions. But how are tight junctions established? Tight

junction assembly is a multistep process that is closely linked to two evolutionarily conserved protein complexes, the Crumbs and partitioning defective (PAR) complexes, which drive the apical-basal polarization of epithelial cells.

1.2.5.1 Crumbs polarity complex The Crumbs complex is composed of three proteins: Crumbs (CRB), protein associated with Lin Seven 1 (PALS1), and PALS1-associated tight junction protein (PATJ). The Crumbs complex was first identified in Drosophila where it plays a key role in regulating apical-basal polarity (Campanale et al., 2017). The mammalian CRB3 protein, which contains a single transmembrane domain, is the main mammalian homolog that regulates apical-basal polarity and tight junction formation through its interactions with the PDZ domains of PALS1 and Par6 at the apical membrane (Lemmers et al., 2004; Roh et al., 2003). PALS1, the mammalian ortholog of Drosophila Stardust, is a member of the MAGUK family and is composed of two Lin 27 domains, one PDZ domain, an SH3 domain, a band 4.1-binding domain, and a GUK domain (Roh et al., 2002b). PALS1 mediates the indirect interaction between CRB3 and PATJ (Roh et al., 2002b). PATJ, the third member of the Crumbs complex and mammalian homologue of Drosophila discs lost, has one Lin 27 domain that interacts with PALS1 and 10 PDZ domains through which it interacts with transmembrane and adaptor proteins of the tight junction (Kamberov et al., 2000). PATJ is structurally similar to the adaptor protein MUPP1 and, thus, interacts with many of the same proteins as MUPP1 including ZO-3 and claudins (Roh et al., 2002a).

1.2.5.2 Par polarity complex The second polarity complex involved in tight junction assembly is the Par polarity complex which consists of Par3, Par6, and αPKC (Campanale et al., 2017). The Par proteins were first identified in C. elegans (Kemphues et al., 1988). They are evolutionarily conserved in structure, protein interactions, and their role in establishing apical-basal polarity. The Par proteins interact with each other as well as with the Rho GTPase Cdc42 to form a complex that regulates tight junction formation. Par6 binds to Par3, αPKC, Cdc42 (Qiu et al., 2000), and the Crumbs polarity complex proteins PALS1 and CRB3 (Hurd et al., 2003). Cdc42 and αPKC kinase activity are required for the association between Par6 and PALS1. Par3, like Par6, directly interacts with αPKC (Tabuse et al., 1998). These interactions link the Crumbs and Par

11 polarity complexes leading to establishment of apical-basal polarity. Once epithelial cells become polarized, tight junctions become positioned at the apical lateral membrane where they reinforce apical-basal polarity by preventing the mixing of apical and basolateral membrane proteins.

1.2.6 Regulation of tight junction assembly/disassembly Once formed, tight junctions are not static structures and are constantly being remodelled. Adaptor proteins at the tight junction cytoplasmic plaque interact with RhoGTPase regulators, guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs), to regulate tight junction assembly/disassembly (Quiros and Nusrat, 2014). Cingulin interacts with GEF-H1 sequestering it at tight junctions where GEF-H1 is inactive and is unable to activate RhoA leading to tight junction disassembly (Aijaz et al., 2005). Paracingulin works redundantly in GEF-H1 regulation (Guillemot et al., 2008a). In contrast, p114RhoGEF/ARHGEF18 is active at tight junctions where it drives RhoA signaling to promote tight junction formation. p114RhoGEF is recruited to tight junctions by cingulin (Terry et al., 2011). Paracingulin recruits Tiam1 to adherens junctions leading to increased Rac1 activity which, in turn, recruits membrane and cytoplasmic proteins required for tight junction formation (Guillemot et al., 2008a). ZO-1 recruits ARHGEF11/PDZ-RhoGEF to tight junctions to promote RhoA-driven tight junction formation (Itoh et al., 2012). The Cdc42 GAP Rich1 regulates tight junction stability by inhibiting Cdc42-mediated endocytosis of tight junction proteins. Rich1 is recruited to tight junctions by the scaffolding protein angiomotin (Amot) which, in turn, interacts with the Par polarity complex protein PALS1 at the tight junction cytoplasmic plaque (Wells et al., 2006). In summary, tight junction cytoplasmic proteins regulate tight junction assembly and maintenance by regulating the activity of RhoGTPases through interaction with RhoGEFs and GAPs.

1.2.7 Functions of tight junctions Classically, tight junctions are described as having two functions that resemble those of a gate and fence (Fig. 1.4). The gate function refers to the ability of tight junctions to regulate the paracellular movement of ions and small molecules by acting as semi-permeable diffusion barriers. The fence function refers to the ability of tight junctions to restrict the movement of lipids and proteins to apical and basolateral membrane compartments thus maintaining apical- basal polarity. New insight into tight junction function has challenged this traditional model,

revealing that tight junctions are much more functionally diverse and can also function as signaling platforms that guide cell behaviors including cell adhesion, proliferation and gene transcription. This section describes the roles of tight junctions.

1.2.7.1 The gate functions of tight junctions The gate function refers to the ability of tight junctions to regulate the passage of ions, small molecules and water through the paracellular space based on charge and size and can be assayed using two parameters: transepithelial electrical resistance (TER) and flux assays. TER is a measurement of electrical resistance across a cellular monolayer and reflects the paracellular resistance of the tight junction to the passage of an electric current. Therefore, the TER value reflects the barrier strength of the tight junction, which is determined by the combination of claudins expressed in a tissue or cell monolayer. While some claudins (Cldn2, -10, -15, -16, 17) form pores to permit the passage of ions (Alberini et al., 2017; Furuse et al., 2001; Hou et al., 2008; Van Itallie et al., 2006), small molecules and water, others act as barriers (Cldn1, -3, -4, -5, -8, -11, -14, -18, -19) (Angelow et al., 2006; Hou et al., 2008; Inai et al., 1999; Jovov et al., 2007; Van Itallie et al., 2001; Wen et al., 2004; Yu et al., 2003). The crucial role of claudins in regulating the gate function of tight junctions is highlighted by the difference in the TER values of the two strains of MDCK renal epithelial cells: low resistance MDCK II cells express Cldn2, a pore-forming claudin, while high resistance MDCK I cells do not express Claudin-2 (Stevenson et al., 1988). Exogenous expression of Cldn2 in MDCK I cells results in decreased TER with a value similar to MDCK II cells (Furuse et al., 2001). The difference in the barrier properties of these renal cell lines reflects what is observed in vivo in the kidney where the combination of claudins expressed across the different nephron segments determines the permeability properties of each segment (Hou et al., 2013). Flux assays measure the paracellular movement of tracer molecules with fluorescent (e.g. FITC-dextran) or radioactive labels (e.g. 3H-mannitol) to study the size selective properties of the tight junction. The paracellular movement of small uncharged molecules across the tight junction occurs through both tight junction pores formed by claudins and through breaks in tight junction strands, which occur as tight junctions are being remodelled (Guo et al., 2003; Rosenthal et al., 2017). The knowledge about the role of individual claudin family members in regulating paracellular size selectivity is limited. Overexpression of Cldn2 in cultured epithelial

cells increases the permeability for solutes smaller than 4 Å, whereas Cldn4 and -18 do not (Van Itallie et al., 2008b). The blood-brain-barrier of Cldn5-deficient mice shows increased permeability to small molecules (less than 800 D), but not larger molecules (Nitta et al., 2003). These data suggest that different claudins control the permeability for uncharged solutes.

1.2.7.2 The fence function of tight junctions The fence function refers to the ability of the tight junction to block the movement of membrane proteins and lipids between apical and basolateral domains, thereby maintaining apical-basal polarity. The Par and Crumbs polarity complexes, which establish apical-basal polarity, position the tight junction at the apical membrane. Once formed, the tight junction strands provide a physical barrier between the apical and basolateral membrane domains to prevent the lateral diffusion of proteins and lipids and maintain apical-basal polarity. Evidence of the fence function of tight junctions comes from cell culture experiments. Early studies showed that disruption of intercellular junctions results in mixing of membrane proteins initially restricted to apical and basolateral domains (Pisam and Ripoche, 1976; Ziomek et al., 1980). When a C-terminally truncated form of occludin is overexpressed in MDCK II cells, fluorescently labelled sphingomyelin added to the apical membrane is redistributed to the basolateral domain (Balda et al., 1996). Together, these data suggest that tight junctions contribute to separating the apical and basolateral membranes into distinct compartments. Whether individual claudin family members differentially regulate the fence function of tight junctions has not been explored. However, a crucial role for claudins in maintaining apical-basal polarity and epithelial identity is well-established. Apical-basal polarity of hepatocytes is disrupted in zebrafish Cldn15-like b mutant livers, (Cheung et al., 2012). In Xenopus Cldn6 morphant embryos, pronephric tubules exhibit defects in apical-basal polarity where the apical protein ezrin is expressed basolaterally and expression of the basolateral protein Na+K+ATPase is reduced.

1.2.7.3 Tight junctions and adhesion There is some evidence that tight junctions play a role in cell adhesion. In cell dissociation assays, L-fibroblasts exogenously expressing claudins take longer to dissociate into single cells compared to parental L-fibroblasts that do not have tight junctions (Kubota et al., 1999). In the converse experiment, L-fibroblasts transfected with claudins re-aggregate while parental L-

fibroblasts do not. A role for claudins in mediating adhesion has also been shown in vivo. In zebrafish, Claudin E is required to maintain tight junction contacts between the extra-embryonic yolk syncytial layer and the enveloping layer, which are undergoing extensive migratory movements called epiboly (Siddiqui et al., 2010). In the absence of cldne, initiation and progression of epiboly are delayed due to reduced tension between these cell layers. Further evidence that claudins affect cell adhesion in vivo comes from the observation that Xenopus blastomeres overexpressing the Xcla remain in tight clusters, while cells injected with a control mRNA dissociate and disperse over the surface of the blastula (Brizuela et al., 2001). These data demonstrate that tight junctions promote cell adhesion through the interaction of claudins between neighboring cells.

1.2.7.4 Tight junctions and cell proliferation and gene expression Roles for tight junctions in mediating cell proliferation and gene transcription are mediated by the ZO proteins, which have nuclear localization (NLS) and nuclear export (NES) signals that allow them to shuttle between the nucleus and tight junction. ZO-1 regulates cell proliferation by binding to the Y-box transcription factor ZONAB, which, interacts with the cell cycle regulators CDK4, cyclin D1 and proliferating cell nuclear antigen (PCNA). These interactions are regulated by cell density. Low density cells express low levels of ZO-1, allowing ZONAB- CDK4 complexes to accumulate in the nucleus and promote G1/S phase transition (Balda et al., 2003). In a separate pathway, nuclear ZONAB acts a transcription factor to upregulate the expression of cyclin D1 and PCNA (Sourisseau et al., 2006). In a confluent cell monolayer, ZO- 1 inhibits ZONAB by sequestering it at the tight junction cytoplasmic plaque. ZO-2 affects cell proliferation by interacting with c-myc to regulate the transcription of cyclin D1 or through interactions with the transcription factors Jun and Fos to regulate transcription of genes with AP- 1 sites in their promoter (Balda and Matter, 2009). These diverse interactions suggest ZO proteins regulate cell proliferation.

In summary, numerous studies conducted over the last 50 years have revealed that tight junctions are not simple barriers but are composed of a network of transmembrane and cytosplasmic proteins that act as a signaling platform to regulate paracellular permeability, apical-basal polarity, cell adhesion, gene expression, and cell proliferation. At the center of this

signaling hub are claudins, which are the building blocks of tight junction strands and regulate most of the tight junction functions described above.

1.3 THE CLAUDIN FAMILY OF TIGHT JUNCTION PROTEINS

1.3.1 Identification of claudins Occludin was the first tight junction transmembrane protein identified, but it was unable to form a continuous network of tight junction strands when transfected into L-fibroblasts. This led to the idea that there must be another unidentified integral transmembrane protein. Claudins (Cldns) were identified by re-examining the junction fraction of chicken liver used to isolate occludin (Furuse et al., 1998b). Peptide sequencing identified two tetraspan transmembrane proteins designated Cldn1 and -2 that showed no sequence homology to occludin, but co- localized with occludin at the tight junction when transfected into MDCK cells. Claudins form a well-developed network of tight junction strands when transfected into L-fibroblasts suggesting that they are the main structural components of tight junction strands (Furuse et al., 1999). There have now been ~24 claudin family members identified in vertebrates.

1.3.2 Structure of claudins Claudins are 20-27kDa tetraspan transmembrane proteins with two extracellular loops and cytoplasmic N- and C-termini (Fig. 1.5). Although little is known about the function of the short, intracellular N-terminal domain of 2-7 amino acids or the short intracellular loop, specific functions have been ascribed to the other domains. The first extracellular loop is ~50 amino acids in length and varies considerably in amino acid sequence. The combination of charged amino acids in the first extracellular loop determines the ion selectivity of each claudin molecule (Piontek et al., 2017b). Based on their ability to regulate the movement of cation and anions, claudins have been categorized as cation barriers (Cldn1, -5, -8, -11, -14, -19), cation pores (Cldn2, -10b, -15, 16), anion barriers (Cldn6 and -9) or anion pores (claudin-10a) (González- Mariscal et al., 2012). The first extracellular loop also contains a conserved W-GLW-C-C motif, where the two cysteine residues are thought to form a disulfide bond that stabilizes the conformation of the loop. The smaller second extracellular loop (~16-33 amino acids) participates in trans interactions of claudin molecules between apposing cells (Piontek et al., 2008) and in cis interactions of claudins within the same cell (Suzuki et al., 2014). In addition,

16 the second extracellular loop of some family members acts as a receptor for Clostridium perfringens enterotoxin (CPE). The transmembrane domains are involved in cis-interactions of neighboring claudins within a cell and are required for the correct folding and assembly of claudin proteins (Rossa et al., 2014). The C-terminal domain shows the most sequence and size heterogeneity between the different claudin species. However, in all claudins, except for Cldn12 and -23, the final two amino acids form a conserved PDZ binding motif that interacts with the PDZ domain of tight junction cytoplasmic proteins. These interactions link claudins to the actin cytoskeleton (Hamazaki et al., 2002; Itoh et al., 1999a; Jeansonne et al., 2003) and to intracellular signaling events that affect cell polarity, cell proliferation and differentiation, and morphogenesis (Fredriksson et al., 2015). Compared to the other domains of the claudin protein, the C-terminus is enriched in serine, threonine and tyrosine residues which undergo post-translational modifications that affect targeting of the claudin protein to the tight junction complex (D'Souza et al., 2005; Ikari et al., 2006), protein stability (Van Itallie et al., 2004), paracellular barrier properties of the tight junction (Fujibe et al., 2004; Soma et al., 2004), and interaction of claudins with cytoplasmic adaptor proteins (Nunbhakdi-Craig et al., 2002). Palmitoylation of cysteine residues in the C-terminal tail of claudins is also required for efficient targeting of claudins to the membrane (Van Itallie et al., 2005).

1.3.3 Claudin-claudin interactions Claudins participate in homophilic interactions with like family members and heterophilic interactions between different claudin family members. These interactions can occur in cis or in trans. Most epithelial cells express a combination of different claudins forming heteropolymeric tight junction strands. However, not all claudins are compatible with each other. For example, Cldn1 can interact with Cldn3 in trans to form paired strands, but not with Cldn2 (Furuse et al., 1999). Cldn4 can interact in cis with Cldn3 and -8, but does not interact with Cldn7 or itself (Gong et al., 2015b). Furthermore, heteromeric and heterotypic interactions occur through distinct mechanisms. For example, Cldn3 can form cis, but not trans, interactions with Cldn4 (Daugherty et al., 2007). Claudin-claudin interactions have been studied for only a limited subset of family members. Interactions of the transmembrane domains form the backbone of tight junction strands (Milatz et al., 2015; Rossa et al., 2014), while homotypic and heterotypic

trans interactions of the extracellular loops between claudins in adjacent cells form paired strands that contribute to the paracellular barrier (Colegio et al., 2003; Piontek et al., 2017b). The interaction between claudin family members is also critical for their trafficking and integration into tight junction strands. For example, interactions between Cldn4 and -8 enhance their integration into tight junction strands. Cldn8 shows reduced localization to tight junctions in the kidney of Cldn4 knockout mice (Fujita et al., 2012) and Cldn4 localization to the tight junction is lost in Cldn8 knockout mouse kidneys (Gong et al., 2015b). Heteromeric interactions between Cldn16 and -19 via their third and fourth transmembrane domains are required for their assembly into tight junction strands (Gong et al., 2015a; Hou et al., 2009). Cldn16 localization to tight junctions is lost in Cldn19 knockdown mouse kidneys and Cldn19 shows reduced localization to tight junctions in Cldn16 knockdown mouse kidneys. In summary, the ability of claudins to interact with other family members, both within a cell and between adjacent cells, provides the opportunity to generate tight junctions that are functionally diverse.

1.3.4 Post-translational modifications Two types of post-translation modifications have been reported to affect the localization and permeability properties of claudins: phosphorylation and palmitoylation.

1.3.4.1 Phosphorylation A number of studies have demonstrated that phosphorylation of serine, threonine, and tyrosine residues in the C-terminal cytoplasmic tail of claudins can regulate tight junction assembly/disassembly and paracellular permeability (D'Souza et al., 2005; Ikari et al., 2006). Phosphorylation of Cldn5 by myosin light chain kinase (MLCK) and rho kinase (ROCK), downstream effectors of RhoA signaling, is associated with tight junction disassembly and increased paracellular permeability. During brain infection, ROCK-mediated phosphorylation of Cldn5 disrupts the integrity of the blood-brain barrier allowing the passage of monocytes and ultimately leading to inflammation of the brain (Persidsky et al., 2006; Yamamoto et al., 2008). Phosphorylation of Cldn5 by MLCK during chronic alcohol abuse contributes to tight junction disassembly in brain endothelial cells resulting in enhanced paracellular permeability (Haorah et al., 2005). Phosphorylation of claudins by ephrin signaling can also affect tight junction integrity. EphA2 binds to the first extracellular loop of Cldn4. This assocication leads to phosphorylation of the C-terminal tail of Cldn4, which disrupts the association of Cldn4 with

ZO-1 and this leads to reduced Cldn4 localization to tight junctions and enhanced paracellular permeability (Tanaka et al., 2005a). Signaling between claudins and ephrins seems to be bidirectional as claudins can also affect the phosphorylation of ephrin ligands. The cis interaction between Cldn1 or -4 and ephrin-B1 leads to increased phosphorylation of ephrinB1 (Tanaka et al., 2005b). In summary, numerous studies have shown that phosphorylation of claudin tails affects claudin membrane localization and the barrier properties of the tight junction. While these studies have looked at individual claudins, residues that are conserved across family members have the potential to be regulated in a similar manner. However, most of these studies are conducted in cell lines and it remains to be determined whether claudins are regulated in the same way in a tissue-specific manner in vivo.

1.3.4.2 Palmitoylation Palmitoylation involves the addition of acyl groups to cysteine residues by palmitic acid and is known to affect protein trafficking and stability of transmembrane proteins. Cldn14 is the only claudin that has been shown to be palmitoylated on cysteines proximal to both the second and fourth transmembrane domains. Modification of these residues is required for efficient localization of Cldn14 to tight junctions and proper barrier formation; mutation of these cysteines to serines results in decreased TER and increased accumulation of Cldn14 in lysosomes (Van Itallie et al., 2005). The di-cysteine palmitoylation motifs found in Cldn14 are conserved throughout the claudin family, thus it is likely that most claudins are palmitoylated at these residues.

1.3.5 Genomic organization and evolutionary conservation of claudins Tight junctions are considered a unique feature of vertebrates. However, many of the protein components of the tight junction predate the origin of the vertebrate tight junction. The claudin family of tight junction proteins is one such example as claudins appear to have arisen prior to the development of chordates. Consistent with this hypothesis, claudin-like molecules have been identified in sponges (e.g. Amphimedon queenslandica) and lower chordates (e.g. the ascidian Halocynthia roretzi) (Lal-Nag and Morin, 2009; Mukendi et al., 2016). C. elegans has five claudins (clc-1 to -5) and five claudin-like molecules (nsy-1 to -4 and vab-9). These

proteins are functionally similar to vertebrate claudins regulating cell adhesion and paracellular permeability; however, they localize to adherens junctions (Simske, 2013; Simske and Hardin, 2011; Simske et al., 2003). Claudin-like molecules are also present in the septate junctions of invertebrates, the functional equivalent to the vertebrate tight junction that lie immediately basal to adherens junctions. The Drosophila septate junction contains three claudin-like molecules, Sinuous, Kune-kune, and Megatrachea that regulate the barrier function of septate junctions, but not apical-basal polarity (Behr et al., 2003; Nelson et al., 2010; Wu et al., 2004). As with most vertebrate claudins, the C-terminal domains of these claudin-like molecules have a PDZ binding domain that interacts with PDZ domain adaptor proteins bridging septate junctions to the actin cytoskeleton. Through these interactions, these claudin-like proteins affect morphogenesis of the trachea regulating its size and shape (Behr et al., 2003; Nelson et al., 2010; Wu et al., 2004). Thus, although the relative position of claudin and claudin-like molecules within the membrane is not conserved, they share a functional role in mediating epithelial permeability, cell adhesion and interactions with the actin cytoskeleton. The appearance of tight junctions in vertebrates was accompanied by the expansion of the claudin gene family. There are 20 claudins in Xenopus (Baltzegar et al., 2013), 19 in the chick (Collins et al., 2013; Consortium, 2004), and ~24 family members in mammals (Lal-Nag and Morin, 2009). Interestingly, the genome of teleosts has undergone extensive expansion such that 56 claudin family members have been identified in the puffer fish Takifugu (Loh et al., 2004) and 54 claudins are present in the zebrafish genome (Baltzegar et al., 2013). Several pairs of highly homologous CLDN genes (CLDN3 and 4, CLDN6 and 9, CLDN8 and 17, and CLDN22 and 24) are located in close proximity in the vertebrate genome (Collins et al., 2013; Lal-Nag and Morin, 2009). These data suggest that gene duplication events have contributed to the expansion and evolution of the claudin family. Whether this genomic arrangement means that claudins share regulatory elements that allow their expression to be coordinated is unknown. Sequence analysis of claudins led to the subdivision of the claudin family into two groups designated as classic claudins (Cldn1 to -10, -14, 15, -17, and -19) or non-classic claudins (Cldn11 to -13, -16, -18, and -20 to -24) based on sequence homology (Krause et al., 2008). The classic claudins contain conserved residues in their second extracellular loop that are important for protein folding and the trans-interaction of claudin family members between adjacent cells (Krause et al., 2008). The sequence of the first extracellular loop of these claudins is also highly

homologous, however, these claudins differ in their barrier properties with some family members functioning as barriers and others as pores. The classic and non-classic claudins also differ in their tissue-specific expression profiles.

1.3.6 Physiological and developmental roles for claudins Generation of knockout, knockdown, and transgenic animal models together with identification of human mutations in CLDN genes have revealed that claudins play a critical role in regulating the tight junction barrier, apical-basal polarity, cell adhesion, and changes in cell shape.

1.3.6.1 Claudin gene deletion models in mouse and mutations in human disease Knockout mice have been generated for 15 of the 27 mouse CLDN genes. While some mice are complete knockouts, others are tissue-specific. In the few examples where CLDN mutations have been identified in humans, the phenotypes observed in the knockout mouse models phenocopy what is seen in human patients confirming that the function of claudins is evolutionarily conserved. These are described below and summarized in Table 1.1.

1.3.6.1.1 Cldn1 and -6 create a water barrier in the skin The skin forms a barrier against pathogens, chemical and physical injury, and prevents excessive water loss that can lead to dehydration. Cldn1 and -6 regulate the skin barrier in mice. Cldn1 knockout mice have wrinkled skin and die within 1 day of birth from dehydration due to transepidermal water loss (Furuse et al., 2002). While Cldn6 knockout mice are phenotypically normal (Anderson et al., 2008), transgenic mice overexpressing Cldn6 in the epidermis have a severe skin phenotype including a defect in the epidermal barrier and terminal differentiation (Turksen and Troy, 2002). Together, these data suggest that the balance of claudins is critical for regulating the skin epidermal barrier. CLDN1 plays a conserved role in humans to regulate the skin barrier (Feldmeyer et al., 2006; Hadj-Rabia et al., 2004; Kirchmeier et al., 2014). Mutations in the CLDN1 gene cause neonatal ichthyosis and sclerosing cholangitis, an autosomal recessive disorder associated with scaling of the skin and progressive scarring and obstruction of the bile duct. To date, no human mutations in CLDN6 have been identified.

1.3.6.1.2 Cldn5 is essential for the blood-brain barrier Cldn5 and -12 are the only claudins expressed in the blood-brain-brain barrier. Knockout of Cldn5 in mice results in a selective increase in paracellular permeability of the blood-brain- barrier to small molecules (Nitta et al., 2003). To date, a Cldn12 knockout mouse has not been reported. Thus, it is possible that Cldn12 regulates paracellular permeability of larger molecules. No human mutations in CLDN5 and 12 have been reported to date.

1.3.6.1.3 Cldn9, -11, and -14 in the inner ear prevent deafness The cochlea of the inner ear is separated into two distinct compartments containing extracellular fluids with different ionic compositions by the stria vascularis: the endolymph (high K+, low Na+) and the perilymph (low K+, high Na+). This ion gradient generates a voltage potential that is critical for hearing. Cldn9, -11, and -14 are expressed in tight junctions of the stria vascularis where they play a critical role in its barrier function (Ben-Yosef et al., 2003; Gow et al., 2004; Kitajiri et al., 2004; Nakano et al., 2009). A Cldn9 ENU-induced mouse mutant exhibits deafness due to loss of sensory hair cells, which is caused by increased K+ in the perilymph (Nakano et al., 2009). Cldn11 knockout mice also suffer from deafness due to a decreased endocochlear potential caused by a defective paracellular barrier (Gow et al., 2004; Kitajiri et al., 2004). Cldn14 is expressed in inner and outer hair cells and in tight junctions of the stria vascularis. Like humans with mutations in CLDN14 (Bashir et al., 2013; Wattenhofer et al., 2005; Wilcox et al., 2001), Cldn14 null mice are deaf due to degeneration of outer hair cells (Ben-Yosef et al., 2003).

1.3.6.1.4 Cldn11 and -19 are required in the nervous system for nerve conduction Myelin sheaths wrap around axons to act as an insulator to facilitate axonal conduction (Nave and Werner, 2014). They are formed by oligodendrocytes in the central nervous system and Schwann cells in the peripheral nervous system. Cldn11, originally designated oligodendrocyte-specific protein (OSP), is expressed in myelin sheaths generated by oligodendrocytes of the central nervous system and Cldn19 is expressed in myelin sheaths generated by Schwann cells of the peripheral nervous system, where they function as cation barriers to prevent the leakage of ions and maintain nerve conduction (Denninger et al., 2015; Devaux and Gow, 2008; Gow et al., 1999; Miyamoto et al., 2005). Cldn11 and -19 null mice

exhibit similar neurological deficits including delayed nerve conductance, but the overall morphology of myelinating cells is unaffected.

1.3.6.1.5 Cldn2, -4, -7, -8, -10, -16, and -19 regulate paracellular permeability in the kidney The unique combination of claudins expressed along the different nephron segments determines the paracellular permeability properties of that segment. This idea is supported by studies in claudin knockout mice in which the main phenotype is wasting of ions in the urine as a result of a tight junction barrier defect in the kidney. Cldn2 knockout mice exhibit decreased Na+, Cl-, and water reabsorption in the renal proximal tubules (Muto et al., 2010). Cldn4 knockout mice display increased urinary excretion of Ca2+ and Cl- and urothelial hyperplasia, which causes urinary tract obstruction that leads to hydronephrosis (Fujita et al., 2012). Cldn7 null mice die from dehydration due to decreased paracellular Cl- permeability; Na+ and K+ paracellular permeability are also decreased (Tatum et al., 2010). Deletion of Cldn8 in the collecting duct of the kidney causes renal salt wasting defects including increased urinary excretion of Na+, Cl-, and K+ (Gong et al., 2015b). Cldn10, -16 and -19 regulate cation reabsorption in the thick ascending limb of the loop of Henle: Cldn10 forms a Na+ channel, while Cldn16 and -19 regulate reabsorption of Ca2+ and Mg2+. Deletion of Cldn10 in the thick ascending limb leads to decreased paracellular permeability to Na+, hypermagnesemia (elevated blood magnesium levels), and nephrocalcinosis (calcium deposits in the kidney) (Breiderhoff et al., 2012). Knockdown or deletion of Cldn16 leads to Ca2+ and Mg2+ wasting and nephrocalcinosis (Hou et al., 2007; Will et al., 2010). Similarly, siRNA knockdown of Cldn19 results in Ca2+ and Mg2+ wasting (Hou et al., 2009). The renal abnormalities observed in the Cldn10, -16 and -19 deficient mice are similar to those observed in patients with mutations in CLDN10B, CLDN16 and CLDN19 suggesting that these claudins play a conserved role in regulating paracellular cation permeability in the kidney (Al-Haggar et al., 2009; Hadj-Rabia et al., 2017; Klar et al., 2017; Milatz et al., 2017; Weber et al., 2001; Yamaguti et al., 2015; Yuan et al., 2015). Patients with mutations in CLDN19 also have ocular abnormalities such as myopia, nystagmus, and macular coloboma.

1.3.6.1.6 Cldn2, -7, -15 and -18 regulate paracellular permeability in the gastrointestinal tract In the gastrointestinal tract, the claudin-based epithelial barrier functions to absorb nutrients, ions and fluids and to prevent infection (Lameris et al., 2013). Cldn2, -7, and -15 regulate paracellular permeability in the intestine, while Cldn18 regulates the tight junction barrier in the stomach. The role of Cldn2 and -15 in the intestine has been examined in single and double knockout mice. Individual knockout mice for Cldn2 and -15 show reduced Na+ permeability (Tamura et al., 2011; Tamura et al., 2008). Cldn15, but not Cldn2, null mice also show reduced glucose absorption, which is accompanied by an enlarged small intestine designated the ‘megaintestine phenotype’. These studies suggest that Cldn2 and -15 play both unique and functionally redundant roles in regulating paracellular permeability in the intestine. Both claudins form Na+-selective permeable pores, while only Cldn15 regulates Na+-driven glucose absorption. The double knockout mice exhibit a more severe paracellular permeability defect: they die from malnutrition at postnatal day 25 due to a deficiency in sodium-driven nutrient absorption (Wada et al., 2013). Cldn7 knockout mice also show increased intestinal paracellular permeability to small solutes leading to inflammation and epithelial cell death (Ding et al., 2012; Tanaka et al., 2015). Knockout of Cldn18 causes increased paracellular leakage of H+ ions in the stomach leading to prolonged inflammation and subsequent development of gastritis (Hayashi et al., 2012). Together, these data suggest that claudins are required to maintain the epithelial barrier in the gastrointestinal tract. To date, no mutations or polymorphisms in CLDNs have been associated with increased susceptibility to inflammatory bowel diseases (IBDs) in humans. However, downregulation of Cldn1, -3, -4, -5 and -8 and upregulation of Cldn2 has been observed in IBD patients (Zeissig et al., 2007). These data suggest that misregulation of claudin expression may contribute to the disease phenotype and that claudins play an important role in regulating paracellular permeability in the gastrointestinal tract of humans.

1.3.6.1.7 Cldn2 plays a role in the hepatobiliary system Bile formation and secretion are essential functions of the hepatobiliary system, which consists of the liver, gall bladder and bile ducts (Boyer, 2013). Bile is produced by the liver and stored in the gallbladder. After eating, bile ducts carry bile to the small intestine where it helps

digest fat and get rid of toxins Bile is composed of cholesterol, phospholipids, and bile acids and an imbalance in the components of bile can lead to formation of gallstones. Production of bile involves the secretion of water and solutes through the paracellular and transcellular pathways. Cldn1, -2 and -3 are the main claudins expressed in the liver. In Cldn2 knockout mice, the concentrations of bile acids, phospholipids, and cholesterol are increased, while water permeability and the rate of bile flow are decreased (Matsumoto et al., 2014). All Cldn2 null mice develop gallstones when placed on a lithogenic diet that promotes formation of gallstones. These data suggest that Cldn2 forms a water and ion permeable pore in the liver to regulate the composition and flow of bile. Patients with CLDN1 mutations exhibit bile leakage due to a tight junction barrier defect that leads to development of neonatal ichthyosis-sclerosing cholangitis syndrome (Feldmeyer et al., 2006; Hadj-Rabia et al., 2004; Kirchmeier et al., 2014). The early lethality of Cldn1 knockout mice means that the functional requirement of Cldn1 in hepatic paracellular permeability cannot be assessed (Furuse et al., 2002).

1.3.6.1.8 Cldn4 and 18 regulate the lung epithelial barrier The alveolar epithelial barrier prevents leakage of fluid into the airspaces of the lung. Cldn3, -4, -7 and -18 are the most highly expressed claudins in alveolar epithelial cells (Kaarteenaho et al., 2010). Cldn18 null mice exhibit increased solute permeability and alveolar fluid clearance (Li et al., 2014). Cldn3 and -4 are upregulated but are unable to compensate for the loss of Cldn18. Cldn4 knockout mice exhibit increased permeability to solutes and decreased fluid clearance, but expression of Cldn3 and -18 is unchanged (Kage et al., 2014; Wray et al., 2009).

1.3.6.1.9 Cldn11 is required for spermatogenesis The blood-testis-barrier (BTB) divides the seminiferous epithelium into the basal and adluminal compartments (Stanton, 2016). It is formed by tight junctions between Sertoli cells where Cldn3 and -11 are the main claudins in mouse and humans. As germ cells differentiate, they pass through the BTB as they move from the basal to the adluminal compartment. For this to occur, Sertoli-germ cell and Sertoli-Sertoli cell junctions must be remodelled. While Cldn11 is continuously expressed in tight junctions of the BTB, Cldn3 is transiently expressed in newly formed tight junctions. Cldn3 is also expressed in migrating germ cells located in the basal compartment. While Cldn3 knockout mice are viable and fertile, Cldn11 null mice are sterile

due to disorganization of Sertoli cells, which detach from the basement membrane and change cell shape resulting in apoptosis of germ cells and ultimately leading to failure of spermatogenesis (Gow et al., 1999; Mazaud-Guittot et al., 2010). Furthermore, Cldn11 is mislocalized in the BTB of infertile men (Chiba et al., 2012). These data indicate that Cldn11 plays an important role in establishing a microenvironment for germ cells.

1.3.6.2 Claudin function in adhesion and cell shape changes during embryonic morphogenesis Claudins have been extensively studied for their tight junction barrier function in epithelial cell layers of adult tissues. More recently, studies in the embryo have highlighted their importance in maintaining cell contacts during the movement and extension of cell layers and in regulating changes in cell shape during morphogenesis. Studies in zebrafish and Xenopus have shown that claudins can act as adhesion molecules to maintain cell contacts during epiboly, a developmental process in which a multilayered cell sheet spreads and thins. In zebrafish, Cldn E is required to maintain tight junction contacts between the extra-embryonic yolk syncytial layer and the enveloping layer (Siddiqui et al., 2010). In the absence of cldne, initiation and progression of epiboly are delayed due to reduced tension between these cell layers. In Xenopus, overexpression of the claudin Xcla alters the cell adhesion properties of blastomeres (Brizuela et al., 2001). Blastomeres overexpressing Xcla remain in tight clusters and fail to undergo epiboly, while cells injected with a control mRNA dissociate and disperse over the surface of the blastula. These data demonstrate a role for tight junctions in cell adhesion through the interaction of claudins with neighboring cells. The Drosophila trachea is a branched network of epithelial tubes that functions as a combined pulomonary and vascular system. The three Drosophila claudin orthologs, Sinuous, Kune-kune, and Megatrachea, have been shown to regulate tracheal tube morphogenesis (Behr et al., 2003; Nelson et al., 2010; Wu et al., 2004). In all three Drosophila claudin mutant embryos, the trachea is elongated and tortuous due to a change in the characteristic cuboidal shape of tracheal cells. These claudin-like molecules may regulate the shape of individual tracheal cells through indirect interactions with the actin cytoskeleton through PDZ-domain adaptor proteins. In summary, studies in invertebrate species have shown that claudins play key roles in morphogenesis during embryonic development regulating cell adhesion and changes in cell

shape. However, even though claudins are expressed in developing epithelial tissues in vertebrates, the phenotypes observed in knockout and knockdown mice mainly affect adult tissues and are frequently due to paracellular permeability defects. The lack of embryonic phenotypes observed in claudin-depleted mice could be attributed to the fact that the claudin family has undergone extensive expansion in vertebrates and that some claudins are functionally redundant during early stages of embryonic development. Tools that target multiple claudins family members, such as the C-terminal domain of Clostridium perfingens enterotoxin (C-CPE), may identify morphogenetic events that are regulated by functionally redundant claudins but hidden when claudins are studied individually.

1.3.7 Clostridium perfringens enterotoxin: a tool to study claudins The ~35kDa Clostridium perfringens enterotoxin (CPE) is a single chain 319 amino acid polypeptide that causes the symptoms associated with C. perfringens food poisoning in humans (Wieckowski et al., 1994). The action of CPE starts with its binding to claudin receptors on host cells. Once bound, CPE forms a series of complexes ultimately leading to formation of a β- barrel that inserts into the membrane, creating a cation permeable pore. Calcium influx through this pore, which is composed of at least six CPE molecules, CPE-binding and non-CPE-binding claudins and occludin, causes cell death by triggering calmodulin or calpain-induced oncosis or caspase-3-mediated apoptosis (Chakrabarti et al., 2003; Robertson et al., 2007).

1.3.7.1 The interaction between CPE and claudins Extensive screens to identify the CPE receptor led to the cloning of Cldn4, which was initially named CPE Receptor (CPE-R) (Katahira et al., 1997a). Subsequent studies revealed that Cldn3, -6, -7, -8 and -14 also act as receptors for CPE or the C-terminal CPE (C-CPE) receptor-binding domain. Amongst the claudins receptors, not all bind CPE with equal affinity. Cldn3, -4, -6, and -7 are high affinity CPE receptors with association constants of 8.4 x 107 M-1, 1.1 x 108 M-1, 9.7 x 107 M-1 and 8.8 x 107 M-1, respectively, while Cldn8 and -14 bind CPE more weakly with association constants of 1 x 106 M-1 and 3.6 x 106 M-1, respectively (Fujita et al., 2000; Sonoda et al., 1999). A study conducted by Winkler et al. suggests that Cldn9 can also bind to CPE (Winkler et al., 2009) and work done by Kimura et al. showed that Cldn1 and -2 interact with CPE with a much lower affinity than claudins classified as CPE-sensitive (Kimura et al., 2010). Cldn17, -21, -22, and -24 have not been tested for their ability to bind to CPE.

All claudins that act as CPE receptors belong to the group of classic claudins supporting the idea that shared homology between these claudins contributes to their ability to bind to CPE. Mutagenesis studies revealed a consensus CPE-binding motif, NP(L/M)(V/L/T)(P/A), in the second extracellular loop of claudins that bind to CPE (Winkler et al., 2009). An alternate hypothesis, the electrostatic attraction model, proposes that the last 12 amino acids of the second extracellular loop of CPE-sensitive claudins are enriched in positively-charged basic amino acids that are attracted to the negatively-charged acidic cleft of CPE (Kimura et al., 2010). Alignment of the amino acid sequence of the second extracellular loop of claudins revealed that the pentapeptide motif NP(L/M)(V/L/T)(P/A) is not conserved among all CPE-binding claudins (Table 1.2) and, therefore, the second hypothesis is assumed correct. The observed discrepancies concerning which claudin family members bind to CPE may be due to differences in the methods used to assay for CPE binding: dot blot assays, pull-downs, cytotoxic assays and cellular binding assays. Dot blot assays use synthetic peptides corresponding to the second extracellular loop of claudins (Winkler et al., 2009). A major drawback to this method is that the influence of the loop structure of claudins on the interaction between claudins and CPE is not taken into account. For example, Cldn9, but not Cldn4, binds to C-CPE using this method (Winkler et al., 2009). In contrast, pull-down assays assess the interaction between C-CPE and full-length claudins using transiently transfected cell lines or epithelial cell lines expressing endogenous claudins (Harada et al., 2007; Lohrberg et al., 2009; Winkler et al., 2009). Transfection of a single claudin assesses the ability of individual claudins to bind to CPE, but transfected claudins often do not localize to the membrane. Claudins are examined in their native environment in epithelial cell lines, but CPE-insensitive claudins that form heterodimers with CPE-sensitive claudins may also be pulled down. For example, non- CPE-binding Cldn5 can interact with C-CPE using this method (Lohrberg et al., 2009). In cytotoxic assays, either full-length CPE or C-CPE fused to a toxic effector domain of another protein is incubated with cells either expressing endogenous or transfected claudins and cell death is used as a measure of CPE-binding (Ebihara et al., 2006; Kimura et al., 2010; Robertson et al., 2010). The cellular binding assay investigates the interaction of C-CPE with full-length claudins in their native membrane environment and is the most relevant method to assess C- CPE-claudin interactions. Cells expressing endogenous or transfected claudins are cultured with C-CPE and immunofluorescence is used to demonstrate that claudins are removed from the

28 membrane (Sonoda et al., 1999) or C-CPE binds to cells expressing specific claudins (Winkler et al., 2009). This is the method used to confirm that C-CPE can selectively remove claudins from tight junctions of cultured chick embryos.

1.3.7.2 The C-terminal claudin-binding domain of CPE Structure/function analysis of CPE revealed that it has two functional domains: the N- terminus mediates its cytotoxic effects through oligomerization, membrane insertion and pore formation and the C-terminal 30 amino acids have receptor-binding activity (Hanna and McClane, 1991; Hanna et al., 1991; Hanna et al., 1989; Kokai-Kun and McClane, 1997; Singh et al., 2000; Smedley and McClane, 2004). While the C-terminal amino acids 290-319 are sufficient for binding to claudins, longer CPE fragments have a higher affinity for claudins. In 2008, the crystal structure of C-CPE from amino acids 194-319 was solved revealing that it is composed of a nine-strand β-sandwich (Van Itallie et al., 2008a). Located at the center of opposing β strands 8 and 9 of C-CPE, the claudin binding site is enriched in negatively charged residues that interact with positively charged residues in the second extracellular loop of claudins. CPE peptides consisting of amino acids 290-319 do not have the same conformation as those containing the full β-sandwich. This, at least partially, explains why C-CPE290-319 binds more weakly to claudins than longer C-CPE fragments. Functional studies in epithelial cell lines revealed that C-CPE is not cytotoxic (Horiguchi et al., 1987), but can bind to a specific subset of claudins removing them from tight junctions (Lohrberg et al., 2009; Moriwaki et al., 2007; Sonoda et al., 1999). Researchers have taken advantage C-CPE’s ability to simultaneously remove multiple claudins from tight junctions to study the functionally redundant role of claudins in cell lines and animal models (Moriwaki et al., 2007; Sonoda et al., 1999). I am using the C-CPE reagent to investigate the role of claudins during neural tube closure in the chick embryo. We previously showed that fourteen claudins are expressed during neural tube closure (Collins et al., 2013), but none of the reported single claudin knockout mice exhibit neural tube defects (Table 1.1). The C-CPE reagent allowed me to investigate whether members of the claudin family play functionally redundant roles in neural tube closure.

1.4 NEURAL TUBE FORMATION The brain and spinal cord are derived from the neural tube, a structure that is formed by a coordinated sequence of morphogenetic events during embryogenesis. Neurulation, the developmental process that gives rise to the neural tube, occurs in two stages (Nikolopoulou et al., 2017). Primary neurulation forms the portion of the neural tube that gives rise to the brain and most of the spinal cord. This process involves converting the neural plate, a single-layered epithelium on the dorsal surface of the embryo, into a closed elongated tube. Following completion of primary neurulation, the posterior neural tube is formed by secondary neurulation. Secondary neurulation involves the condensation of mesenchymal cells in the tail bud of the embryo to form an epithelial rod. Subsequent canalization of this rod results in formation of a lumen that is continuous with that of the primary neural tube. Both primary and secondary neurulation are required to generate the complete neural tube of mammals, amphibians and birds.

1.4.1 Phases of neural tube closure Primary neurulation, which is the stage of neurulation being studied in this thesis, is divided into four phases: neural plate formation, neural plate shaping, bending and elevation of the neural folds, and fusion of the neural folds to form a closed neural tube (Fig. 1.6). These phases of neurulation overlap in time and space and must be tightly coordinated. A defect in any of these events will results in a neural tube defect. The cell behaviors and signaling pathways regulating these events are described below.

1.4.1.1 Phase I: Formation of the neural plate At the end of gastrulation, a flat epithelial sheet called the ectoderm covers the dorsal surface of the embryo. Neurulation begins with induction of the neural plate (Stern, 2005). In the absence of any signal, the ectoderm will acquire a neural fate. In the early embryo, bone morphogenetic proteins (BMPs) are expressed throughout the ectoderm and neural induction requires a blockade of BMP signaling (Hawley et al., 1995; Hemmati-Brivanlou and Melton, 1997). The organizer secretes three BMP antagonists, chordin, noggin and follistatin, inhibiting BMP signaling and allowing ectoderm cells to acquire a neural fate. BMP antagonists are sufficient to induce a neural fate in amphibians and teleosts (Khokha et al., 2005). However, in chick and mouse embryos, fibroblast growth factor (FGF) signaling is also required to inhibit an epidermal fate (Rodriguez-Gallardo et al., 1997; Rogers et al., 2011; Streit et al., 2000; Streit et

al., 1998). Aside from neural induction, neural plate formation involves apical-basal thickening and pseudostratification of the ectoderm (Schoenwolf and Powers, 1987). By the end of this phase of neurulation, the disc-shaped neural plate consists of a pseudostratified, columnar epithelium in which neuroepithelial cells have an inherent apical-basal polarity, are connected by cell-cell junctions on their apical side and express unique molecular markers. Maintaining apical-basal polarity of neural plate cells is critical for neurulation because neural tube morphogenesis depends on apical-basal events (e.g. apical constriction, basal nuclear migration, apical-basal height) and when it is disrupted embryos develop neural tube defects (Chen et al., 2017; Eom et al., 2011).

1.4.1.2 Phase II: Shaping the neural plate The initially flat and broad neural plate thickens along the apical-basal axis, narrows medial-laterally and ultimately extends along the anterior-posterior axis. Tissue isolation experiments revealed that neural plate shaping is regulated by changes in the behavior of neural ectoderm cells: apical-basal elongation, convergent extension and oriented cell division (Moury and Schoenwolf, 1995; Schoenwolf, 1988). In mammals and birds, all three cell behaviors determine the final shape of the neural plate, while in amphibians cell rearrangement is the predominant mechanism regulating neural plate shaping (Kieserman and Wallingford, 2009).

1.4.1.2.1 Apical-basal thickening of neural plate cells contributes to neural plate shaping During neural plate shaping, neuroepithelial cells continue to increase in height and decrease in diameter except at the midline where they become shorter. Apical-basal elongation of neural ectoderm cells distinguishes them from the adjacent flatter non-neural ectoderm cells and contributes to narrowing of the neural plate (Schoenwolf and Franks, 1984; Schoenwolf and Powers, 1987). Elongation of neural plate cells is partly regulated by microtubules as their depolymerization results in a 25% decrease in cell height with a corresponding increase in neural plate width (Schoenwolf and Powers, 1987).

1.4.1.2.2 Convergent extension regulates neural plate shaping Neural ectoderm cells intercalate along the mediolateral plane contributing to the narrowing (convergence) and anterior-posterior lengthening (extension) of the neural plate (Davidson and Keller, 1999; Elul et al., 1997; Keller et al., 1992; Schoenwolf and Alvarez,

1989). Collectively, these cell movements are known as convergent extension. At the molecular level, convergent extension is regulated by the non-canonical Wnt/planar cell polarity (PCP) pathway. The PCP pathway is highly conserved from Drosophila, the species in which it was first identified, to higher vertebrates (Devenport, 2016; Nikolopoulou et al., 2017). In Drosophila epithelia, the six core PCP proteins asymmetrically localize to either the proximal or distal side of a cell and form complexes that act to polarize cells within the plane of a tissue (Fig. 1.7). A similar polarized distribution occurs in the neural plate during convergent extension and is thought to provide a directional cue for cellular motility regulating the lateral-to-medial displacement of neural ectoderm cells (Nishimura et al., 2012; Ossipova et al., 2014; Ossipova et al., 2015). Disrupting key signaling molecules in the PCP pathway leads to neural tube defects caused by disrupted convergent extension. In Xenopus, overexpression and knockdown of PCP components causes convergent extension defects resulting in an abnormally short and wide neural plate in which the neural folds are spaced too far apart for them to fuse at the midline (Goto and Keller, 2002; Wallingford and Harland, 2002). Defective convergent extension is also observed in mice carrying mutations in PCP genes (Curtin et al., 2003; Greene et al., 1998; Paudyal et al., 2010; Wallingford and Harland, 2002; Wang et al., 2006a; Wang et al., 2006b). These mice develop cranial and spinal neural tube defects suggesting that convergent extension, under the control of PCP signaling, is critical at all axial levels during neural tube closure. A similar phenotype is observed in zebrafish PCP mutants (Tawk et al., 2007). Although a role for the PCP pathway in regulating convergent extension during avian neural tube closure has not been assessed, siRNA-mediated knockdown of the PCP component Celsr1 causes neural tube defects in chick embryos (Nishimura et al., 2012). These data suggest that the PCP pathway plays an evolutionary conserved role in vertebrate neural tube closure.

1.4.1.2.3 Oriented cell division is critical for neural plate shaping Oriented cell division also contributes to the lengthening of the neural plate. In the avian and mammalian neural plate, mitotic spindles are preferentially oriented along the anterior- posterior axis compared to the medial-lateral axis suggesting that non-random oriented cell division contributes to the lengthening of the neural plate (Sausedo et al., 1997). The molecular mechanisms regulating oriented cell division in the neural plate remain poorly understood, but studies in amphibians and fish suggest a critical role for components of the apical-basal polarity

pathway. In Xenopus neuroepithelial cells, Cdc42 functions to stabilize the position of the metaphase spindle (Kieserman and Wallingford, 2009). Par3 controls the mirror-symmetric neural progenitor divisions during zebrafish neurulation ensuring that daughter cells integrate into the opposite sides of the developing neural tube (Tawk et al., 2007). Knockdown of Par3 causes neural tube defects in chick embryos (Chen et al., 2017). Whether defects in oriented cell division are responsible for the NTDs observed in Par3-depleted avian embryos has not been assessed. The available evidence suggests that components of the apical-basal polarity pathway regulate asymmetric cell division in the Xenopus and zebrafish neural plate and that they are important for vertebrate neural tube closure. The extent to which proteins in the apical-basal polarity pathway regulate neural tube closure because of their role in oriented cell division in chick and mouse embryos has not been explored.

1.4.1.3 Phase III: Bending of the neural plate and elevation of the neural folds As the neural plate is being shaped along its anterior-posterior and medial-lateral axes, it begins to bend and elevate forming bilateral neural folds which converge towards the midline. The morphogenetic movements driving this process require the coordinated action of intrinsic forces within the neural plate and extrinsic forces generated outside the neural plate (Hackett et al., 1997; Moury and Schoenwolf, 1995; Sausedo et al., 1997; Schoenwolf, 1991a; Schoenwolf and Alvarez, 1989; Schoenwolf and Franks, 1984; Smith and Schoenwolf, 1987; Smith and Schoenwolf, 1988; Smith and Schoenwolf, 1991; Smith et al., 1994). The intrinsic forces responsible for neural plate bending are generated by formation of three hinge points: the median hinge point (MHP) and paired dorsolateral hinge points (DLHPs). These hinge points act as pivot points that direct folding of the neural plate: folding around the MHP causes neural fold elevation, while folding around the DLHPs results in neural fold convergence. Changes in cell shape, position and number of non-neural ectoderm cells generates extrinsic forces that contribute to neural fold elevation. In addition, emigration of neural crest cells, a population of mesenchymal cells that arise from the tips of the neural folds, contributes to dorsolateral bending (Copp, 2005). Precisely how neural crest migration enables dorsolateral bending of the neural folds is unclear, but a reduction in cell density as neural crest cells emigrate could play a permissive role by increasing the mechanical flexibility of the dorsolateral neuroepithelium.

1.4.1.3.1 Median hinge point formation causes neural plate bending and neural fold elevation The MHP extends the entire anterior-posterior length of the neural ectoderm of chick and mouse embryos (Kinoshita et al., 2008; Schoenwolf et al., 1988; van Straaten et al., 2002; Ybot- Gonzalez and Copp, 1999). It is required for cranial and spinal neural tube closure in the chick, and cranial neural tube closure in the mouse. The notochord underlying the neural tube secretes Sonic hedgehog (Shh) and the BMP antagonists chordin and noggin to induce MHP formation (Patten and Placzek, 2002; Ybot-Gonzalez et al., 2002; Ybot-Gonzalez et al., 2007a). During later stages of neurulation, Shh is also produced by the MHP, the precursor of the floor plate (Chiang et al., 1996; Greene et al., 1998). Once specified, the cells of the MHP undergo a characteristic change in cell shape from columnar to wedge-shaped and the nuclei of MHP cells are positioned basally (McShane et al., 2015; Schoenwolf and Franks, 1984; Smith and Schoenwolf, 1987; Smith and Schoenwolf, 1988; Smith et al., 1994). Wedging of MHP cells generates intrinsic forces that deform the midline of the neural plate resulting in neural fold elevation and formation of the neural groove. Wedging is generated by apical constriction and basal expansion of midline neuroepithelial cells. Actin-myosin dynamics play a key role in apical constriction. Inhibiting actin polymerization using cytochalasin D in cultured chick embryos prevents apical constriction and MHP formation resulting in completely open neural tubes (Schoenwolf, 1998; van straaten, 2002; Kinoshita, 2008 (Schoenwolf et al., 1988). Mouse embryos treated with cytochalasin D develop cranial neural tube defects caused by defective MHP formation (Ybot-Gonzalez and Copp, 1999). Ectopic accumulation of actin also causes neural tube defects. Cranial neural tube defects are seen in mouse mutants lacking cofilin, an actin-severing protein (Grego-Bessa et al., 2015), and mouse embryos treated with Jasp, a drug that blocks actin depolymerisation, develop spinal neural tube defects (Escuin et al., 2015). Activation of myosin light chain is also critical for apical constriction (Escuin et al., 2015; Kinoshita et al., 2008). Chick embryos treated with blebbistatin, which inhibits myosin II motor activity and prevents actin-myosin crosslinking, develop caudal neural tube defects due to defective apical constriction (Kinoshita et al., 2008). Mouse embryos treated with blebbistatin develop cranial neural tube defects (Escuin, 2015). Together, these data suggest that a balance between F-actin turnover and actin-myosin contraction is critical for apical constriction and MHP formation.

RhoA-ROCK signaling regulates actin-myosin dynamics during MHP formation. The MHP does not form in chick embryos treated with C3, a RhoA inhibitor, or Y27632, a ROCK inhibitor, and the neural folds fail to elevate. Phosphorylated MLC (pMLC) is reduced in embryos exposed to these inhibitors. Mouse embryos treated with the ROCK inhibitor Y27632 develop spinal neural tube defects indicating that RhoA-ROCK signaling plays a conserved role in regulating neural plate bending in vertebrates (Escuin et al., 2015). Spinal neural tube closure in mouse embryos requires ROCK-dependent disassembly of actin filaments and not ROCK- mediated MLC activation (Escuin et al., 2015). ROCK inhibition causes increased apical accumulation of actin and myosin resulting in a stiff neural plate that resists bending. Hence, neural plate bending requires regulated turnover of the actin-myosin cytoskeleton at the apical surface of neural ectoderm cells. In the chick, the PCP pathway interacts with RhoA-ROCK signaling at the apical surface of neural plate cells to ensure that phosphorylation of MLC and actin-myosin contraction occur in a polarized manner causing cells to constrict at their apical surface and interacalate (Nishimura et al., 2012) (Fig. 1.8). This promotes the simultaneous bending and narrowing of the neural plate. Basal expansion of MHP cells is directly linked to the basal migration of their nuclei (Fig. 1.8). The position of the nucleus is determined by the cell cycle because of interkinetic nuclear migration. The nucleus of a neuroepithelial cell is located basally during DNA synthesis and migrates apically during mitosis. This led to the hypothesis that basal cell expansion at the MHP can be achieved by altering progression of the cell cycle so that nuclei remain positioned basally. Indeed, studies conducted in avian embryos showed that the overall length of the cell cycle, including S-phase, is prolonged and mitosis is significantly shorter in cells of the MHP compared to lateral neuroepithelial cells (Smith and Schoenwolf, 1987; Smith and Schoenwolf, 1988). Recently, this observation has been shown to hold true in the mouse (McShane et al., 2015), suggesting a conserved morphogenetic mechanism in vertebrates regulating basal expansion at the MHP.

1.4.1.3.2 Dorsolateral hinge points contribute to neural fold convergence DLHPs are restricted to the cranial neural tube in the chick, while they are located in the spinal region of the mouse (Nagai et al., 2000; Ybot-Gonzalez et al., 2007a). Cells of the DLHPs become wedge-shaped but, unlike MHP cells, they undergo apical-basal elongation (McShane et

35 al., 2015). An increase in cell number also occurs at the DLHPs (Fig. 1.8). DLHP cells are the most rapidly proliferating cells in the neuroepithelium. Cell labeling experiments showed that neuroepithelial cells move in a ventral-to-dorsal direction during neural plate elevation (Fig 1.8). Together, the enhanced cell proliferation and translocation of neuroepithelial cells contribute to a higher number of cells in the dorsolateral neuroepithelium. An increase in cell density combined with apical constriction of individual cells causes local buckling of the neuroepithelium at the DLHPs resulting in inward bending of the neural folds.

1.4.1.3.3 Expansion of the non-neural ectoderm generates extrinsic forces that contribute to neural fold elevation The non-neural ectoderm also plays a key role in elevation and convergence of the neural folds (Schoenwolf, 1991b; Smith and Schoenwolf, 1991). When isolated pieces of avian neural plate are cultured they undergo normal shaping but the neural folds fail to elevate. In contrast, explants form neural folds when a portion of the non-neural ectoderm is left intact (Moury and Schoenwolf, 1995). Similar findings were reported for mouse and amphibian embryos, suggesting a conserved role for the non-neural ectoderm in neural plate bending among vertebrates (Brun and Garson, 1983). Changes in cell shape, position and number contribute to the extrinsic forces generated by the non-neural ectoderm (Lawson et al., 2001; Schoenwolf and Alvarez, 1991) (Fig. 1.8). The transformation of non-neural ectoderm cells from a cuboidal to a squamous shape by increasing their widths and decreasing their heights generates a pushing forces that causes the neural folds to bend inwards (Lawson et al., 2001; Schoenwolf and Alvarez, 1991). Non-neural ectoderm cells, like neural ectoderm cells, undergo convergent extension movements (Schoenwolf and Alvarez, 1991). These cell rearrangements contribute to anterior-posterior elongation of the non-neural ectoderm and generate a pushing force that contributes to neural fold convergence. In addition, mitotic spindles are preferentially oriented along both anterior- posterior and medial-lateral axes (Sausedo et al., 1997). Preferential divisions along the anterior posterior axis are an autonomous process as they occur when this tissue is isolated from the neural ectoderm and they contribute to elongation of the non-neural ectoderm. In contrast, enrichment of cell divisions in the medial-lateral plane depends on interactions with the neural plate generating a force that brings the paired neural folds into contact at the midline. Together,

36 expansion of the non-neural ectoderm and medial-lateral cell divisions generate a pushing force to ensure that the neural folds bend inwards and not outward. Non-neural ectoderm cells also regulate formation of DLHPs by exerting a localized signal that induces proliferation of cells in these paired hinge points (McShane et al., 2015). Removal of the non-neural ectoderm overlying the neural folds in mouse embryos prevents formation of DLHPs and the enhanced cell proliferation within the dorsal neuroepithelium is lost (McShane et al., 2015).

1.4.1.4 Phase IV: Fusion of the neural folds Completion of neural tube closure involves fusion of the apposing neural folds to form a closed neural tube. Prior to fusion, the non-neural ectoderm is in direct contact with the underlying neural ectoderm forming a continuous ectodermal layer (Fig 1.9). During epithelial fusion, individual cells do no actually fuse with one another, but rather tissue remodeling separates the neural and non-neural ectoderm tissues into two distinct epithelial layers: the inner neural tube, which forms the brain and spinal cord, covered by a continuous layer of non-neural surface ectoderm. The morphogenetic process of epithelial fusion is divided into three distinct phases: 1) initial contact between apposing neural folds 2) epithelial adhesion, and 3) epithelial remodeling.

1.4.1.4.1 Cellular protrusions make the initial contact between the apposed neural folds The first step of neural fold fusion involves apical protrusions emanating from the neural and the non-neural ectoderm that make the initial contact across the space between the neural folds. Cellular protrusions were first described as emanating from the approaching neural folds of amphibian, chick and mouse embryos using scanning and transmission electron microscopy in the 1970s (Bancroft and Bellairs, 1975; Geelen and Langman, 1979; Mak, 1978; Schoenwolf, 1979; Waterman, 1976). These studies revealed that whether the initial contact between the approaching neural folds is made by the neural ectoderm, the non-neural ectoderm or both differs along the body axis and between species. In Xenopus, the non-neural ectoderm closes first and this is followed by independent closure of the neural ectoderm (Mak, 1978). In chick cranial neural tube closure, the neural and non-neural ectoderm cell layers make contact at the same time, but contact in the spinal region is first made by non-neural ectoderm cells (Schoenwolf, 1979). In mouse, the initial contact between the neural folds in the midbrain, hindbrain and

spinal regions is made by the non-neural ectoderm cells, while at the forebrain level contact is first made by neural ectoderm cells (Geelen and Langman, 1979). There are two types of cellular protrusions emerging from ectoderm cells coincident with neural fold fusion: lamellipodia and filopodia. Lamellipodia or membrane ruffles are broad, sheet-like cell protrusions containing a branched network of actin filaments. In contrast, filopodia are finger-like projections with an actin filament core. The formation of these membrane protrusions is regulated by the Rho family of small GTPases, with Rac1 being involved in lamellipodia formation (Sugihara et al., 1998) and Cdc42 regulating filopodia assembly (Chen et al., 2000). While it has long been known that membrane protrusions are present at the edges of approaching neural folds during neural tube closure, there was no evidence supporting the idea that these protrusions are required for neural tube closure at any level of the body axis. This was, in part, due to the embryonic lethality of Cdc42 and Rac1 knockout mice prior to neurulation (Chen et al., 2000; Sugihara et al., 1998). Recent analysis of conditional knockouts of Cdc42 and Rac1 in the neural or non-neural ectoderm solved this problem (Rolo et al., 2016). Conditional inactivation of Rac1 in neural ectoderm cells did not affect neural tube closure. However, embryos in which Rac1 was conditionally ablated in the non-neural ectoderm had no lamellipodia and developed cranial and spinal neural tube defects. These data support the idea that Rac1-mediated formation of lamellipodia in the non-neural ectoderm is required for neural tube closure. The number of filopodia is significantly lower in E8.5 embryos which lack Cdc42 in the non-neural ectoderm, suggesting that Cdc42 plays a critical role in filopodia formation. Unfortunately, the role of Cdc42-induced filopodia in neural fold fusion could not be assessed as these embryos died prior to neural tube closure. Together, these data suggest that cellular protrusions originating from non-neural ectoderm cells are required for neural tube closure in mouse embryos. The importance of cellular protrusions in mediating neural fold fusion in avian embryos remains unclear. While filopodia have been observed emanating from non-neural ectoderm cells in the spinal region of chick embryos, the number of cellular protrusions is far fewer than what is observed in the mouse (Schoenwolf, 1979). Furthermore, no cellular protrusions are seen in the cranial neural tube and membrane ruffles are not present at any axial level. Hence, the importance of cellular protrusion in medating neural fold fusion in avian embryos needs to be further explored.

1.4.1.4.2 Epithelial adhesion stabilizes the initial contact between the neural folds Compared to the cellular protrusion phase, much less is known about the adhesion phase. The predominant view is that the initial contact made by epithelial protrusions between apposing neural folds is stabilized by cell adhesion molecules (Pai et al., 2012). An alternative hypothesis suggests that epithelial adhesion does not exist as a distinct phase, but rather that it is purely the forces generated by epithelial bending that keep the neural folds apposed until epithelial remodeling has progressed sufficiently to ensure a stable connection across the midline. The interaction between Eph receptors and ephrin ligands on the tips of the dorsal neural folds and overlying non-neural ectoderm is thought to stabilize the initial contact between the apposed neural folds (Abdul-Aziz et al., 2009; Pai et al., 2012; Ray and Niswander, 2012). Eph receptors are unlike other receptor tyrosine kinase families because their ligands, the ephrins, are linked to the plasma membrane by a glycophosphatidylinositol (GPI) anchor. This means that bidirectional signaling through Eph-ephrin interactions is largely dependent on cell-cell contact. As such, Eph receptor-ephrin complexes are uniquely positioned to regulate morphogenetic events involving cell adhesion and fusion such as neural tube closure. All five ephrin receptors (EphA1-5) are expressed in the elevating spinal neural folds of mouse embryos with EphA2 expression restricted to the tips of the neural folds just prior to closure (Abdul-Aziz et al., 2009). Interestingly, EphA2 is also detected on the lamellipodia protrusions emanating from the neural folds, suggesting that EphA-ephrinA interactions could be one of the earliest adhesion events in neural fold fusion (Abdul-Aziz et al., 2009). Their ligands, the ephrins, are differentially expressed along the anterior-posterior neural tube with ephrinA1 and A3 being highly expressed in the caudal region of the neural tube and ephrinA2 and A5 strongly expressed in the cranial neural tube. EphrinA4 is expressed throughout the neural tube (Abdul-Aziz et al., 2009). Experimental evidence for a role of EphA-ephrinA interactions in regulating neural tube closure comes from studies done in mouse. Cleavage of GPI-anchored proteins, including ephrinAs, from the embryonic cell surface of whole cultured mouse embryos results in delayed spinal neural tube closure (Abdul-Aziz et al., 2009). In addition, injection of EphA1 and EphA3 fusion proteins intra-amniotically into cultured embryos to disrupt ephrinA-EphA receptor interactions causes spinal neural tube defects (Abdul-Aziz et al., 2009). Furthermore, ephrinA5 and EphA7 null mice exhibit cranial NTDs (Holmberg et al.,

2000). In all cases, neural plate bending and elevation occur normally arguing for a role of Eph- ephrin interactions in the final fusion of the neural folds. A role for ephrin ligands and Eph receptors in avian neural tube closure has not been assessed, however, these genes are evolutionarily conserved displaying similar expression patterns to their mouse orthologs: ephrinA2 and A5 are enriched in the cranial neural tube, while EphA1, A3, and A4 are expressed in the spinal neural tube (Bell et al., 2004; Colas and Schoenwolf, 2003b; Darnell et al., 2007). The expression patterns for ephrinA1, A3, A4 and EphA2 and A5 have not been described. These data argue for a conserved role for Eph-ephrin interactions in chicken neural tube closure. Interestingly, crosstalk between between claudins and ephrin ligands and receptors has previously been reported (Bhavaniprasad et al., 2013; Shang et al., 2014; Swager et al., 2015). Thus, it is tempting to speculate that claudins may interact with ephrin ligands and receptors to regulate fusion of the neural folds. In addition to their role in mediating adhesion between the neural folds, epithelial adhesion molecules maintain the epithelial integrity of non-neural ectoderm cells during neural fold fusion. This is best illustrated in the ENU-induced Grhl21Nisw mutant mice, which exhibit cranial neural tube defects due to failure of the neural folds to fuse (Pyrgaki et al., 2011). Grainyhead- like 2 (Grhl2) is a member of the Grhl family of transcription factors and is expressed in the non- neural ectoderm of mice throughout neurulation. Grhl2 regulates neural tube closure by inhibiting epithelial-to-mesenchymal transition as loss of Grhl2 leads to expression of mesenchymal markers in the non-neural ectoderm (Ray and Niswander, 2016). Consistent with this hypothesis, downstream targets of the Grhl2 transcription factor include genes for cell adhesion molecules such as E-cadherin and Cldn3, -4, -6 and -7 (Pyrgaki et al., 2011; Rifat et al., 2010; Werth et al., 2010). Interestingly, claudins have been implicated in maintaining epithelial integrity in Xenopus embryos. Morpholino knockdown of claudin-6 in Xenopus prevents the pronephric tubules from undergoing mesenchymal-to-epithelial transition (Sun et al., 2015). Knockdown of Grhl2 in inner medullary collecting duct (mIMCD-3) epithelial cell cysts results in impaired lumen formation, a phenotype which is rescued when Cldn4 is overexpressed (Aue et al., 2015). Together, these data suggest that downregulation of claudins is responsible for the neural tube defects observed in Grhl2 mutant mice. However, we cannot rule out the possibility that E-cadherin is the primary target of Grhl2 in neurulation because disruption of E-cadherin

40 expression using antisense oligonucleotides causes cranial neural tube defects in cultured rat embryos (Chen and Hales, 1995).

1.4.1.4.3 Tissue remodeling separates the neural and non-neural ectoderm into distinct cell layers Even less is known about the epithelial remodeling phase that separates the neural and non- neural ectoderm into two distinct cell layers resulting in a closed neural tube overlain by a continuous layer of non-neural ectoderm. Initially, it was thought that apoptosis played a key role in epithelial remodeling as an increase in cell death is observed coincident with neural fold fusion and inhibition or knockout of pro-apoptotic genes causes neural tube defects in chick (Weil et al., 1997) and mouse embryos (Cecconi et al., 1998; Leonard et al., 2002), respectively. However, the current thought is that apoptosis is a consequence of neural fold fusion and cells that lose cell-cell adhesion during epithelial remodeling undergo cell death. In support of this view, mosaic embryos for deletion of the cell-cell adhesion molecule Merlin (also known as Nf2) exhibit fusion defects affecting the neural tube, eye, heart and palate due to failure to establish apical junctional complexes (McLaughlin et al., 2007).

1.4.2 Discontinuous neural tube closure in mammals and birds The phases of neural tube closure are conserved in higher vertebrates: the neural plate is induced to differentiate, it then bends and elevates to form neural folds, which meet at the dorsal midline, and, finally, the neural fold tips fuse forming a closed neural tube. In mammals and birds, unlike amphibians, this process does not occur in a continuous manner, but rather is initiated at several discrete points along the anterior posterior axis of the embryo. The number and location of these closure points and the order in which they close varies across species. In mouse, the initial neural tube closure point (closure 1) is at the hindbrain/cervical boundary and occurs at embryonic day (E) 8 (Greene and Copp, 2009). At E9, two additional closure sites form anteriorly to the first closure point including one at the forebrain/midbrain boundary (closure 2) and one at the anterior end of the future forebrain (closure 3). These multiple closure initiation sites create three neuropores, which are open regions of the neural tube: the anterior and hindbrain neuropores in the cranial region and the posterior neuropore in the lower spinal region. Neural tube closure proceeds bidirectionally from closure sites 1 and 2

and in a caudal or posterior direction from closure 3. While cranial neural tube closure is complete by E9.5, the spinal neural tube closes at E10. The chick has two points of closure: one at the level of the future midbrain (closure 1) and a second at the hindbrain/cervical boundary (closure 2). Neural tube closure initiates at closure site 1 at Hamilton and Hamburger (HH) stage 8 creating the anterior and posterior neuropores (Schoenwolf, 1979; Van Straaten et al., 1996). Closure proceeds bidirectionally from closure site 1 with cranial neural tube closure completed by HH12. Spinal neurulation is more complex. Although it is initiated by caudal progression of closure from closure site 1, by HH9 multiple small contact sites form at the hindbrain/cervical boundary (closure 2) separating the initial posterior neuropore into the rhombencephalic and posterior neuropores. The rhombencephalic neuropore is present for only 4-5 hours; by HH10 the anterior spinal cord is closed. Closure 2 is complete at HH14. In the human embryo, neural tube closure begins 17-18 days after fertilization (O'Rahilly and Muller, 1994). The initial closure site is located in the rhombencephalon or hindbrain, which is slightly anterior to closure site 1 in the mouse or closure 2 in the chick. A second closure site is located at the level of the future forebrain similar to closure site 3 in the mouse. The existence of a site equivalent to closure 2 in the mouse or closure 1 in the chick is still unclear. Analysis of early human embryos suggests that closure 2 occurs more caudally, in the hindbrain (Nakatsu et al., 2000), or does not exist (O'Rahilly and Muller, 2002). However, analysis of late stage fetuses with cranial neural tube defects led to the hypothesis that a closure site at the forebrain/midbrain boundary exists in the human embryo (Seller, 1995; Van Allen et al., 1993). Closure in the cranial region is complete on day 25, while spinal neurulation is complete at 26-28 days post-fertilization. In summary, neural tube closure is initiated at several discrete points along the anterior- posterior axis of higher vertebrates. Failure to initiate closure at a particular closure site or disruption of the progression of closure between closure sites will result in a neural tube defect.

1.4.3 Neural tube defects Neural tube defects (NTDs) represent a group of congenital malformations of the central nervous system and can occur at any level of the neuraxis affecting the developing brain and/or spinal cord. NTDs are classified as “open” when the affected nervous tissue is exposed to the

42 amniotic environment or “closed” when the defect is covered by skin. They are the second most common birth defect with a frequency that typically ranges between 0.5-1 per 1000 pregnancies (Zaganjor et al., 2016). However, in some geographical regions, e.g. Northern China, frequencies as high as 10 per 1000 births have been reported (Li et al., 2006).

1.4.3.1 Types of open NTDs Open NTDs, the type of NTDs being studied in this thesis, occur when the neural folds fail to fuse during primary neurulation (De Marco et al., 2011). These defects occur at different levels of the anterior-posterior and reflects the occurrence of multiple closure sites. Failure of the anterior end of the neural tube to close results in cranial NTDs. In cranial NTDs, the neural folds at the level of the future brain remain open and the neuroepithelial tissue protrudes from the embryo, a phenotype referred to as exencephaly. As the embryo continues to develop, the neural tissue remains exposed to the toxic amniotic environment leading to its degeneration. This gives rise to the characteristic flat brain observed at birth in patients with anencephaly. These babies are stillborn or die shortly after birth. Open spinal NTDs occur when the neural tube fails to close at the posterior end of the embryo leading to spina bifida and are the most common type of open NTD. While the developmental cause of spina bifida is the same, the clinical phenotype can vary. The most severe and common form is myelomeningocele, in which the membranes around the spinal cord (meninges) and the spinal cord protrude at birth, forming a sac on the baby’s back. The sac can easily rupture causing the baby to develop meningitis. A patient with this condition usually has partial or complete paralysis of their lower body and children with myelomeningocele may develop hydrocephalus, swelling of the brain. In myelocele, the neural tissue is exposed directly to the amniotic cavity. Meningocele refers to a form of spina bifida in which the meninges push out through an opening in the back, but the spinal cord develops normally. These membranes can be surgically removed with little or no damage to the nervous tissue. Craniorachischisis refers to a condition in which the entire neural tube remains open from the midbrain to the lower spine. This is the most severe type of NTD and is embryonic lethal. The prevalence of craniorachischisis is difficult to assess as fetuses with this type of defect often die early during pregnancy.

1.4.3.2 Causes of NTDs NTDs have a complex etiology with both genetic and environment factors contributing to their development. The interaction of these factors modulates the incidence and severity of NTDs. Recognized environmental risk factors for the development of NTDs include maternal usage of anticonvulsant drugs, maternal diabetes and obesity, maternal exposure to high temperatures in early pregnancy (e.g. fever and hyperthermia), maternal infection with the Zika virus, and maternal nutrition (e.g. low folate intake) (Reviewed in Nikolopoulou et al., 2017). Indeed, clinical trials revealed that peri-conceptional folic acid supplementation can prevent up to 70% of NTDs (Blencowe et al., 2010; Seller and Nevin, 1984; Smithells et al., 1983; Smithells et al., 2011; Viswanathan et al., 2017). The protective effect of folic acid is further supported by the significant reduction in the prevalence of NTDs in countries with mandatory fortification of food with folic acid. Following folic acid fortification in the 1990s, the incidence of NTDs in the United States and Canada decreased by 36% and 46%, respectively (De Wals et al., 2007; National Center on Birth Defects and Developmental Disabilities, 2016). Some NTDs are not prevented by folic acid, and inositol has emerged as a potential preventative agent in cases that do not respond to folate. A recent pilot clinical trial reported no recurrences of NTDs in women with a previous NTD pregnancy who took inositol pre-conceptionally (Greene et al., 2016). Furthermore, inositol supplementation is able to prevent NTDs in mouse models with folate- resistant NTDs (Greene and Copp, 1997). While these results are promising, larger-scale studies are needed to establish whether inositol is effective in preventing NTDs. Genetic factors also contribute to NTDs. Compared to the general population, there is a 10-20 fold higher recurrence risk to siblings in families with a child affected by a NTD (Pangilinan et al., 2012). The contribution of mutations in folate-related genes to the etiology of NTDs has been driven by the finding that peri-conceptional folic acid supplementation has a protective effect. Indeed, a common single-nucleotide polymorphism, 677C→T, in the gene for 5,10-methylene-tetrahydrofolate reductase (MTHFR), which encodes an enzyme involved in folate metabolism, is associated with an increased risk for developing a NTD (Kibar et al., 2007; Pangilinan et al., 2012). Mutations in many other genes involved in the folic acid pathway have also been associated with an increased risk for NTDs. Studies in animal models represent a powerful tool for identifying genes involved in neural tube closure. Genes that cause NTDs in animal models when mutated or knocked out represent strong candidates to contribute to human

NTDs. More than 200 genetic mouse models for NTDs have been generated and they include genes encoding cytoskeletal proteins and adhesion molecules, planar cell polarity genes, RhoGTPases, cell cycle genes, genes regulating cell death, extracellular matrix genes, transcription factors, and folate-related genes (Harris and Juriloff, 2010). However, mutations in only a few of the genes involved in these pathways have been identified as genetic risk factors for NTDs in humans. The planar cell polarity pathway (PCP) is one such example as mutations in genes of the PCP pathway are implicated in the etiology of human NTDs (De Marco et al., 2013; De Marco et al., 2012; Kibar et al., 2009; Kibar et al., 2011). Other candidate genes found to have mutations predisposing to NTDs in humans include SHROOM3, an actin binding protein involved in apical constriction, the transcription factors grainyhead-like 3 (GRHL3) (Lemay et al., 2015) and paired box (PAX) 1, 3, 7 and 8 (Agopian et al., 2013; Hol et al., 1996; Lemay et al., 2015; Volcik et al., 2002), protein tyrosine phosphatase receptor type S (Lemay et al., 2015), a phosphatase involved in regulating cell proliferation and cell adhesion, Sonic hedgehog (SHH), which is involved in patterning and bending of the neural tube, and PAR3 (Chen et al., 2017), which is involved in establishing apical-basal polarity. In summary, human NTDs represent a complex trait involving both genetic and environmental components. NTDs are thought to be the result of multigenic inheritance with environmental factors modulating the risk for developing a NTD (Kibar et al., 2007). While significant progress has been made in deciphering the complex etiology of neural tube defects, the genetic and environment causes of NTDs have not been fully elucidated. Determining the specific causes of NTDs is best achieved in the context of understanding the mechanisms underlying normal neural tube closure. Given the inaccessibility of the human embryo during neurulation, our knowledge of the main mechanisms regulating neural tube closure comes from the analysis of animal models.

1.4.4 Using the chick as an animal model to study the role of claudins in neural tube closure

1.4.4.1 A brief history of the chick as an experimental model The chick embryo has long been a favoured animal model for developmental biologists since Aristotle examined embryos at different incubation times to examine the different stages of development (Graper, 1929; Waddington, 1932). During the 1600s, anatomists described early

structures of the chick embryo including the neural tube, somites, and the heart. By the early 1900s, scientists learned how to culture chick embryos ex ovo which made the chick embryo amenable to microsurgical and fate mapping methods leading to landmark studies, including Grãper’s time-lapse movies monitoring the movements of labelled cells in living embryos from gastrulation to organogenesis (Graper, 1929) and Waddington’s cross-species transplant experiments which provided the first evidence that Hensen’s node has inducing properties (Waddington, 1932). Since then, the chick has been extensively used in the field of developmental biology and has contributed to advances in the fields of virology, cancer, and immunology.

1.4.4.2 The chick as an experimental model of neural tube closure The chick embryo is an excellent model with which to study neural tube closure because the tissue and cell behaviors that regulate neurulation in the chick have been well studied since the 1970s (Chen et al., 2017; Kinoshita et al., 2008; Lawson et al., 2001; Moury and Schoenwolf, 1995; Nishimura et al., 2012; Sausedo et al., 1997; Schoenwolf and Alvarez, 1989; Schoenwolf and Franks, 1984; Schoenwolf and Powers, 1987; Smith and Schoenwolf, 1987; Smith and Schoenwolf, 1988; Smith and Schoenwolf, 1991; Smith et al., 1994). Furthermore, the morphogenetic movements regulating neural tube closure in the chick closely resemble the human embryo. In addition, the chick offers several advantages over mammalian animal models: low cost, the ease in which the embryo can be cultured ex vivo that recapitulates normal in ovo development, and a short period of embryogenesis. Fertilized eggs are commercially available year-round and can be grown in an incubator at 39ºC until the desired Hamilton and Hamburger (HH) stage (Hamburger and Hamilton, 1951). The easy accessibility of eggs, combined with the fact that neural tube closure in the chick embryo occurs over a short period of time (~28 hours), makes it is easy to collect a large number of embryos at each stage of neural tube closure as defined by HH staging criteria. Furthermore, the chick, like the human embryo, develops as a flat blastodisc so that it can be easily cultured ex ovo, recapitulating normal in ovo development. Chicken embryos can be cultured ex ovo on solid agar-albumen plates for the first three days of development (Collins and Ryan, 2011) or placed in liquid culture for the first two days (Nagai et al., 2011). Thus, neural tube closure can be monitored in real time.

1.4.4.3 Experimental manipulations in the chick Functional analysis of candidate neurulation genes requires gain- and loss-of-function studies, which can easily be performed in the chick embryo. Gain-of-function experiments are performed by electroporating a eukaryotic expression plasmid or by injecting retroviral particles generated from the the replication competent avian retrovirus plasmid, RCAS (Collins and Ryan, 2011; Hughes, 2004; Voiculescu et al., 2008). Expression plasmids encoding a gene of interest can be focally or broadly electroporated into the embryo. This is sufficient for short-term experiments (24-36 h) as the plasmid is diluted with each round of cell division, while retroviral injection is preferred for long-term experiments since the viral DNA will stably integrate into the genomic DNA and will be maintained in daughter cells. Two methods are available for loss-of- function in the chick that function by blocking RNA translation: electroporation of short interfering RNAs (siRNAs) or morpholino antisense oligonucleotides (MOs) (Voiculescu and Stern, 2017). Unlike siRNAs, morpholinos do not cause degradation of their RNA targets and are more stable because of their uncharged backbone, so effects of knockdown can be assessed over a longer time period (GeneTools).

1.4.4.4 Evolutionary conservation of claudins in the chicken genome The chicken was the first avian species whose genome was sequenced, and the first build of the genome was published in 2004 (Consortium, 2004). Since then, many gaps in the chicken genome have been filled with the most recent chicken genome assembly being released in 2015. In vertebrates, 25 claudin family members have been identified to date. There are 19 claudins annotated in the most recent build of the chicken genome (CLDN6, 7, 13, 21, 22 and 24 are absent; NCBI Gallus_gallus-5.0, GCF_000002315.4). Phylogenetic analysis shows that chicken claudins are highly similar to their mammalian homologs with each chicken claudin clustering more closely with its mouse or human ortholog than with other claudin family members (Collins et al., 2013). In addition, the genomic organization of claudins in chicken is similar to what is observed in humans and mice. Claudins are distributed throughout the genome with the exception of three closely linked pairs: CLDN3 and 4, CLDN6 and 9, and CLDN8 and 17. CLDN3 and 4 are located 40 kb apart on chicken chromosome 19, and CLDN8 and 17 are located 5 kb apart on chromosome 1. (Consortium, 2004; Lal-Nag and Morin, 2009) The

chicken ortholog of CLDN6 lies within a syntenic block that has not been annotated in the chicken genome.

1.4.4.5 Expression pattern of claudins during chick neural tube closure We performed an mRNA expression pattern analysis of the 17 chicken claudins annotated as of 2013 in the chick embryo during the first ten days of embryonic development (Collins et al., 2013; Simard et al., 2005). A thorough analysis of the expression pattern of claudins during primary neurulation revealed that claudins are differentially expressed in the neural and non- neural ectoderm throughout the different phases of neural tube closure (Collins et al., 2013; Simard et al., 2005). Cldn1, -2, -4, -10, -15, -16, -18 and -23 are uniformly expressed in the ectoderm throughout neural tube closure. Four claudins, Cldn3, -11, -12 and -22, are more highly expressed in the non-neural ectoderm than the neural ectoderm at the different phases of neural tube closure. In contrast, three claudins, Cldn8, -14 and -17, are enriched in the neural plate compared to the adjacent non-neural ectoderm and continue to be more highly expressed in the open neural folds. Expression of all claudins is downregulated in the closed neural tube. Cldn5 is expressed in the developing vasculature (Collins et al., 2012). The cDNA sequence used to examine the mRNA expression pattern of Cldn22 was subsequently designated as Cldn25. The expression pattern of claudins during neural tube closure in mouse embryos is not well-characterized. Grainyhead-like 2 (Grhl2) is a transcription factor that is exclusively expressed in the non-neural ectoderm of mice during neural tube closure. Cldn4, -6, and -7 are downregulated in the non-neural ectoderm of Grhl2 mutant mice, which exhibit NTDs (Pyrgaki et al., 2011; Rifat et al., 2010; Werth et al., 2010). These data suggest that at least Cldn4, -6 and -7 are expressed in the non-neural ectoderm of mouse embryos during neural tube closure. The expression patterns of the remaining claudin family members have not been analyzed during neurulation in the mouse. However, we found that the expression patterns of several claudin family members are well-conserved during organogenesis in chick and mouse embryos, suggesting that they are likely to have conserved expression patterns during neural tube closure (Collins et al., 2013). In summary, the chick embryo is an ideal model organism to study the role of claudins in neural tube closure because the expression pattern of claudins is well-defined, it is easy to

perform gain- and loss-of-function experiments, and the extensive knowledge of the cell biology regulating neural tube closure allows the distinction of abnormal from normal morphogenetic movements.

1.5 HYPOTHESIS AND OBJECTIVES Claudins are integral tight junction transmembrane proteins whose cytoplasmic C- terminal tails contain a PDZ-binding domain that interacts with PDZ domain scaffolding proteins at the tight junction cytoplasmic linking claudins to signaling and polarity complex proteins. Claudin C-terminal domains are highly variable between family members and preferentially interact with different PDZ domain proteins. Previously our lab showed that 14 claudins are expressed during neurulation in the chick and a subset of claudins is differentially expressed in the neural and non-neural ectoderm: Cldn8, -14 and -17 are enriched in the neural ectoderm, while Cldn3, -11, -12 and -22 are more highly expressed in the non-neural ectoderm. I, therefore, hypothesized that the combinations of claudins expressed in the ectoderm create functional compartments that regulate differentiation and morphogenesis of the neural and non- neural ectoderm during neural tube closure and, if true, deleterious missense mutations in CLDN genes contribute to increased susceptibility to human neural tube defects.

To address this hypothesis, my first objective was to determine if claudins are required for neural tube closure. I used the C-terminal domain of Clostridium perfringens enterotoxin (C- CPE) to simultaneously remove Cldn3 and -4 from the non-neural ectoderm and Cldn4 and -8 from the neural ectoderm during neural tube closure in the chick embryo. Furthermore, mouse embryos were treated with C-CPE to determine if claudins play an evolutionarily conserved role in neural tube closure in vertebrates. The results of these experiments are described in Chapter II. The role of claudins in the non-neural ectoderm at later phases of neural tube closure could not be assessed in Chapter II due to the early defects in neural tube closure. My second objective was to determine if Cldn3 is required in the non-neural ectoderm to regulate the last phase of neural tube closure. In Chapter III, I used a C-CPE variant that specifically targets Cldn3 (C- CPELDR) to investigate the role of Cldn3 in fusion of the neural folds. To test the second part of my hypothesis, I identified rare and novel missense variants in CLDN genes in a cohort of 125

49 patients with spinal neural tube defects and performed functional analysis of these variants. These results are described in Chapter IV.

Figure 1.1 Schematic of the apical junctional complex in vertebrate epithelial cells The apical junctional complex is composed of three main types of junctions. At the apical lateral end, tight junctions (TJ, purple) encircle the cell sealing the intercellular space to prevent the passive flow of material between cells. Immediately below them, adherens junctions (AJ, turquoise) form a continuous ribbon around the cell to hold neighboring cells together. Bound to intermediate filaments inside the cell, desmosomes (D, orange) form strong attachment points to stabilize adhesion of adjacent cells.

Figure 1.2 Morphology of the tight junction (A) Ultrathin electron microscope image of tight junctions. Arrowheads point to the ‘kissing points’ where the intracellular space is eliminated. Scale bar, 50 nm. (B) Freeze-fracture replica electron microscope image of intestinal epithelial cells. Arrowheads point to the network of continuous, anastamosing tight junction strands. Scale bar, 200 nm. Mv, microvilli; Ap, apical membrane; Bl, basolateral membrane. From (Tsukita et al., 2001). Reproduced with permission from Nature Publishing Group.

Figure 1.3 Schematic of components of the tight junction and polarity complexes. Tight junctions are composed of the transmembrane proteins junctional adhesion molecules (JAMs), claudins, and the tight junction-associated MARVEL protein family (TAMPs) that are linked to a cytoplasmic plaque, which is formed by a nework of scaffolding proteins such as the ZO family, MUPP1, and cingulin and paracingulin bound to cell signaling proteins such as RhoGTPases and RhoGEFs, the Par3/Par6/αPKC and Crumbs/Pals/PATJ polarity complexes and actin filaments.

Figure 1.4 Schematic of tight junction fence and gate functions. As shown on the left, tight junctions (represented by purple rectangles) act as a fence to prevent the mixing of apical (pink stars) and basolateral (orange circles) membrane components. For their function as gates, tight junctions can form a tight seal between epithelial cells that blocks the movement of ions and small molecules through the paracellular space (middle pair of cells). Alternatively, as depicted on the right, some tight junctions are permissive and allow the passage of ions and water through the paracellular space.

Figure 1.5 Schematic of a claudin molecule. Claudins contain four transmembrane domains (TM1-4), two extracellular loops (ECL1 and 2), and a C-terminal cytoplasmic tail. ECL1 contains several basic (green) and acidic (red) amino acid residues that regulate ion selectivity of the tight junction and a highly conserved W-GLW- CC motif with two conserved cysteines that are thought to form a disulphide bond (pink). EL2 interacts with EL2 of other claudin family members. The C-terminal cytoplasmic tail contains phosphorylation sites (orange) and a binding site for PDZ-domain scaffolding proteins (grey circles). Palmitoylation sites (yellow stars) have been identified in the cytoplasmic domains immediately following the second and fourth transmembrane domains.

Figure 1.6 Phases of neural tube closure in the chick embryo. Top, whole mount images of chick embryos at the different phases of neural tube closure. Bottom, schematics of transverse sections through the chick embryo at the level of the dashed line. 1. The ectoderm differentiates into the neural ectoderm (grey) that thickens to form the neural plate and the non-neural ectoderm (green). 2. The neural plate and non-neural ectoderm extend along the anterior-posterior axis and narrow medial-laterally. 3. Cells at the midline of the neural plate (red) undergo apical constriction to form the median hinge point resulting in elevation of bilateral neural folds. Formation of dorsolateral hinge points (blue) causes convergence of the neural folds towards the midline. 4. The neural folds fuse to form a closed neural tube that disconnects from the non-neural ectoderm, which becomes a continuous layer of overlying ectoderm.

A. B.

Figure 1.7 The vertebrate non-canonical Wnt/planar cell polarity pathway. (A) Epithelial tissues display apical-basal and planar polarity. Planar polarity mediates cell polarity in the plane of the epithelia and is perpendicular to apical-basal polarity. (B) Planar polarity is generated through asymmetric protein localization. Vangl and Frizzled (Fzd) are located on opposites sides of a cell. Celsr participates in homophilic interactions between adjacent cells, while Frizzled (Fzd) and Vangl form heterophilic interactions. PK: prickle; Dvl: dishevelled.

Figure 1.8 Schematic of key mechanisms regulating hinge point formation. A number of cellular mechanisms involved in the formation of the median hinge point (MHP) and dorsolateral hinge points (DLHPs) are depicted. (1) Planar polarized actomyosin contraction: planar-polarized-controlled apical constriction causes bending along the medial-lateral axis. (2) Interkinetic nuclear migration: nuclei migrate apically to divide with daughter nuclei returning to a basal position for S-phase. A prolonged S-phase in MHP cells causes basal expansion. (3) Enhanced proliferation: DLHP cells proliferate more rapidly than lateral and MHP neuroepithelial cells. (4) Translocation: Neuroepithelial cells translocate in a ventral to dorsal direction contributing to the DLHPs. (5) Cell spreading: non-neural ectoderm cells change from a cuboidal to a squamous shape by increasing in width and decreasing their apical-basal height. (6) Cells in the non-neural ectoderm preferentially divide along the medial-lateral axis.

Figure 1.9 Schematic of mechanisms involved in neural fold fusion. (1) Protrusions: non-neural ectoderm cells extend membrane ruffles or lamellipodia that make the initial contact during neural fold fusion. (2) Adhesion: Eph-ephrin interactions stabilize the interaction between apposing neural folds. (3) Epithelial remodeling: the neural and non-neural ectoderm separate into two distinct epithelial cell layers resulting in formation of a closed neural tube overlain by a continuous layer of non-neural ectoderm.

Table 1.1. Phenotypes of mouse claudin knockout and knockdown models and human diseases caused by claudin mutations

Claudin Mouse Knock-Out Phenotype Human Disease 1 Skin barrier defect (Furuse et al., 2002) Neonatal ichthyosis and sclerosing cholangitis (OMIM: 607626) 2 Intestine barrier defect; decreased Na+, Cl- and water reabsorption in the renal proximal tubules (Matsumoto et al., 2014; Muto et al., 2010) 3 Dedifferentiated colon epithelium; blood- cerebrospinal fluid barrier defect; enamel defects (Ahmad et al., 2017; Bardet et al., 2017; Kooij et al., 2014) 4 Late-onset hydronephrosis and increased susceptibility to lung injury (Fujita et al., 2012) 5 Size-selective blood-brain barrier defect (Nitta et al., 2003) 6 Phenotypically normal (Anderson et al., 2008) 7 Die from dehydration (Tatum et al., 2010) 8 Conditional deletion in the collecting duct causes renal wasting of Na+, Cl- and K+ (Gong et al., 2015b) 9 Deafness (Nakano et al., 2009) 10 Conditional deletion in the thick ascending limb leads to hypermagnesemia and nephrocalcinosis (Breiderhoff et al., 2012) 11 Deafness, infertility, oligodendrocyte barrier defect (Gow et al., 1999; Kitajiri et al., 2004) 14 Deafness (Ben-Yosef et al., 2003) Non-syndromic deafness (OMIM: 614035) 15 Megaintestine (Tamura et al., 2008) 16 Hypercalciuria, hypomagnesemia (Hou et al., Familial hypomagnesemia, 2009; Hou et al., 2007; Will et al., 2010) hypercalciuria, nephrocalcinosis (OMIM: 248250) 18 Lung barrier defect; increased H+ permeability in the stomach leads to gastritis (Hayashi et al., 2012; Li et al., 2014) 19 Schwann cell barrier defect; siRNA knockdown Familial hypomagnesemia, model exhibits renal wasting of Mg2+ and Ca2+ hypercalciuria, nephrocalcinosis; (Hou et al., 2009; Miyamoto et al., 2005) visual impairment (OMIM: 248190)

Table 1.2 CPE sensitivity of the second extracellular loop of claudins

Claudin Amino acid sequence Highly sensitive hCldn3 TIIRDFYNPVVPEAQKREMGAG mCldn3 TIIRDFYNPLVPEAQKREMGAG cCldn3 TIIRDFYNPLVIDAQKRELGTS hCldn4 NIIQDFYNPLVASGQKREMGAS mCldn4 AIIRDFYDPLVADSQKRELGSS cCldn4 NIIRDFYNPLVVSGQKREMGAS hCldn6 AVIRDFYNPLVAEAQKRELGAS mCldn6 SIIQDFYNPLVADAQKRELGAS hCldn7 QIVTDFYNPLIPTNIKYEFGPA mCldn7 QIVTDFYNPLTPMNVKYEFGPA hCldn8 AIIRDFYNSIVNVAQKRELGEA mCldn8 SIIRDFYNPLVDVALKRELGEA cCldn8 TIIRDFYNPVVNVAQKRELGEA hCldn14 DVVQNFYNPLLPSGMKFEIGQA mCldn14 DVVQNFYNPLLPSGMKFEIGQA cCldn14 DVVTDFYNPLLPQGMKYEIGQA Slightly sensitive hCldn1 RIVQEFYDPMTPVNARYEFGQA mCldn1 RIVQEFYDPLTPINARYEFGQA cCldn1 RVARAFYDPFTPVNTRFEFGSA hCldn2 ILRDFYSPLVPDSMKFEIGEA mCldn2 ILRDFYSPLVPDSMKFEIGEA cCldn2 VLRDFHNPLLPDSTKFEMGEA Insensitive hCldn5 IVVREFYDPSVPVSQKYELGAA mCldn5 IVVREFYDPTVPVSQKYELGAA cCldn5 IVISDFYDPSVPPSQKREIGAA hCldn10 KITTEFFDPLFVEQKYELGAA mCldn10 KITTEFFDPLYMEQKYELGAA cCldn10 RITSEFFDPSFVAQKYELGAA

The amino sequences of the second extracellular loops of human (h), mouse (m) and chicken (c) Cldn1, -2, -3, -4, -5, -6, -7, -8, -10 and -14 are shown. Claudins are classified into three categories based on CPE sensitivity. Residues corresponding to the predicted CPE binding consensus motif are shown in red.

2 CHAPTER II: Claudins are essential for cell shape changes and convergent extension movements during neural tube closure

Amanda I. Baumholtz, Annie Simard, Evanthia Nikolopoulou, Marcus Oosenbrug, Michelle M. Collins, Anna Piontek, Gerd Krause, Jörg Piontek, Nicholas D. E. Greene, and Aimee K. Ryan

Published in Developmental Biology, 2017, 428 (1): 25-28.

2.1 ABSTRACT During neural tube closure, regulated changes at the level of individual cells are translated into large-scale morphogenetic movements to facilitate conversion of the flat neural plate into a closed tube. Throughout this process, the integrity of the neural epithelium is maintained via cell interactions through intercellular junctions, including apical tight junctions. Members of the claudin family of tight junction proteins regulate paracellular permeability, apical-basal cell polarity and link the tight junction to the actin cytoskeleton. Here, we show that claudins are essential for neural tube closure: the simultaneous removal of Cldn3, -4 and -8 from tight junctions caused folate-resistant open neural tube defects. Their removal did not affect cell type differentiation, neural ectoderm patterning nor overall apical-basal polarity. However, apical accumulation of Vangl2, RhoA, and pMLC were reduced, and Par3 and Cdc42 were mislocalized at the apical cell surface. Our data showed that claudins act upstream of planar cell polarity and RhoA/ROCK signaling to regulate cell intercalation and actin-myosin contraction, which are required for convergent extension and apical constriction during neural tube closure, respectively.

2.2 INTRODUCTION The molecular interactions that regulate many of the morphogenetic changes required for neural tube closure occur at the apical cell surface (Colas and Schoenwolf, 2001; Lawson et al., 2001). Apical localization of Wnt/planar cell polarity (PCP) pathway components is required for the convergent extension movements that lead to anterior-posterior elongation and medial-lateral narrowing of the neural plate (Goto and Keller, 2002; Wallingford and Harland, 2002; Wang et al., 2006a). Apical constriction of cells at the midline to form the median hinge point depends upon the actin-myosin contractile force driven by localized myosin light chain phosphorylation downstream of RhoA-ROCK signaling (Kinoshita et al., 2008; Suzuki et al., 2012). This allows the neural plate to bend and the neural folds to elevate. During the final phase of neural tube closure, Cdc42 RhoGTPase signaling regulates the formation of apical protrusions which are essential for the epithelial remodeling at the boundary of the neural and non-neural ectoderm to form the closed neural tube and a continuous layer of overlying surface ectoderm (Lawson et al., 2001; Moury and Schoenwolf, 1995; Pai et al., 2012; Quiros and Nusrat, 2014).

Apical tight junctions are a hallmark of vertebrate epithelial cell layers. They maintain apical-basal polarity by preventing the mixing of apical and basolateral membrane proteins and regulate the paracellular movement of ions, solutes and water (Gunzel and Yu, 2013; Suzuki et al., 2015). The claudin family of tetraspan membrane proteins are essential for formation of the tight junction backbone through homo- and heteropolymerization: their four transmembrane helix bundles interact within one cell membrane and their two extracellular loops interact with those of claudins in the apposing cell (Daugherty et al., 2007; Gunzel and Yu, 2013; Krause et al., 2015). While the claudin extracellular loops define the paracellular barrier properties of the junction, their cytoplasmic C-termini interact with adaptor and scaffolding protein complexes at the tight junction cytoplasmic plaque. Thus, claudins are poised at the apical surface to bridge intercellular interactions to cytoplasmic regulatory events that affect cell behaviours.

Our previous analysis of claudin expression patterns in chick embryos revealed that 14 claudins are expressed during neurulation (Collins et al., 2013). Neural tube defects have not been reported in any of the single claudin knockout mouse lines (Anderson et al., 2008; Ben- Yosef et al., 2003; Fujita et al., 2012; Furuse et al., 2002; Gow et al., 1999; Kage et al., 2014; Li et al., 2014; Miyamoto et al., 2005; Morita et al., 1999; Tamura et al., 2008). However, Cldn4, -

6, and -7 are downregulated in Grhl2 mutant mouse lines, which exhibit open neural tube defects due to a failure in the final stage of neural tube closure (Pyrgaki et al., 2011; Rifat et al., 2010; Werth et al., 2010). These data suggest that claudins may have functionally redundant roles during neural tube closure.

The C-terminal domain of C. perfringens enterotoxin (C-CPE) is an ideal tool for studying claudin redundancy. Full-length CPE, a common cause of food poisoning, crosses the intestinal barrier at tight junctions by binding to the second extracellular loop of Cldn4, which was first cloned as the CPE cell-surface receptor (CPE-R) (Katahira et al., 1997a). CPE also interacts with Cldn3, -6, -7, -8, -9, -14 and -19 (Katahira et al., 1997b; Shrestha and McClane, 2013; Winkler et al., 2009) but not with other cell surface proteins (Katahira et al., 1997b; Lohrberg et al., 2009). C-CPE contains the claudin binding domain but not the cytotoxic domain, which is located in the N-terminal region of CPE. C-CPE can bind to and internalize the same subset of claudins without affecting the localization of other claudin family members (Sonoda et al., 1999; Winkler et al., 2009). C-CPE was used to demonstrate the redundant functions of Cldn4 and -6 in mouse blastocysts: mice null for only Cldn4 (Fujita et al., 2012) or only Cldn6 (Anderson et al., 2008) are viable but removal of both claudins by C-CPE caused a loss of the hydrostatic pressure that maintains blastocoel shape (Moriwaki et al., 2007).

Here, we determined that C-CPE-sensitive claudins are essential for neural tube morphogenesis. Targeted removal of Cldn3, -4 and -8 from the neural and non-neural ectoderm of neural plate stage chick embryos resulted in folate-resistant open neural tube defects (NTDs). We showed that C-CPE-sensitive claudins were required for convergent extension movements and apical constriction of cells at the median hinge point. Furthermore, our data suggest that claudin family members differentially regulate localization of components of the PCP polarity complex and RhoGTPase signaling to the apical cell surface.

2.3 RESULTS

2.3.1 C-CPE-sensitive claudins are required for neural tube closure in chick embryos To simultaneously remove multiple claudins from tight junctions during neurulation we used the nontoxic GST-C-CPE (hereafter referred to as C-CPE) reagent and compared its effects

to either GST alone or C-CPEYL, a variant of C-CPE that does not bind claudins (Protze et al., 2015). Previously we showed by whole mount in situ hybridization that Cldn3, -4, -8 and -14 are the only C-CPE-sensitive claudins expressed during neural tube closure in chick embryos. We first confirmed that the protein expression patterns of these claudins during neural tube closure matched that of their transcripts (Collins et al., 2013). As expected Cldn4, -8 and -14 were expressed in the neural ectoderm, while Cldn3 was absent from the neural folds but was highly expressed in non-neural ectoderm (Suppl. Fig. 2.1). Next, we tested the ability of C-CPE to effectively remove these claudins from tight junctions as compared to effects on Cldn1, which does not interact with C-CPE (Fig. 2.1A,B). In GST-treated embryos, all five claudins co- localized with the tight junction scaffolding protein ZO-1 at apical cell-cell contacts in the neural (Cldn1, -4, -8, -14) and non-neural (Cldn1, -3, -4, -14) ectoderm (Fig. 2.1B). Co-localization analysis using Pearson’s correlation coefficient confirmed that Cldn1 (R=0.6267), -3 (R=0.5583), -4 (R=0.5867), -8 (R=0.5070), and -14 (R=0.7156) co-localized at tight junctions with ZO-1, which was used as a marker of tight junctions. After 5h of C-CPE treatment, only Cldn1 (R=0.6975) and Cldn14 (R=0.6083) remained co-localized with ZO-1 at tight junctions; localization of Cldn3 (R=0.09563), -4 (R=0.09) and -8 (R=0.2519) was discontinuous and often absent (Fig. 2.1B). Similar effects were observed after 20h (data not shown). The unexpected observation that Cldn14 remained localized to tight junctions in C-CPE-treated embryos may reflect context-dependent sensitivity to C-CPE. As predicted, C-CPEYL had no effect on the localization of Cldn3, 4 or 8 (Fig. 2.1C). To determine if C-CPE-sensitive claudins are required for neural tube closure, HH4 neural plate stage embryos were cultured ex ovo in GST or in C-CPE media for 20h. GST-treated embryos and embryos treated with the C-CPEYL variant were indistinguishable from wild-type embryos grown in ovo (Fig. 2.1D). C-CPE-treatment did not affect embryo viability: at 20h their hearts were beating, of normal size and exhibited normal rightward looping (Movie 1; Suppl. Fig. 2.2). However, C-CPE-treated embryos showed a dose-dependent increase in the incidence of open NTDs (Fig. 2.1E). NTDs were characterized as ‘complete’ when the opening was along the entire anterior-posterior axis, ‘caudal’ when the opening was posterior to the hindbrain or ‘cranial’ when the opening was in the region of the future brain (Fig. 2.1D,E). The lowest dose of C-CPE that caused NTDs in 100% of embryos (200 μg/ml) was used for all subsequent experiments. Folic acid supplementation, which reduces the incidence of NTDs in humans by

60-70% (van der Linden et al., 2006) and rescues NTDs induced in chick embryos (Guney et al., 2003; Weil et al., 2004), was unable to rescue the NTDs in C-CPE-treated embryos (n=24; Fig. 2.1F), suggesting that the C-CPE-induced NTDs are a model of folate-resistant NTDs.

2.3.2 Evolutionarily-conserved requirement for claudins in neural tube closure in mice To determine if removal of C-CPE-sensitive claudins would affect neural tube closure in mouse embryos, C-CPE was injected into the amniotic cavity of embryos that were then cultured for approximately 18h (Fig. 2.2A). Among embryos injected at the 0-4 somite stage, exposure to 0.5 or 1 mg/ml led to shortening of the caudal end of the embryo and a ‘wiggly’ neural tube suggestive of caudal elongation defects and a proportion of these embryos (n=4/7) also failed to undergo axial turning. As these embryos did not progress to a stage where neural tube closure could be reliably assessed due to other confounding phenotypes, we focused our analysis on embryos in which treatment was performed after initiation of neural tube closure at the 9-13 somite stage. Treated embryos appeared healthy and viable after the culture period; yolk sac circulation and mean number of somites in C-CPE-treated embryos did not significantly differ from controls. Spinal neural tube closure continued to progress in all treatment groups. However, the open region of spinal neural folds (posterior neuropore) was significantly larger in embryos treated with C-CPE (Fig. 2.2C,D) than in GST controls (Fig. 2.2B), indicating suppression of closure. Moreover, cranial NTDs were also observed in embryos exposed to 0.5 (n=1/8) and 1 mg/ml (n=3/12) C-CPE (Fig. 2.2D). These data indicate that C-CPE-sensitive claudins are also required for neural tube closure in mice embryos. The ‘milder’ NTDs observed in the C-CPE- treated mouse embryos most likely reflects the fact that neural tube closure was already initiated at the time of treatment

2.3.3 Depletion of C-CPE-sensitive claudins does not affect ectoderm differentiation Differentiation of the ectoderm into neural and non-neural progenitors is the first step of neurulation and disrupting this process can affect neural tube closure (Kimura-Yoshida et al., 2015; Ybot-Gonzalez et al., 2007b). To determine if the open neural tube defects caused by C- CPE-treatment were downstream of effects on cell type differentiation, we examined gene expression patterns in embryos treated with C-CPE for 20 hours (Suppl. Fig. 2.3). Despite the clear morphological defects following C-CPE treatment, all genes examined exhibited spatially- restricted expression boundaries that were similar to those observed in GST-treated control

embryos. Sox2 was expressed in the neural ectoderm and AP2 was expressed in the non-neural ectoderm indicating that the initial differentiation event that distinguishes neural versus non- neural progenitors occurred normally (Suppl. Fig. 2.3A-D). Otx2 and Pax6 exhibited normal boundaries of expression along the anterior-posterior axis (Suppl. Fig. 2.3E-H). Finally, differentiation of ventral versus dorsal cell types in the neural tube was also initiated normally in C-CPE-treated embryos based on the ventrally- and dorsally-restricted expression domains of Pax6 and Pax7, respectively, in the open neural tubes (Suppl. Fig. 2.3G-J). Thus, failure of the neural tube to close in C-CPE-treated embryos is not due to a global defect in differentiation or patterning of the ectoderm. Furthermore, we did not observe a difference in cell death between GST- and C-CPE-treated embryos at 5h and 20h (p>0.05) (Suppl. Fig. 2.4 and data not shown).

2.3.4 Claudins are required during neural plate shaping and median hinge point formation. To identify the phases of neural tube closure dependent on C-CPE-sensitive claudins, we performed media swap experiments. C-CPE does not act at the level of claudin transcription or translation, therefore, upon its removal C-CPE-sensitive claudins are able to re-populate tight junctions (Sonoda et al., 1999). All embryos transferred from C-CPE to GST media at 5h and collected at 20h had closed neural tubes (Fig. 2.3A). However, ~50% of the embryos transferred from C-CPE to GST at 10h had cranial NTDs. In a complementary experiment, we found that embryos transferred from GST to C-CPE at 5h had either complete or caudal NTDs (Fig. 2.3A), while none of the embryos transferred from GST to C-CPE at 10h exhibited NTDs. These data demonstrate that C-CPE-sensitive claudins are required for the phases of neural tube closure occurring between 5h and 10h in our culture system. To identify the morphological events occurring between 5h and 10h, we incubated HH4 embryos with GST or C-CPE for 5h, 10h, or 20h (Fig. 2.3B). At 5h, GST-treated and C-CPE- treated embryos were HH6-8. In HH6 GST embryos, elongation of the neural plate and formation of the neural groove had begun and by HH8 the neural folds were elevated and met at the level of the midbrain (Fig. 2.3B). In contrast, C-CPE-treated HH6 embryos had broader neural grooves and their neural folds failed to meet at the level of the midbrain at HH8 (Fig. 2.3B). After 10h of culture, the GST-treated control embryos were HH8+-10, had 5-10 pairs of somites and their neural folds were fused at the level of the brain and apposed at the anterior end

of the spinal cord. Embryos cultured in C-CPE for 10h were clearly distinguishable as their neural folds failed to converge towards the midline. After 20h, GST-treated embryos had closed neural tubes, while all C-CPE-treated embryos had open neural tubes (Fig. 2.1D) and were significantly shorter than GST control embryos (Fig. 2.1D, 4A). Cldn4 and -8 are the only C-CPE-sensitive claudins expressed in the neural ectoderm during these early phases. While we have successfully used morpholinos to knockdown expression of claudins in gastrulation-stage chick embryos (Collins et al., 2015), we and others have been unsuccessful at using this approach in neurulation-stage embryos (data not shown) (Colas and Schoenwolf, 2003a). Therefore, we took advantage of a C-CPE variant that specifically targets Cldn4, C-CPELSID (Veshnyakova et al., 2012). C-CPELSID removed Cldn4 from tight junctions in the neural and non-neural ectoderm, had no effect on the localization of Cldn3 in non-neural ectoderm, and led to increased levels of Cldn8 at tight junctions in the neural ectoderm (Suppl. Fig. 2.5A). Embryos treated with C-CPELSID did not have NTDs (Suppl. Fig. 2.5B). Based on these data we hypothesize that either Cldn4 is not essential for neural tube closure or that the increased localization of Cldn8 to tight junctions compensates for the loss of Cldn4. Unfortunately, a C-CPE variant that specifically targets removal of Cldn8 from tight junctions does not exist and so we cannot distinguish between these two possibilities at the present time.

2.3.5 C-CPE-treated embryos have convergent extension and apical constriction defects The morphology of the C-CPE-treated embryos was consistent with phenotypes caused by defective convergent extension movements (Lawson et al., 2001; Moury and Schoenwolf, 1995; Ybot-Gonzalez et al., 2007b). First, C-CPE-treated embryos were shortened along their anterior-posterior axes (p<0.0001), wider (p<0.001) and displayed a reduced length-to-width ratio (p<0.001) at all stages of neurulation examined (Fig. 2.4A,B). Second, the somites of 20- 40h C-CPE-treated embryos were irregularly shaped and sometimes fused (Fig. 2.4C-F). Third, the notochord marker Brachyury (T) had a broadened posterior expression domain and the notochords were misshapen in 50% of the C-CPE-treated embryos (Fig. 2.4G). Finally, we observed an effect on oriented cell division, which helps shape the neural plate (Sausedo et al., 1997). In GST-treated embryos, mitotic spindles in the neural ectoderm were preferentially oriented rostrocaudally (p=0.031), while in C-CPE-treated embryos there was a decrease in the

proportion of rostrocaudal divisions and corresponding increase in mediolateral divisions (p=0.125) (Fig. 2.4H). Together these data suggest that removal of C-CPE-sensitive claudins prevents normal convergent extension movements during neurulation, contributing to the increased distance between the neural folds thereby impeding neural fold fusion. The broadened neural groove in C-CPE-treated embryos suggested a failure in median hinge point formation. In GST-treated embryos, Shh exhibited a normal wedge-shaped expression domain at the neural plate midline reflecting the triangular shape of these cells (Fig. 2.5A). In C-CPE-treated embryos Shh expression was unaffected but the shape of its expression domain was rectangular rather than the triangular shape observed in control embryos (Fig. 2.5A). Indeed, transmission electron microscopy confirmed that midline neural tube cells in C-CPE- treated embryos were not apically constricted and triangular (Fig. 2.5B). The apical-to-basal width ratio of the midline cells was three times greater in C-CPE-treated embryos than in GST- treated embryos (p=0.012; Fig. 2.5C). Microtubule-dependent basal migration of the nuclei also contributes to the basal widening of the median hinge point cells (Karfunkel, 1971; Karfunkel, 1972; Smith and Schoenwolf, 1987). As expected, median hinge point cell nuclei were situated basally in GST-treated embryos (Fig. 2.5D). However, in C-CPE-treated embryos, these nuclei were randomly localized and on average closer to the apical surface than in GST-treated control embryos (p=0.0005) (Fig. 2.5D,E). In addition, the apical-basal microtubule network was discontinuous in C-CPE-treated embryos (Fig. 2.5F). Together these data demonstrate that C- CPE-sensitive claudins are required in the neural ectoderm for cell shape changes associated with median hinge point formation.

2.3.6 Claudins function upstream of PCP and RhoA/ROCK signaling Convergent extension and apical constriction during neurulation require the coordinated activities of the PCP and RhoA/ROCK signaling pathways, respectively (Kinoshita et al., 2008; Nishimura et al., 2012; Ossipova et al., 2015). In order to position claudins in the context of these pathways, we examined the expression of Vangl2, a core component of the PCP pathway that is required for neural tube closure (Kibar et al., 2011; Murdoch et al., 2001), RhoA, a small GTPase that shuttles between the cytoplasm and membrane, and phosphorylated myosin light chain (pMLC), the downstream target of RhoA/ROCK signaling. Vangl2 was greatly reduced after 5h of C-CPE-treatment (p=0.0024) (Fig. 2.6A). RhoA partly co-localized with ZO-1 and

Cldn14 at the membrane in the neural ectoderm of GST-treated embryos and its localization was unchanged in C-CPEYL-treated embryos (Fig. 2.6B, Suppl. Fig. 2.6A). However, after 5h of C- CPE-treatment, we observed the loss of RhoA from tight junctions and a significant reduction in phosphorylated myosin light chain (pMLC) in the neural ectoderm of C-CPE-treated embryos (Fig. 2.6B,C). Quantitative analysis revealed that although the total amount of RhoA did not change (p=0.7882), the correlation between RhoA and ZO-1 in C-CPE-treated embryos was ~40% that of GST-treated embryos (p=0.0052). Both the total level of pMLC (p=0.0023) and its distribution to tight junctions (p<0.0001) were significantly reduced in C-CPE-treated embryos. Incubating embryos with the ROCK inhibitors GSK 269962 and Y-27632 did not affect the expression or localization of Cldn3, -4 or -8 (Suppl. Fig. 2.7), although they did reduce the level of pMLC and cause NTDs as previously reported (Fig. 2.6C) (Escuin et al., 2015; Kinoshita et al., 2008). These data further support that claudins function upstream of PCP and Rho/ROCK signaling during neural tube closure.

Analysis of apical-basal polarity suggested that the decreased apical localization of Vangl2, RhoA and pMLC in C-CPE-treated embryos was not due to effects on the establishment or maintenance of apical-basal polarity. Cldn1, Cldn14, ZO-1 and 2, and F-actin were apically localized in C-CPE-treated embryos (Fig. 2.1, 5D and 6). In both GST- and C-CPE-treated embryos Cldn1 was localized apical to the adherens junction protein E-cadherin (Fig. 2.7A), Par3 co-localized with ZO-1 (Fig. 2.7B) and ZO-2 localized apical to the Scribble complex protein Dlg1 (Fig. 2.7C). These data suggest that localization and/or retention of Vangl2, RhoA and pMLC to the apical membrane is dependent on the presence of C-CPE-sensitive claudins in the neural tube.

2.3.7 C-CPE treatment affects cell morphology and protein localization at the apical surface. Apical surface views showed that claudins, ZO scaffolding proteins, E-cadherin and F- actin localized normally to the lateral cell membranes in C-CPE-treated embryos (Figs. 2B,C, 6A,B and 7A). This was not the case for Par3; although restricted to the apical domain, large aggregates of Par3 were present in the cytoplasm (Fig. 2.7D). Similarly, the RhoGTPase Cdc42 was observed across the apical surface and less enriched at lateral membranes as compared to embryos treated with GST or C-CPEYL (Fig. 2.7E, Suppl. Fig. 2.6B). Consistent with our visual

analysis, we found that the total amount of Par3 (p=0.107) and Cdc42 (p=0.523) was not affected by C-CPE treatment, but that their distribution was altered (p<0.05) (Fig. 2.7F). These data also support unique roles for individual claudins in maintaining protein complexes at the apical tight junction cytoplasmic plaque. TEM analysis of tight junction ultrastructure revealed that electron dense tight junctions between cells at the neural plate midline in C-CPE-treated embryos were indistinguishable from those in GST-treated embryos (Fig. 2.7G). However, in C-CPE-treated embryos the apical surface of these cells was smooth, convex and devoid of microvilli (Fig. 2.7G) as compared to cells in GST-treated embryos, which contained numerous microvilli. Changes in morphology of the apical cell surface may be secondary to the altered localization of Cdc42, which is a critical regulator of the microtubule network. Together these data revealed roles for C-CPE-sensitive claudins in regulating the shape of the apical cell surface.

2.4 DISCUSSION Claudins are known to have distinct roles in determining cell shape (Nelson et al., 2010; Wu et al., 2004), maintaining epithelial integrity (Ikenouchi et al., 2003) and embryonic morphogenesis (Siddiqui et al., 2010). In this study, we demonstrated for the first time that claudins have a direct role in neural tube closure: C-CPE removal of Cldn3, -4 and -8 from tight junctions in the neural and non-neural ectoderm resulted in NTDs in chick and mouse embryos. Although disrupting ectoderm differentiation can affect neural tube closure (Kimura-Yoshida et al., 2015; Ybot-Gonzalez et al., 2007b), gene expression analyses support that this is not the underlying cause of the C-CPE-induced NTDs. However, PCP and Par complex polarity proteins, and the small GTPases RhoA and Cdc42 were inappropriately localized at the apical cell surface in C-CPE-treated chick embryos. Consequently, convergent extension and apical constriction, two critical morphogenetic events that drive neural tube closure, were disrupted. Folic acid supplementation was unable to rescue the open NTDs in C-CPE-treated chick embryos suggesting that claudin-mediated morphogenetic events underlie at least a subset of NTDs not prevented by folate dietary supplementation and fortification.

2.4.1 Claudins in convergent extension movements Convergent extension, which coordinates anterior-posterior lengthening and mediolateral narrowing of the neural plate, is dependent on PCP signaling. C-CPE-treated embryos exhibited

hallmarks of convergent extension defects similar to those observed in the PCP mouse mutants Lp/Lp and Dvl1–/–;Dvl2–/– (Wang et al., 2006a; Ybot-Gonzalez et al., 2007b), suggesting that claudins and PCP signaling function in the same molecular pathway. The reduced expression of the core PCP protein Vangl2 in C-CPE-treated embryos positions claudins upstream of the PCP pathway in directing convergent extension cell movements. Furthermore, Vangl2 has been implicated in regulating anterior-posterior cell divisions during zebrafish neurulation (Ciruna et al., 2006) and we observed similar defects in oriented cell division in C-CPE-treated embryos. Therefore, we speculate that the loss of Vangl2 is responsible for the oriented cell division defects observed in these embryos. However, we cannot rule out the possibility that this is caused by disrupted localization of Par3 (Tawk et al., 2007) and Cdc42 (Kieserman and Wallingford, 2009) as they also play a role in oriented cell division. While a role for PCP- dependent convergent extension in avian neural tube closure has not been assessed, siRNA- mediated knockdown of the PCP component Celsr1 causes NTDs in chick embryos (Nishimura et al., 2012). This observation, along with the fact that overexpression and knockdown of PCP components causes convergent extension defects during neurulation in Xenopus (Goto and Keller, 2002; Wallingford and Harland, 2002) and zebrafish embryos (Tawk et al., 2007), argues for a conserved role of the PCP pathway in regulating convergent extension during neural tube closure. Convergent extension defects were not observed in C-CPE-treated mouse embryos because C-CPE treatment was initiated at E8.5 after convergent extension is underway and neural tube closure has initiated. Interestingly, caudal elongation defects were observed in mouse embryos that were treated at the 0-4 somite stage (EN and NDEG, unpublished observations). C-CPE removed claudins from the neural (Cldn4 and -8) and non-neural ectoderm (Cldn3 and -4), both of which undergo convergent extension movements. Decreased expression of Cldn3, -4, -6 and -7 in the non-neural ectoderm is also observed in Grhl2 mutant mice, which have folate-resistant NTDs. Grainyhead-like 2 (Grhl2) is a transcription factor that is expressed exclusively in the non-neural ectoderm where it regulates expression of genes involved in epithelial identity, including claudins (Brouns et al., 2011; Pyrgaki et al., 2011; Werth et al., 2010). Although the neural folds remain widely spaced in the Grhl2 mutants, convergent extension and apical constriction occur normally (Pyrgaki et al., 2011; Rifat et al., 2010; Werth et al., 2010). These data support our hypothesis that the convergent extension and apical

73 constriction defects observed in C-CPE-treated embryos are due to the depletion of C-CPE- sensitive claudins from the neural ectoderm.

2.4.2 Claudins function upstream of RhoGTPase signaling during neural tube closure Morphological, molecular and ultrastructure analysis of C-CPE-treated embryos revealed cells in the neural plate midline were not apically constricted and consequently the median hinge point did not form. Establishment of the apical membrane domain in epithelial cells depends on the coordination of the intracellular trafficking machinery, RhoGTPase signaling and the Par3- Par6-aPKC polarity complex. Despite a clear demarcation of the apical and basolateral domains in C-CPE-treated embryos, Par3 was abnormally localized in large aggregates. Knockdown of Par3 in HH12 chick embryos, following neural tube closure, results in closed NTDs (Chen et al., 2017). Our data suggest that Par3 also plays a role during earlier stages of neurulation in avian embryos. We presume that the aggregated Par3 either cannot fulfill its normal function and/or interferes with other protein trafficking at the apical surface. During neural tube closure, apical localization of RhoA-ROCK signaling components at the neural plate midline is required for phosphorylation of myosin light chain, which then moves along actin filaments to generate the contractile force required for apical constriction (Kinoshita et al., 2008; van Straaten et al., 2002). Chick embryos treated with C3, an inhibitor of Rho GTPases, develop NTDs (Kinoshita et al., 2008). Similarly, inhibiting Rho kinase (by Y27632) or myosin II motor activity (by blebbistatin) causes open NTDs in chick (Kinoshita et al., 2008) and mouse embryos (Escuin et al., 2015). In C-CPE-treated embryos, junctional localization of RhoA was reduced leading to reduced levels of pMLC. Treatment with ROCK inhibitors, which significantly decreased the level of pMLC and caused open NTDs (Escuin et al., 2015; Kinoshita et al., 2008), did not affect claudin expression or localization. These data indicate that claudins function upstream of RhoA-ROCK signaling in the neural ectoderm during neural tube closure and that recruitment and retention of RhoA at the cytoplasmic plaque is dependent on direct or indirect interactions with claudins expressed in the neural ectoderm. RhoA signaling is required for changing the claudin composition of tight junctions via endocytosis (Quiros and Nusrat, 2014). In Xenopus embryos endocytosis acts downstream of actin-myosin contraction to remove excess membrane from the apical surface; inhibiting endocytosis leads to neural tube defects due to defective apical constriction (Lee and Harland,

2010). Currently there is no evidence to suggest that endocytosis is required for apical constriction of neural ectoderm cells in the chick embryo. However, it is tempting to speculate that RhoA interactions with Cldn8 and/or Cldn4 contribute to remodeling of the apical cell surface in the chick neural ectoderm via endocytosis. Actin-myosin-dependent apical constriction in midline neural plate cells occurs in concert with apical-basal shortening and basal nuclear migration, which are dependent on an apically- basally aligned microtubule network. Nuclear migration was disrupted in C-CPE-treated embryos and Cdc42, a critical regulator of the microtubule network (Cadot et al., 2012; Palazzo et al., 2001b), was expressed diffusely across the apical surface rather than being enriched at the lateral membrane. We hypothesize that the altered localization of Cdc42 affects microtubule- dependent events during median hinge point formation. Defects in the microtubule cytoskeleton may also underlie the altered shape of the apical surface of midline cells in C-CPE-treated embryos, which were smooth, convex and devoid of the microvilli present in control embryos. Our data indicate that Cldn8 and/or Cldn4, C-CPE-sensitive claudins expressed in the neural ectoderm, uniquely contribute to the retention of Par polarity complex proteins and Cdc42 at the lateral cell membrane at the neural plate midline.

In conclusion, we have defined new roles for C-CPE-sensitive claudins in regulating cell movements and shape changes that are critical for neural tube morphogenesis. Our data support a model where disturbing the claudin composition of tight junctions impacts protein complexes at the cytoplasmic plaque. Previous studies have suggested that Vangl2 polarization and Rho- induced phosphorylation of myosin light chain participate in a mutually dependent feedback loop (Nishimura et al., 2012; Ossipova et al., 2015). Our data suggest that claudin function intersects with this pathway to regulate Vangl2 localization to the apical cell surface and phosphorylation of myosin light chain during convergent extension and apical constriction, respectively. Both of these events are intrinsic to the neural ectoderm, where Cldn4 and -8 are the only C-CPE- sensitive claudins. Cldn4 null mice have closed neural tubes (Fujita et al., 2012) and Cldn8 null mice have not yet been reported. In the mouse kidney Cldn4 and -8 are known to interact and Cldn8 is required for recruitment of Cldn4 to the tight junction (Gong et al., 2015b; Hou et al., 2010). Therefore, we hypothesize that Cldn8 collaborates with Cldn4 to participate in unique interactions with signaling complexes at the tight junction cytoplasmic plaque, which regulate

the morphogenetic movements and cell shape changes required for neural plate elongation and median hinge point formation. Claudin C-terminal domains are highly variable between family members and are uniquely regulated by phosphorylation and other post-translational modifications, which can influence their interactions with proteins in the tight junction cytoplasmic plaque (Findley and Koval, 2009). While individual claudins may not be required to maintain the apical domain or tight junction integrity, they are likely to each have unique roles in regulating intracellular events that affect the actin cytoskeleton and cell behavior, potentially in a cell-type or tissue-specific manner. Future studies will be needed to understand how claudins anchor these proteins to the apical cell surface, to determine if these interactions are direct or indirect, and how they are influenced by post-translational modifications.

Few of the more than 200 NTD mouse models have led to the identification of genetic mutations associated with human NTDs (Harris and Juriloff, 2010; Kibar et al., 2007). This is likely due to the complex etiology of the disease which results from gene-environment and gene- gene interactions. While single claudin knock-out mice do not exhibit NTDs, we showed that simultaneous removal of Cldn3, -4 and -8 from tight junctions results in NTDs that parallel the phenotypically diverse NTDs that can result from a common underlying gene defect in humans (Juriloff and Harris, 2012; Lei et al., 2013; Robinson et al., 2012). There are numerous examples of diseases where birth defects result from mutations in multiple genes that act synergistically (Kajiwara et al., 1994; Katsanis et al., 2001; Kim et al., 2016; Nanni et al., 1999). While there has been limited success in identifying mutations in the human orthologues of mouse candidate NTD genes, an approach that looks at multiple genes that interact in a signaling pathway may be more successful. Our discovery of roles for C-CPE-sensitive claudins in neural tube closure, and their intersection with pathways known to be involved in neural tube closure, including RhoGTPase and PCP signaling, suggests that the claudin family represents new candidates for NTDs in humans.

2.5 MATERIALS AND METHODS

2.5.1 Production of GST and GST-C-CPE fusion protein Expression of GST, GST fused N-terminally to the C-terminal amino acids 185-319 (Moriwaki et al., 2007) and the variants C-CPELSID (L254A/S256A/I258A/D284A) and C-CPEYL

(Y306A/L315A) cloned into pET14b or pGEX6P1 was induced by IPTG. GST fusion proteins were purified from E. coli BL21 as described previously (Protze et al., 2015; Veshnyakova et al., 2012) and dialyzed against PBS. Protein concentration was determined using the BCA Protein Kit (Thermo-Scientific, Rockford, USA) or Bio-Rad Protein Assay (BioRad, Mississauga, Canada).

2.5.2 Embryo culture Fertilized eggs (Couvoir Simetin, Mirabel, Canada) were incubated at 38.5°C. Embryos were collected, staged according to Hamilton and Hamburger (HH) criteria (Hamburger and Hamilton, 1951) and cultured using the modified Cornish pasty method (Nagai et al., 2011). Purified GST or C-CPE (50-500 µg/ml), ROCK inhibitors Y27632 (50μM; Abcam, Cambridge, UK) or GSK 269962 (1μM; Axon Medchem, Reston, USA), and folic acid (100µM or 1mM) were added directly to the Cornish pasty culture medium to achieve the final concentrations indicated. E8.5 mouse embryos (E0.5 is noon on the day after overnight mating) were dissected in Dulbecco’s Modified Eagle’s Medium containing 10% fetal calf serum. The yolk sac and amnion were left intact for whole embryo culture (Pryor et al., 2012). GST or C-CPE were diluted with PBS. Fast Green (0.5% solution) was added to visualize injection. Microinjection was performed using a hand-held glass micropipette. Needle tips were positioned in the amniotic cavity by traversing the yolk sac and amnion and 1-2µl were injected, as previously described (Abdul-Aziz et al., 2009). Following 18h of culture, embryos were examined. Yolk sac circulation was used as an indicator of viability, somites were counted as a measure of developmental progression, and length of the posterior neuropore was measured using an eyepiece graticule.

2.5.3 Immunofluorescence staining Immunostaining was performed on paraffin sections (for α-tubulin), cryosections or whole embryos fixed in 4% PFA or 10% trichloroacetic acid at 4°C. Samples were incubated overnight at 4°C with primary antibodies: Cldn1, -4 and -8 (Invitrogen, Carlsbad, USA, 1:25- 1:50), Cldn3 (Abcam, Cambridge, UK, 1:50), Cldn14 (Sigma Aldrich, Oakville, Canada, 1:25), ZO-1 and ZO-2 (Invitrogen, 1:50, Carlsbad, USA), anti-disphospho-myosin light chain (Thr18/Ser19) (Cell Signaling, Ipswich, USA, 1:50), RhoA and Cdc42 (Santa Cruz, Santa Cruz, USA, 1:50), Par3 (Millipore, Etobicoke, Canada, 1:250), Dlg1 (US Biologicals, Salem, USA,

1:50), E-cadherin (BD Transduction, San Jose, USA, 1:100), α-tubulin (Abcam, Cambridge, UK, 1:100) or Vangl2 (gift from Dr. M Montcouquiol, 1:500). Alexa Fluor-conjugated secondary antibodies (1:500) were added for 1h at RT. F-actin was detected using Alexa Fluor-conjugated Phalloidin (Molecular Probes, Eugene, USA). Sections and flat-mounted embryos were coverslipped with SlowFade Gold with DAPI (Molecular Probes, Eugene, USA) and imaged using a Zeiss LSM780 laser scanning confocal microscope. Colocalization and immunoquantification was performed using ZEN 2012 SP1 software (Carl Zeiss Microscopy, Germany).

2.5.4 Whole mount in situ hybridization Antisense digoxygenin-labeled riboprobes for Sox2, AP-2α, Pax7 (Khudyakov and Bronner-Fraser, 2009), Otx2 (Chapman et al., 2002; Chapman et al., 2003), Shh, Brachyury, and Pax6 were used as previously described (Collins and Ryan, 2011). Embryos were photographed using a Leica M125 dissecting microscope with Infinity Capture software v5.0.2 (Lumenera Corp.). Paraffin-embedded embryos were sectioned and photographed using a Zeiss Axiophot compound microscope and AxioCamMRc camera with Axiovision v4.7.1.0 software.

2.5.5 Transmission electron microscopy Specimens were prepared at the Facility for Electron Microscopy Research, McGill University. Embryos were fixed with 2.5% glutaraldehyde, 0.1% calcium chloride and 4% sucrose in 0.1M sodium cacodylate buffer overnight at 4°C, post-fixed with 1% aqueous osmium and 1.5% potassium ferrocyanide in 0.1M sodium cacodylate for 2h at 4°C, dehydrated with a graded series of acetone, and embedded in epon. 100nm sections were cut using an UltraCutE ultramicrotome (Reichert-Jung) and stained with uranyl acetate and Reynold’s lead. Images were acquired on an FEI Techai 12 120kV transmission electron microscope equipped with an AMT xR80C 8 megapixel CCD camera.

2.5.6 Morphometric assessment of axial length, length-to-width ratios, apical constriction and oriented cell division. Dorsal images of embryos were photographed using a Leica M125 dissecting microscope with Infinity Capture software v5.0.2 (Lumenera Corp.) and imported into SPOT Advanced image capture software (SPOT Imaging Solutions). Width of the neural folds or neural tube at the level of the first somite and anterior-posterior length were measured. To assess apical constriction, individual cells of the median hinge point were examined in transverse ultrathin sections processed for transmission electron microscopy from embryos treated for 20h with GST or C-CPE. Cell surfaces were measured using ImageJ software. Basal nuclear migration was determined by drawing a perpendicular line from the apical surface to the apical tip of each DAPI-positive nucleus in GST- and C-CPE-treated embryos at 5h. To quantify the orientation of cell division, z-stacks of embryos stained with DAPI were exported into ImageJ. A line was drawn between the mitotic spindles of anaphase cells and the angle between this axis and the midline of embryos was measured. The angles of divisions were then exported to Microsoft

Excel and converted into a 0° to 90º range plot as previously described (Shafer et al., 2017). For statistical purposes, spindle orientation measurements were separated into three bins: 0-30º (rostral-caudal), 30-60º (diagonal) or 60-90º (medial-lateral).

2.5.7 Whole mount TUNEL assay GST- or C-CPE-treated embryos (200µg/ml) were fixed with 4% PFA overnight at 4ºC and assessed using the in situ cell death detection kit, AP (Roche, Montreal, Canada). Neural and non-neural ectoderm of each embryo was imaged at the level of the hindbrain and first and last somites. Cell death index was calculated by dividing the number of TUNEL-positive nuclei by the total number of cells, as determined by Phalloidin staining in each 125µm by 125µm region.

2.5.8 Statistical Analyses Statistical significance was evaluated using the Mann-Whitney U-test, student t-test or Wilcoxon’s signed-rank test through GraphPad InStat software or SigmaStat v.2 (SPSS Inc.).

2.6 ACKNOWLEDGEMENTS We thank Drs. K. Christensen C. Goodyer, I. Gupta, L. Hermo, L. Jerome-Majewska, L. McCaffrey, R. Rozen, P. Siegel and Y. Yamanaka for helpful discussions, M. Furuse for GST-C- CPE construct, M. Bronner for Pax7, S. Chapman for Otx2, and J. Mui (McGill University FEMR), M. Fu (RI-MUHC Imaging Platform) and D. Savery (UCL) for technical assistance. The authors acknowledge the facilities and the scientific and technical assistance of the Facility for Electron Microscopy Research, McGill University. AIB is the recipient of a doctoral studentship from Fonds de recherche du Québec – Santé (FRQS) (29939). This work was supported by the UK Medical Research Council (NDEG), the Deutsche Forschungsgemeinschaft (grant number 1273/3-2 to DFG, KR and grant number PI 837/2-1 to GK, JP), the Natural Sciences and Engineering Research Council of Canada (234319) (AKR), and Canadian Institutes of Health Research (MOP-84583) (AKR). AKR is a member of the RI-MUHC, which is supported in part by the FRQS.

2.7 COMPETING INTERESTS The authors have no competing interests.

Figure 2.1 C-CPE-treated embryos exhibit dose-dependent, folic acid resistant neural tube defects. (A) Dorsal view of a neural groove stage embryo. Dashed line outlines the neural plate. The areas of the neural (box 1) and non-neural (box 2) ectoderm imaged in (B) are shown. (B) Apical surface view of ZO-1 (red) and Cldn1, -4, -8 or -14 (green) in the neural ectoderm and Cldn-3 (green) in non-neural ectoderm of 5h GST- or C-CPE-treated embryos. Three embryos per treatment were analyzed. Scale bar, 10 µm. (C) Apical surface views of ZO-1 (red) and Cldn3, -4 or -8 (green) in embryos treated with C-CPEYL for 5h. Three embryos per treatment were analyzed. Scale bar, 10 µm. (D) Dorsal views of chick embryos treated with 200 µg/ml GST, C-CPEYL, or C-CPE for 20h. Dashed lines indicate open neural tubes. Scale bar, 0.2 mm. (E) Distribution of complete, cranial and caudal open NTDs following 20h incubation in 200 or 500 µg/ml GST or 50, 100, 200 or 500 μg/ml C-CPE. (F) Dorsal views of embryos treated with 200 µg/ml GST or C-CPE in the presence of 0 μM, 100 μM, or 1 mM folic acid. Dashed lines indicate open neural tubes. Scale bar, 0.2 mm.

Figure 2.2 C-CPE prevents closure of the posterior neuropore in mouse embryos. (A-D) Cultured mouse embryos injected with GST or C-CPE into the amniotic cavity prior to culture. (A) Blue dye indicates retention of solution in yolk sac during 18-20h culture period. (B- D) Representative embryos injected at E8.5 with (B) GST, (C) 0.5 mg/ml C-CPE or (D) 1.0 mg/ml C-CPE. Dashed lines indicate normal neuropore (B’) and closed neural tube (B’’) in GST-treated embryos and enlarged neuropore in (C,C’) and open hindbrain neural folds (cranial NTD) in (D,D’) in C-CPE-treated embryos. (E) The length of the posterior neuropore relative to the somite number in individual embryos at the end of the culture period is shown. The mean (± s.e.m.) for each treatment group is shown in the red rectangle. (*p<0.025, ANOVA).

Figure 2.3 C-CPE-sensitive claudins are required for apposition of neural folds. (A) Time course of media swap (left) and phenotypes of neural tube defects at the end of the 20h of culture period (right). Numbers of embryos per phenotype are indicated. (B) Dorsal views of embryos cultured from the neural plate stage (HH4) in 200 µg/ml GST or C-CPE for 5h or 10h. Double-headed arrows indicate distance between midbrain neural folds (inset). Dashed lines outline open NTDs. Scale bar, 0.2 mm.

Figure 2.4 C-CPE-treated embryos exhibit convergent extension defects. (A) Graph shows average anterior-posterior (AP) length of GST- and C-CPE-treated embryos at different stages of neural tube closure (mean±s.e.m, ***p<0.0001, Mann-Whitney U-test). Corresponding Hamilton and Hamburger (HH) stages are indicated below. (B) Graph shows length-to-width ratio for GST- and C-CPE-treated embryos at neural fold and neural tube stages (mean±s.e.m., ***p<0.0001, Mann-Whitney U-test). (C-F) Dorsal views of embryos treated with 200 µg/ml GST (C, D) or C-CPE (E, F) for 20h. Brightfield images (C,E) and Phalloidin- stained (D,F) embryos are shown. (C’-F’) Higher magnification images showing paired somites in GST-treated embryos (C’,D’) and unpaired and fused somites in C-CPE-treated embryos (E’,F’). (G) Left, dorsal view of Brachyury (T) expression following 20h treatment with 200 µg/ml GST or C-CPE for 20h. Scale bar, 0.5mm. Right, transverse section corresponding to position indicated. Bottom images are higher magnification images of sections. Scale bar, 50µm. nc, notochord; nt, neural tube; so, somite. (H) Rose diagram showing orientation of cell divisions relative to the midline in GST- (n=6 embryos/810 cells) and C-CPE-treated embryos (n=4 embryos/629 cells) at 5h. Division angles are binned in bins of 10º from 0º to 90º. Purple bars represent rostrocaudal (RC) divisions, blue are mediolateral (ML) divisions, and green are diagonal (D) divisions. In GST-treated embryos, mitotic spindles are oriented preferentially in the RC plane compared to the ML plane (p=0.031) in C-CPE-treated embryos cell division was random (p=0.125) (Wilcoxon’s signed-ranks test).

Figure 2.5 Neural plate midline cells are not apically constricted in C-CPE-treated embryos. (A) Transverse sections showing Shh expression in 5h and 20h GST- and C-CPE-treated embryos. Dashed line in higher magnification view marks Shh expression domain in floor plate. Scale bar, 50µm. (B) Ultrathin sections through the neural tube of 20h GST- and C-CPE-treated embryos processed for TEM. Nuclei are indicated by asterisks. Schematic representation is shown below. Apical and basal cell surfaces are outlined in red. (C) Graph illustrates apical-to- basal width ratio of median hinge point cells from embryos treated with GST (n=8 cells/2 embryos) or C-CPE (n=10 cells/2 embryos) (mean±s.e.m, *p =0.012), student t-test). Scale bar, 5µm. (D) Transverse sections of 5h GST- or C-CPE-treated embryos stained for Cldn1 (green) and DAPI (blue). Midline (white arrowhead) and nuclei (asterisks) are indicated. Scale bar, 10µm. (E) Graph shows average distance of nuclei from apical cell surface in GST- (n=108 cells/2 embryos) and C-CPE-treated embryos (n=212 cells/3 embryos) (mean±s.e.m, ***p=0.0005, student t-test). (F) Top, transverse sections of 5h GST- or C-CPE-treated embryos stained for α-tubulin (green) and DAPI (blue). Bottom, high magnification view of microtubules in the median hinge point of GST- and C-CPE-treated embryos. Scale bar, 100 µm.

Figure 2.6 PCP and RhoGTPase signaling proteins are reduced at tight junctions in C- CPE-treated embryos. (A) Apical surface and orthogonal (XZ) views of Vangl2 (green) and F-actin (Phalloidin, red) in neural ectoderm of GST- or C-CPE-treated embryos at 5h. Scale bar, 10 µm. Graph shows the ratio of the immunofluorescence intensity of Vangl2 to F-actin. Histogram represents an average of three embryos and values were calculated from maximum intensity projections (mean±s.e.m, **p=0.0024, student t-test). (B) Apical surface and orthogonal (XZ) views of Cldn14 (pink), ZO-1 (red), and RhoA (green) in the neural ectoderm of GST or C-CPE-treated embryos at 5h. Scale bar, 10 µm. Graphs show the ratio of the immunofluorescence intensity of RhoA to ZO-1 (upper), and Pearson’s correlation coefficient for RhoA and ZO-1 (lower). Histograms represent an average of three different fields from three embryos and values were calculated from maximum intensity projections (mean±s.e.m, ns = not significant, **p=0.0052, student t-test). (C) Transverse sections of embryos treated with GST, C-CPE or Y27632 for 5h showing pMLC (green) and ZO-1 (red). Scale bar, 100µm (left and middle) and 10µm (right). Graphs show the ratio of the immunofluorescence intensity of pMLC to ZO-1 (upper), and Pearson’s correlation coefficient for pMLC and ZO-1 (lower). Data were collected from the enclosed areas of sectioned neural plates. Histograms represent an average of three embryos and values were calculated from maximum intensity projections (mean±s.e.m, **p=0.0023, ***p<0.0001, student t-test).

Figure 2.7 C-CPE-sensitive claudins are required for Par3 and Cdc42 localization to the lateral membranes at the apical cell surface. (A) Apical surface and XZ orthogonal views of Cldn1 (red) and E-cadherin (green) in 5h GST- and C-CPE-treated embryos. Scale bar, 20 µm. (B) Transverse section from 5h GST- and C- CPE-treated embryos showing Par3 (green) and ZO-1 (red). Scale bar, 20µm. (C) Transverse section from from 5h GST- and C-CPE-treated showing ZO-2 (red) and Dlg1 (green). Scale bar, 10µm. (D) Apical surface and XZ orthogonal views of Par3 (green) and F-actin (Phalloidin, red) in the neural ectoderm of 5h GST- or C-CPE-treated embryos. Scale bar, 10µm. (E) Apical surface and XZ orthogonal views of Cldn14 (pink), ZO-1 (red) and Cdc42 (green) in the neural ectoderm of 5h GST- or C-CPE-treated embryos. Scale bar, 10µm. Three embryos per treatment were analyzed. (F) Graphs show the ratio of the immunofluorescence intensity of Par3 to F-actin or Cdc42 to ZO-1 (upper), and Pearson’s correlation coefficient for Par3 and F-actin or Cdc42 and ZO-1 (lower). Histograms represent an average of three different fields from three embryos and values were calculated from maximum intensity projections (mean±s.e.m, ns = not significant, *p=0.0132, **p=0.0072, student t-test). (G) TEM micrographs of transverse sections at the neural tube midline of 5h and 20h GST- and C-CPE-treated embryos. Arrows indicate tight junctions and arrowheads indicate microvilli. Scale bar, 500nm.

Movie 2.1 The heart of C-CPE-treated embryos is beating. Movie showing the beating heart of a C-CPE-treated embryo (left) cultured ex ovo using the Cornish Pasty method for 20h.

Supplemental Figure 2.1 Expression of claudins at neurulation in chick embryos. (A) Dorsal views of chick embryos at the different stages of neural tube closure. Dashed lines indicate the level of sections shown in B. Scale bar, 0.5mm. (B) Immunostaining for Cldn3, -4, -8 and -14 (green) on transverse sections of chick embryos during neurulation corresponding to the level of the dashed lines in A. Nuclei are stained with DAPI (blue). Asterisk indicates the midline. ne, neural ectoderm; nne, non-neural ectoderm. Scale bar, 100μm.

Supplemental Figure 2.2 Embryos treated with C-CPE are viable. (A, C) Dorsal view of embryos treated with 200µg/ml GST (A) or C-CPE (C) for 20h. Scale bar, 0.5 mm. (B, D) Ventral view showing the rightwardly looped heart tube of embryos shown in A and C. h, heart. Scale bar, 0.5 mm. (E, F) Troponin expression in embryos cultured in 200µg/ml GST (E) or C-CPE (F) for 40h. h, heart. Scale bar, 0.5 mm.

Supplemental Figure 2.3 Differentiation and patterning of the neural and non-neural ectoderm were not affected by C-CPE treatment. Wholemount in situ hybridization of embryos treated with GST or C-CPE for 20h with Sox2, which marks the neural tube and endoderm (A, B), AP-2α, which marks non-neural ectoderm (C, D), Otx2, which marks the anterior neural tube (E,F), Pax6, which marks the optic vesicles, the first rhombomere and the ventral neural tube (G,H), and Pax7, which marks the dorsal neural tube (I, J). A’-J’ are transverse sections through whole embryos at the level of the dashed line in A-J. Red arrowheads mark expression boundaries. mhp, median hinge point; ne, neural ectoderm; nne, non-neural ectoderm; nt, neural tube; ph, pharynx; s, somite. Scale bar, 0.2 mm.

Supplemental Figure 2.4 Embryos treated with C-CPE are viable and do not exhibit increased cell death. (A) Apical surface views of neural and non-neural ectoderm in embryos treated with GST or C- CPE for 5h and assayed for apoptosis by TUNEL assay. TUNEL-positive nuclei are red. Phalloidin (green) outlines cells. Three embryos per treatment were analyzed. Scale bar, 10μm. (B) Graph shows cell death index in the neural and non-neural ectoderm of embryos treated with GST or C-CPE for 5h (mean±s.e.m, p > 0.05, student-t-test).

Supplemental Figure 2.5 Cldn4-specific C-CPE variant, C-CPELSID, does not cause NTDs. (A) Apical surface views of Cldn3, -4 or -8 (green) and ZO-1 (red) in embryos treated with C- CPEYL or C-CPELSID for 5h. Scale bar, 10 µm. Three embryos per treatment were analyzed. (B) Dorsal view of representative embryos incubated with C-CPEYL (n=9) or C-CPELSID (n=12) for 20h. Higher magnifications of bracketed region are shown to the right. Scale bar, 0.5 mm.

Supplemental Figure 2.6 Localization of Cldn14, RhoA, and Cdc42 is not affected in embryos treated with C-CPEYL. (A) Confocal views of triple-immunostaining for Cldn14, ZO-1 and RhoA at the apical surface of neural ectoderm of embryos treated with GST or C-CPEYL for 5 hours. The bar graphs show Pearson’s correlation coefficient for RhoA and ZO-1 (p=0.182), RhoA and Cldn14 (p=0.608) or ZO-1 and Cldn14 (p=0.08). Histograms are the average of three different fields from three embryos (mean±s.e.m, student t-test). (B) Confocal views of triple-immunostaining for Cldn14, ZO-1 and Cdc42 at the apical surface of neural ectoderm of embryos treated with GST or C- CPEYL for 5 hours. The bar graphs show Pearson’s correlation coefficient for Cdc42 and ZO-1 (P=0.642), Cdc42 and Cldn14 (P=0.344) or ZO-1 and Cldn14 (P=0.652) (mean±s.e.m, student t- test).

Supplemental Figure 2.7 Inhibitors of ROCK signaling did not affect Claudin expression or localization. (A) Apical surface views of ZO-1 (red) and Cldn3, -4, -8 and -14 (green) expression in embryos incubated in DMSO (control) or GSK269962 for 5h. (B) Left, dorsal views of neural fold stage embryos treated with GST or Y27632 for 5h. Transverse sections showing Cldn4 or -14 (green) and ZO-1 (red) expression and apical surface view of an embryo co-stained for Cldn8 (green) and ZO-1 (red) expression.

2.8 CONNECTING TEXT BETWEEN CHAPTER II AND III

The role of claudins in fusion of the dorsal tips of the neural folds to form a closed neural tube could not be evaluated in C-CPE-treated embryos due to defects in earlier phases of neural tube closure. In Chapter III, I explore the role of Cldn3 in the non-neural ectoderm in regulating neural fold fusion.

3 CHAPTER III: Cldn3 is required in the non-neural ectoderm to regulate neural fold fusion

Amanda I. Baumholtz, Anna Piontek, Gerd Krause, Jörg Piontek, and Aimee K. Ryan

Manuscript in preparation

3.1 ABSTRACT Claudins are integral tight junction proteins that regulate paracellular permeability, apical- basal cell polarity, cell adhesion and link the tight junction to the actin cytoskeleton. Cldn3 is exclusively expressed in the non-neural ectoderm during neural tube closure. To ascertain the importance of Cldn3 in neural tube closure, we used a variant of the C-terminal domain of the Clostridium perfringens enterotoxin, C-CPELDR, which specifically targets Cldn3 and removes it from tight junctions. Cldn3-depleted embryos progress normally through the initital phases of neural tube closure. However, they fail to complete the final phase of epithelial remodeling during which the neural folds fuse and separate from the non-neural ectoderm to yield a closed neural tube and a continuous layer of overlying non-neural ectoderm. Analysis of Cldn3- depleted embryos revealed the non-neural ectoderm cells have normal apical-basal polarity, although the Par polarity complex protein Par3 is mislocalized at the apical surface of non-neural ectoderm cells. Furthermore, a fibrous extracellular matrix-based meshwork bridging non-neural ectoderm cells of the apposed neural folds was present in control but absent in Cldn3-depleted embryos. We have shown that Cldn3 is a critical tight junction protein required for neural tube closure.

3.2 INTRODUCTION The neural tube is the embryonic precursor of the central nervous system. Formation of a closed neural tube is highly complex involving a series of morphogenetic movements that must be coordinated in time and space. The neural tube initiates as a flat sheet of neuroepithelium that thickens apical-basally to form the neural plate, which is connected laterally to the non-neural ectoderm. The cells at the midline of the neural plate bend as the lateral edges rise generating neural folds which meet at the dorsal midline and fuse. Remodeling separates the neural and non-neural ectoderm from each other leading to formation of a closed neural tube and a continuous overlying layer of non-neural ectoderm.

Studies from animal models have shown that both the neural and non-neural ectoderm are required for neural tube closure. The intrinsic forces generated by the neural ectoderm that regulate neural tube closure have been extensively characterized and include changes in cell shape, cell rearrangement and oriented cell division. Apical constriction of midline neural ectoderm cells causes the neural plate to bend while convergent extension and anterior-posterior cell divisions contribute to lengthening and narrowing of the neural plate. The non-neural ectoderm is not only required for proper patterning of the dorsal neural tube and induction of neural crest cells, but it also regulates neural tube morphogenesis (Dickinson et al., 1995; Selleck and Bronner-Fraser, 1995). Studies in amphibian, chick and mouse embryos showed that removal of the non-neural ectoderm inhibits neural fold elevation and convergence towards the midline (Hackett et al., 1997; Jacobson and Moury, 1995; Morita et al., 2012; Moury and Schoenwolf, 1995). It has been proposed that this is due to a pushing force generated by the non-neural ectoderm against the neural ectoderm caused by changes in cell shape, position, and number. In chick and Xenopus, cells in the non-neural ectoderm change from cuboidal to squamous resulting in a decrease in height and increase in width. This changes in cell shape, along with medial-lateral cell divisions, generates a pushing force that causes the neural folds to bend inwards (Sausedo et al., 1997). In the avian embryo, cell rearrangement in the form of convergent extension causes the non-neural ectoderm to move towards the midline. Although these changes in the behavior of non-neural ectoderm cells are thought to generate extrinsic forces that drive neural fold elevation and convergence, this is purely speculative. Furthermore, even less is known about the molecular mechanisms that regulate these cell behaviors.

100

Our previous analysis of claudin expression patterns in chick embryos revealed that 14 claudins are expressed during neurulation and a subset of claudins are differentially expressed in the neural and non-neural ectoderm (Collins et al., 2013): Cldn8, -14 and -17 are enriched in the neural ectoderm and Cldn3, -11, -12 and -22 are more highly expressed in the non-neural ectoderm. Furthermore, we recently showed that targeted removal of Cldn3, -4 and -8 from the ectoderm using the C-terminal domain of Clostridium perfingens enterotoxin (C-CPE) results in open neural tube defects in chick and mouse embryos (Baumholtz et al., 2017b). Claudins are a family of ~24 tetraspan transmembrane proteins that form the backbone of tight junctions not only through the interaction of their extracellular loops between adjacent cells, but also by polymerization of their transmembrane domains within a cell. Members of the claudin family perform diverse functions including regulating paracellular permeability, apical-basal polarity and cell adhesion and they link the tight junction to the actin cytoskeleton and intracellular signaling molecules. Our previous study showed that claudins regulate convergent extension, apical constriction and oriented cell division in the neural ectoderm during neural tube closure; however, C-CPE sensitive claudins are also expressed in the non-neural ectoderm: Cldn4 is expressed in both neural and non-neural ectoderm, while Cldn3 is exclusively expressed in the non-neural ectoderm.

We previously showed that removing only Cldn4 had no effect on neural tube closure (Baumholtz et al., 2017b). Here, we studied the role of Cldn3 in the non-neural ectoderm using C-CPE variants that have altered claudin binding affinity. We found that targeted removal of Cldn3 from the non-neural ectoderm using a C-CPE variant that binds specifically to Cldn3 (C- CPELDR) resulted in open neural tube defects. Unlike embryos treated with wild-type C-CPE, convergent extension and apical constriction occurred normally, thus the inability of the neural folds to fuse at the midline appeared to be the primary cause of the neural tube defects. Furthermore, our data suggest that Cldn3 is required for formation of a mesh-like network that mediates the initial contact between apposed neural folds.

101

3.3 RESULTS

3.3.1 Cldn3 is required for neural tube closure in chick embryos We previously showed that simultaneous removal of Cldn3, -4 and -8 from the ectoderm of neural plate stage chick embryos using C-CPE causes neural tube defects (NTDs) in 100% of treated embryos due to defective convergent extension and apical constriction, events that are intrinsic to the neural ectoderm (Baumholtz et al., 2017b). However, C-CPE-sensitive claudins are also expressed in the non-neural ectoderm during neural tube closure. Cldn4 is expressed apically in the neural plate and adjacent non-neural ectoderm at HH6 and continues to be expressed in elevating neural folds at and overlying non-neural ectoderm at HH8 (Baumholtz et al., 2017b; Collins et al., 2013). At HH12, Cldn4 is strongly expressed in the overlying non- neural ectoderm cells but its expression is decreased in the closed neural tube. Cldn3 is highly expressed in the non-neural ectoderm at all stages of neural tube closure. How the loss of Cldn3 and -4 from tight junctions in the non-neural ectoderm affects neural tube closure could not be assessed in embryos treated with C-CPE due to the early defects in neural tube closure.

We showed that culturing chick embryos with GST, a C-CPE variant that binds specifically to Cldn4 (C-CPELSID) and a C-CPE variant that does not bind to claudins (C-CPEYL) does not affect the localization of Cldn3 and does not cause NTDs (Baumholtz et al., 2017b) (Fig 3.1A,B). Here, we tested a variant that was shown to specifically interact with Cldn3 in claudin- transfected HEK293 cells (C-CPELDR) (Veshnyakova et al., 2012). After 5 h of treatment with C-CPELDR, Cldn3 showed reduced localization to tight junctions while Cldn4 remained co- localized with the tight junction marker ZO-1 (Fig. 3.1A). To determine if removal of Cldn3 would cause NTDs, HH4 neural plate stage chick embryos were cultured ex ovo in media containing 200µg/ml C-CPELDR for 20 h (Fig. 3.1B). 83.3% of embryos treated with C-CPELDR exhibited wiggly spinal neural tubes that failed to close (n=20/24). 25% of embryos with spinal NTDs also exhibited cranial NTDs (n=6/24). The severity of the NTDs was not enhanced when C-CPELDR was increased to 400μg/ml (n=2) or when embryos were cultured with both C- CPELDR and C-CPELSID (n=7). These data show that Cldn3, but not Cldn4, is required in the non-neural ectoderm for neural tube closure. Folic acid supplementation, which prevents up to 70% of human NTDs (Viswanathan et al., 2017) and reduces the incidence of NTDs in chick embryos (Guney et al., 2003; Weil et al., 2004), was unable to prevent the open NTDs in C-

102

CPELDR-treated embryos (n=6) suggesting that C-CPELDR-induced NTDs are a model of folate- resistant NTDs (Fig. 3.1C).

3.3.2 Cldn3 is not required for bending and elevation of the neural folds Elevation of the neural folds is regulated by intrinsic forces arising within the neural ectoderm and extrinsic forces generated by the non-neural ectoderm. The intrinsic forces involve formation of focal bending sites at the midline called the median hinge point (MHP) that extends along the entire anterior-posterior axis of the embryo and dorsolaterally called dorsolateral hinge points (DLHPs) which are restricted to future brain levels in the chick (Lawson et al., 2001; Schoenwolf and Franks, 1984; Smith and Schoenwolf, 1987). The non-neural ectoderm also plays a role in DLHP formation by secreting signaling factors that induce their formation (McShane et al., 2015). Histological sections and analysis of the Shh expression domain, which marks the MHP, showed normal bending of the neural folds and formation of the MHP and DLHPs (Fig. 3.2). This was expected because MHP formation is intrinsic to the neural ectoderm, which does not express Cldn3.

Non-neural ectoderm cell divisions are preferentially oriented along the anterior-posterior and medial-lateral axes with anterior-posterior divisions contributing to lengthening of the embryo and medial-lateral divisions, along with convergent extension movements, generating a pushing force that causes the neural folds to converge towards the midline (Hackett et al., 1997; Sausedo et al., 1997; Smith and Schoenwolf, 1991). In contrast to previous reports, we did not observe an enrichment of medial-lateral cell divisions in GST-treated embryos (n=4 embryos/59 cells) and there was no significant difference in the proportion of rostral-caudal, medial-lateral or diagonal cell divisions in C-CPELDR-treated embryos (n=3 embryos/67 cells, p=0.6250) (Fig. 3.3A). C-CPELDR -treated embryos displayed a normal length-to-width ratio (n=11) compared to embryos treated with C-CPEYL (n=8, p=0.0523) and C-CPELSID (n=7, p=0.3191) suggesting that convergent extension was also not affected (Fig. 3.3B). A change in cell shape of non-neural ectoderm cells from cuboidal to squamous also generates a pushing force that causes the neural folds to bend inwards. The width (p=0.7) and surface area (p=1.0) of non-neural ectoderm cells was not significantly different in C-CPELDR-treated embryos (n=3 embryos/72 cells) (Fig.3.3C). Together, these data suggest that Cldn3 does not play a role in the extrinsic pushing forces generated by the non-neural ectoderm.

103

3.3.3 Cldn3 is required for neural fold fusion Together, the above data indicated that C-CPELDR-treated embryos cannot be distinguished from wild-type embryos grown in ovo or cultured with GST, C-CPEYL, or C-CPELSID prior to the final step of neural fold fusion. We examined the gene expression pattern of AP-2α, a marker of the non-neural ectoderm and neural crest cells (Fig. 3.4). RNA in situ hybridization showed that AP-2α is expressed normally in the non-neural ectoderm and migrating neural crest cells of CPELDR-treated embryos compared to GST-treated control embryos. However, AP-2α-positive neural crest cells were observed in the lumen of the neural tube of CPELDR-treated embryos confirming that the neural folds have failed to fuse. Interestingly, AP-2α null mice exhibit severe fusion defects including failure of cranial neural tube closure (Zhang et al., 1996). The fact that AP-2α expression is not affected indicates that Cldn3 acts in a separate pathway from AP-2a or that AP-2α functions upstream of Cldn3 to regulate neural tube closure.

3.3.4 C-CPELDR-treatment affects protein localization at the apical surface, but not apical-basal polarity, of non-neural ectoderm cells Claudins play a role in the maintenance of apical-basal polarity in epithelial cell layers and disrupting apical-basal polarity affects neural tube closure (Cheung et al., 2012; Eom et al., 2011; Wessely and Tran, 2011). In both GST- and C-CPELDR-treated embryos, the tight junction scaffolding protein ZO-1 was localized apical to the basolateral marker Na+K+ATPase (Fig. 3.5A) and Par3 was restricted to the apical domain (Fig. 3.5B) suggesting that apical-basal polarity of non-neural ectoderm cells was maintained. Furthermore, transmission electron microscopy analysis of tight junction ultrastructure revealed electron dense tight junctions between non-neural ectoderm cells of C-CPELDR-treated embryos that were indistinguishable from those of GST-treated embryos (Fig. 3.5C).

Apical surface views showed that Cldn4 and ZO-1 localized normally to the lateral cell membranes of non-neural ectoderm cells in C-CPELDR-treated embryos (Fig. 3.1A). This was not the case for Par3 which was restricted to the apical domain but the total levels of Par3 were increased and this was accompanied by reduced localization of Par3 to the lateral membrane (Fig. 3.5B). These data support a role for Cldn3 in maintaining protein complexes at the apical tight junction cytoplasmic plaque.

104

3.3.5 Cldn3 is required to form a meshwork of fibrils that connects the apposed neural folds Our data suggest that removing only Cldn3 affects fusion of the neural folds during the final phase of neural tube closure. In amphibians (Mak, 1978) and mammals (Geelen and Langman, 1979; Waterman, 1976), the initial contact between the apposed neural folds is mediated by cellular protrusions emanating from neural and non-neural ectoderm cells. Inhibiting formation of cellular protrusions emanating from non-neural ectoderm cells causes spinal NTDs in mouse embryos (Rolo et al., 2016). While cellular protrusions are observed at the tips of the approaching neural folds of avian embryos by transmission electron microscopy (TEM) and scanning electron microscopy (SEM), the number of protrusions is far fewer than what is observed in amphibian and mammalian embryos suggesting that the initial events regulating neural fold fusion in the chick may be different (Schoenwolf, 1979). We, therefore, analyzed the apposed neural folds throughout neurulation using SEM. In control embryos, we did not observe cellular protrusions emanating from the neural or non-neural ectoderm at any axial level at any stage of neural tube closure examined (n=5, data not shown). Instead, we observed a fibrous material connecting the closely apposed neural folds which may be functionally equivalent to the cellular protrusions observed in amphibian and mammalian embryos in mediating the initial contact between the neural folds (Fig. 3.6). A similar meshwork of thin fibrils overlying the approaching neural folds was previously reported in the chick embryo (Bancroft and Bellairs, 1975) but was not examined in further detail. We next analyzed the apposed neural folds of embryos treated with C-CPELDR. A thin network of fibrils overlying the approaching neural folds was never observed in C-CPELDR-treated embryos (n=5) (Fig. 3.7). These results indicate that Cldn3 is required for formation of the fibrous material emanating from non-neural ectoderm cells.

3.4 DISCUSSION We previously showed that C-CPE removal of Cldn4 and -8 from the neural ectoderm of neural plate stage chick embryos causes NTDs due to defective apical constriction of midline neural ectoderm cells and convergent extension movements (Baumholtz et al., 2017b). In this study we showed that claudins also regulate morphogenesis of the non-neural ectoderm during neural tube closure. Removal of only Cldn3 from the non-neural did not affect neural fold

105

elevation and formation of the median and dorsolateral hinge points but the dorsal tips of the neural folds failed to fuse resulting in NTDs. These data suggest that the combination of claudins expressed in a tissue creates a claudin ‘code’ that regulates cell and tissue behaviors. Analysis of C-CPELDR-treated embryos revealed that convergent extension movements, oriented cell division and cell spreading of non-neural ectoderm cells occurred normally, but that Cldn3 plays a role in forming a fibrous network that bridges the apposed neural folds. Folic acid supplementation was unable rescue the open NTDs in C-CPELDR-treated embryos suggesting that Cldn3-mediated morphogenetic events underlie a subset of the human NTDs not prevented by dietary folic acid fortification.

3.4.1 Cldn3 mediates the initial contact between the apposed neural folds Scanning electron microscopy revealed a mesh-like network bridging the apposed neural folds in GST-treated control embryos that was not present in Cldn3-depleted chick embryos. This meshwork was first identified in the 1970s and was proposed to act as a temporary adhesive (Bancroft and Bellairs, 1975; Lee et al., 1983; Lee and Nagele, 1985). Histological staining, its ability to bind to concanavalin A and its removal by trypsin suggest that it is a mix of proteoglycans and glycoproteins, extracellular matrix proteins. Cldn3 is expressed in tight junctions of ameloblasts in the mouse tooth germ where it plays a role in secretion of extracellular matrix proteins; Cldn3 knockout mice exhibit enamel defects caused by excess accumulation of extracellular matrix proteins in the forming enamel (Bardet et al., 2017). A proximity ligation assay in MDCK II epithelial cells showed that the proteoglycan syndecan-1 and the extracellular matrix proteins receptors integrin α2 and β1 and -3 were enriched around the N-terminus of Cldn4 (Fredriksson et al., 2015). Through their cytoplasmic C-terminal domains, claudins interacts with many proteins at the tight junction cytoplasmic including scaffolding proteins, RhoGTPases and polarity complex proteins anchoring them to the apical membrane (Baumholtz et al., 2017b; Itoh et al., 1999a; Quiros and Nusrat, 2014; Roh et al., 2002a). It is possible that Cldn3 is required to anchor extracellular matrix proteins or their receptors, the integrins, to the apical membrane of non-neural ectoderm cells where extracellular matrix proteins are secreted into the extracellular space.

Cldn3 null mice do not exhibit NTDs (Ahmad et al., 2017; Kooij et al., 2014). We hypothesize that other claudins act functionally redundantly to Cldn3 in mediating

106

morphogenesis of the non-neural ectoderm in the mouse embryo. Good candidates are Cldn6 and -7, which are absent in the chick genome. Cldn3, -4, -6 and -7 are downstream targets of the Grhl2 transcription factor (Pyrgaki et al., 2011; Senga et al., 2012; Werth et al., 2010). Grhl2 mutant mice exhibit NTDs defects caused by defective fusion of the neural folds (Pyrgaki et al., 2011), a phenotype similar to what we observed in Cldn3-depleted chick embryos. A recent study suggests that the NTDs observed in the Grhl2 mutant mouse are due to loss of epithelial integrity of non-neural ectoderm cells (Ray and Niswander, 2016). However, there is no evidence to show that the loss of epithelial identity is responsible for the NTDs observed in these mice. Further studies are needed to determine if other processes are affected in Grhl2 mutant mice that are required for neural fold fusion. Depletion of Cldn3 did not affect the structure of tight junctions or apical-basal polarity of non-neural ectoderm cells in chick embryos suggesting that epithelial integrity is maintained. We hypothesize that the loss of epithelial integrity observed in the Grhl2 mutant mice is due to loss of claudins in conjunction with other junctional proteins such as E-cadherin. In support of this hypothesis, loss of Cldn4 and -8 from neural ectoderm cells in C-CPE-treated embryos did not affect epithelial integrity of the neural ectoderm (Baumholtz et al., 2017b).

3.4.2 Cldn3 regulates protein localization at the apical surface of non-neural ectoderm cells Establishment of the apical membrane domain in epithelial cells depends on the coordination of intracellular trafficking pathways, the Crumbs-Pals-Patj and Par3-Par6-αPKC polarity complexes and RhoGTPase signaling. Despite a clear demarcation of the apical and basolateral membrane domains in C-CPELDR-treated embryos, Par3 showed reduced membrane and increased cytoplasmic localization. This differs from what is observed in C-CPE-treated embryos where Par3 is abnormally localized in large aggregates near the apical membrane of Cldn4 and -8- depleted neural ectoderm cells (Baumholtz et al., 2017b). These data indicate that claudins are required to anchor proteins to the apical membrane of epithelial cells where they function to localize proteins. Furthermore, these data suggest that the localization of proteins at the apical membrane is differentially regulated by individual claudins.

In conclusion, we have defined a new role for Cldn3 in regulating morphogenesis of the non- neural ectoderm during neural tube closure. Our data support a model where changing the

107

combination of claudins at tight junctions influences protein interactions at the tight junction cytoplasmic plaque. Previously we showed that removing Cldn4 and -8 from the neural ectoderm disrupts localization of the planar cell polarity protein Vangl2, the RhoGTPases RhoA and Cdc42, and the polarity complex protein Par3 at the apical membrane of neural ectoderm cells (Baumholtz et al., 2017b). Our data supports a model where the combination of claudins expressed in an epithelial tissue forms structural and functional compartments that organize the apical domain of epithelial cells within that tissue. The claudin cytoplasmic C-terminal tail is the most variable region between family members and, thus, likely mediates the unique interactions of individual family members. In support of this hypothesis, the C-terminus of claudins binds to PDZ domain scaffolding proteins at the tight junction cytoplasmic plaque and post-translational modifications of the C-terminus of claudins may influence their relative affinities for interacting partners (Findley and Koval, 2009). Future studies will be needed to understand how individual claudins interact with proteins at the apical cell surface and how these interactions are influenced by post-translation modifications.

3.5 MATERIALS AND METHODS

3.5.1 Production of GST and GST-C-CPE variants Expression of GST and the variants C-CPELDR (L223A/ D225A/R227A), C-CPELSID (L254A/ S256A/I258A/D284A) and C-CPEYL (Y306A/L315A) cloned into pET14b or pGEX6P1 was induced by IPTG. GST fusion proteins were purified from E. coli BL21 as described previously and dialyzed against PBS (Protze et al., 2015; Veshnyakova et al., 2012). Protein concentration was determined using the BCA Protein Kit (Thermo-Scientific, Rockford, USA) or Bio-Rad Protein Assay (BioRad, Mississauga, Canada).

3.5.2 Embryo culture Fertilized eggs (Couvoir Simetin, Mirabel, Canada and Ferme GMS, Saint-Liboire, Canada) were incubated at 38.5°C. Embryos were collected, staged according to Hamilton and Hamburger (HH) criteria (Hamburger and Hamilton, 1951), and cultured using the modified Cornish pasty method (Nagai et al., 2011). Purified GST or C-CPE variants (200-400 µg/ml) were added directly to the Cornish pasty culture medium to achieve the final concentrations indicated.

108

3.5.3 Immunofluorescence staining Whole embryo immunostaining was performed on embryos fixed in 10% trichloroacetic acid at 4°C for 1 hour. Samples were incubated overnight at 4°C with primary antibodies: Cldn3 (Abcam, Cambridge, UK, 1:50), Cldn4 (Invitrogen, Carlsbad, USA, 1:50), ZO-1 (Invitrogen, Carlsbad, USA, 1:50), Par3 (Millipore, Etobicoke, Canada, 1:250), Na+K+ATPase (DSHB, Iowa, USA. 1:3). Alexa Fluor-conjugated secondary antibodies (1:500) were added for 1 h at room temperature. Embryos were flat-mounted and coverslipped with SlowFade Gold with DAPI (Molecular Probes, Eugene, USA) and imaged using a Zeiss LSM780 laser scanning confocal microscope. Colocalization and immunoquantification was performed using ZEN 2012 SP1 software (Carl Zeiss Microscopy, Germany).

3.5.4 Whole mount in situ hybridization Antisense digoxygenin-labeled riboprobes for AP-2α and Shh were used as previously described (Baumholtz et al., 2017b). Embryos were photographed using a Leica M125 dissecting microscope with Infinity Capture software v5.0.2 (Lumenera Corp.). Paraffin- embedded embryos were sectioned and photographed using a Zeiss Axiophot compound microscope and AxioCamMRc camera with Axiovision v4.7.1.0 software.

3.5.5 Scanning electron microscopy Specimens were prepared at the Facility for Electron Microscopy Research, McGill University. Embryos were fixed overnight in 2% glutaraldehyde, 2% paraformaldehyde in 0.1M phosphate buffer, pH7.4 at 4°C. After rinsing with distilled water, embryos were dehydrated in a series of graded ethanol-water washed to 100% ethanol. The samples were then critical point

dried using CO2 and mounted on aluminum metal holders using sticky carbon tape. The mounted samples were then coated with a thin layer of Pd using an ion beam coater and imaged using an FEI Helio Nanolab 660 DualBeam scanning electron microscope.

3.5.6 Transmission electron microscopy Specimens were prepared at the Facility for Electron Microscopy Research, McGill University as previously described (Baumholtz et al., 2017b). Images were acquired on an FEI Techai 12 120kV transmission electron microscope equipped with an AMT xR80C 8 megapixel CCD camera.

109

3.5.7 Morphometric analysis of length-to-width ratios and oriented cell division Dorsal images of embryos were photographed using a Leica M125 dissecting microscope with Infinity Capture software v5.0.2 (Lumenera Corp.) and imported into ImageJ. Width of the neural tube at the level of the first somite and anterior-posterior length were measured. Oriented cell division was quantified as previously described (Baumholtz et al., 2017b). Mitotic spindles of anaphase and metaphase cells were pooled to increase statistical power.

3.6 ACKNOWLEDGEMENTS We thank M. Bronner for AP-2α, and D. Lui, L. Monaghan and J. Mui (McGill University FEMR) for technical assistance.

AIB is the recipient of a doctoral fellowship from the Fonds recherché du Québec – Santé (FRQS, 29939). This work was supported by the Deutsche Forschungsgemeinschaft (grant number 1273/3-2 to DFG, KR and grant number PI 837/2-1 to GK, JP) and a Natural Sciences and Engineering Research Council of Canada Discovery grant (234319) (AKR). AKR is a member of the RI-MUHC, which is supported in part by the FRQS.

110

Figure 3.1 Cldn3-specific C-CPE variant, C-CPELDR, causes NTDs. (A) Apical surface view of Cldn-3 or -4 (green) and ZO-1 (red) in embryos treated with C- CPEYL, C-CPELDR, or C-CPELSID for 5h. Scale bar, 20 µm. (B) Dorsal view of chick embryos treated with 200 µg/ml GST, C-CPELDR, or C-CPELDR + C-CPELSID for 20 h. (C) Dorsal view of embryos treated with 200 µg/ml GST or C-CPELDR in the presence of 100 µM folic acid. Dashed lines indicate open neural tubes. Scale bar, 0.2 mm.

111

Figure 3.2. Formation of the hinge points is not affected in C-CPELDR-treated embryos. (A) Transverse sections showing Shh expression in the cranial and spinal neural tube of GST and C-CPELDR-treated embryos at 20 h. Dashed line marks the Shh expression domain in the floor plate. (B) H&E staining of transverse sections of the cranial and spinal neural tube of HH11 chick embryos. Note that the neural folds of C-CPELDR-treated embryos bend inwards indicating normal dorsolateral hinge point formation (asterisks). Scale bar, 50 µm.

112

Figure 3.3. Oriented cell division, convergent extension and changes in cell shape of non- neural ectoderm cells are not affected in C-CPELDR-treated embryos. (A) Graphs showing orientation of cell divisions relative to the midline in GST- (n=4 embryos/59 cells) and C-CPELDR-treated embryos (n=3 embryos/67 cells) at the neural tube stage. Blue bars represent rostrocaudal (RC) divisions, green are mediolateral (ML) divisions, and red are diagonal (D) divisions (mean±s.e.m., Wilcoxon’s signed-ranks test). (B) Graph shows length-to-width ratios for C-CPEYL- (n=8), C-CPELDR- (n=11), or C-CPELSID-treated embryos (n=7) at the neural tube stage (mean±s.e.m., Mann-Whitney U-test). (C) Scanning electron microscope images showing surface views of the apposed neural folds of embryos treated with GST or C-CPELDR. Bottom images show higher magnification views of the non- neural ectoderm in the boxed region. Double arrow indicates the cell width. Scale bar, 50 µm for top images and 20 µm for bottom images. Graphs show the average surface area and the average cell width of non-neural ectoderm cells of GST- (n=3 embryos/84 cells) and C-CPELDR- treated embryos (n=3 embryos/72 cells).

113

Figure 3.4. Aberrant neural crest migration in C-CPELDR-treated embryos. Transverse sections showing AP-2α expression in the cranial neural tube of embryos treated with GST or C-CPELDR for 20 h. Neural crest cells migrate along the edges of the non-neural ectoderm in the GST-treated embryo whereas they migrate into the lumen of the open neural tube in C-CPELDR-treated embryos (arrows). Scale bar, 50 µm.

114

Figure 3.5. Non-neural ectoderm cells of C-CPELDR-treated embryos exhibit loss of epithelial integrity and an altered apical surface. (A) Apical surface and XZ orthogonal views of ZO-1 (red) and Na+K+ATPase (green) in the non-neural ectoderm of 20 h GST- and C-CPELDR-treated embryos. (B) Apical surface and XZ orthogonal views of Par3 (green) and ZO-1 (red) in the lateral non-neural ectoderm of 20 h GST and CPELDR-treated embryos. Three embryos per treatment were analyzed. Graphs show the ratio of the immunofluorescent intensity of Par3 to ZO-1 (left) and Pearson’s correlation coefficient for Par3 and ZO-1 (right). Histograms represent an average of three different fields from three embryos and values were calculated from maximum intensity projections (mean ± s.e.m., *p=0.0186, **p=0.0049, student t-test). Scale bar, 20 µm. (C) Top, transmission electron micrographs of transverse section of the dorsal neural tube of GST- and CPELDR-treated embryos at the neural tube stage. Bottom, high magnification images of the non-neural ectoderm. Arrowheads indicate tight junctions. Scale bar, 2 µm.

115

Figure 3.6. A meshwork of thin fibrils connects the non-neural ectoderm of apposing neural folds during spinal neurulation. (A) Brightfield) and (A’) scanning electron microscopy images of an HH10 chick embryo. (i, i’) A thin network of fibrils connects the edges of the apposed neural folds. Note the absence of this meshwork in more posterior regions of the embryo where the neural folds are elevated but spaced further apart (ii, ii’). Three different embryos were analyzed. Scale bars: 500 µm (A’), 25 µm (i, ii), and 5 µm.

116

Figure 3.7. The apposed neural folds of C-CPELDR-treated embryos are not connected by fibrils. (A-A’’) Scanning electron microscope images of HH10 control and (B-B’’) C-CPELDR-treated embryos. SEMs of the apposed spinal neural folds of control embryos show a network of fibrils (A’ and A’’), whereas this network is absent in C-CPELDR-treated embryos (B’ and B’’). Scale bars: 500 µm (A and B), 25 µm (A’ and B’), and 5 µm (A’’and B’’).

117

3.7 CONNECTING TEXT BETWEEN CHAPTER III AND IV

In Chapter III, I showed that Cldn3 localization to tight junctions in the non-neural ectoderm regulates neural tube closure in chick embryos. Together with the data presented in Chapter II showing that Cldn4 and -8 are required in the neural ectoderm for neural tube closure and C-CPE-mediated depletion of claudins causes NTDs in mouse embryos, our results suggest that claudins play an evolutionarily conserved role in vertebrate neural tube closure. In Chapter IV, I identified rare and novel missense mutations in CLDN1-25 in a cohort of patients with spinal NTDs and performed functional studies to test the pathogenicity of these variants. The data from Chapter II strongly support that Cldn8 plays a critical role in neural tube closure. I did not identify rare or novel missense variants in Cldn8 in the cohort of patients with spinal NTDs. This is expected as data from Chapter II suggest that deleterious mutations in CLDN8 would cause an embryonic lethal form of NTDs in which the neural tube remains open along the entire anterior-posterior axis. As part of this study, I performed mutagenesis experiments in the C- terminal tail of Cldn8. It is possible that these data may form part of a fourth manuscript.

118

4 CHAPTER IV: Mutations in the claudin family of tight junction proteins contribute to the etiology of human neural tube defects

Amanda I. Baumholtz, Elisa Merello, Patrizia De Marco, Valeria Capra, and Aimee K. Ryan

Manuscript in preparation

119

4.1 ABSTRACT Neural tube defects (NTDs) are severe malformations of the central nervous system affecting 1-2 individuals per 2000 births. Their etiology is complex involving both genetic and environmental factors. Our recent discovery that removal of Cldn3, -4 and -8 from tight junctions results in cranial and spinal NTDs in both chick and mouse embryos suggests that claudins play a conserved role in neural tube closure in vertebrates. The aim of the present study was to determine if genetic variants in CLDN loci were present in patients with NTDs. Using a Fluidigm next generation sequencing approach to sequence CLDN coding sequences in 125 patients with spinal NTDs, we identified ten rare and four novel missense mutations in nine CLDN genes. Functional analysis revealed that overexpression of the CLDN19 I22T and E209G variants caused a significant increase in NTDs and convergent defects in chick embryos. Our findings implicate rare nonsynonymous variants in CLDN genes as risk factors for spinal NTDs and identify a new family of proteins involved in the pathogenesis of these malformations.

120

4.2 INTRODUCTION Neural tube defects (NTDs) are a group of congenital central nervous system malformations that affect the brain and spinal cord. They occur when the neural tube fails to close at any level of the embryonic anterior-posterior axis during the 3rd and 4th week of gestation. NTDs are the second most common birth defect with a varying incidence affecting 1-2 per 2000 pregnancies in North America and as high as 20 per 1000 births in certain regions of China (Zaganjor et al., 2016). NTDs are classified as ‘open’ when the affected nervous system tissues are exposed to the environment or ‘closed’ when the defect is covered by skin. The two most common forms of NTDs are anencephaly and myelomeningocele (spina bifida), which result from failure of the neural tube to close in the cranial or spinal region, respectively. A third rare form of open NTDs, craniorachischisis, occurs when the neural tube fails to close along the entire body axis. Of these, only myelomeningocele is compatible with survival beyond birth.

NTDs have a complex etiology involving both genetic and environmental factors. One of the most significant findings in the field has been the protective effect of periconceptional folic acid supplementation, which reduces the incidence of NTDs by as much as 50-70% (Atta et al., 2016; De Wals et al., 2007; Osterhues et al., 2013). Despite this, 300,000 newborns affected by these defects are born annually (Zaganjor et al., 2016) and, thus, there is an urgent need to better understand the underlying pathogenic mechanisms of NTDs, particularly those not prevented by folic acid fortification.

Primary neurulation, the developmental process that converts the flat neural plate into a closed neural tube, occurs in four distinct phases. First, the neural plate is induced to differentiate. The neural plate then undergoes mediolateral convergence and anterior-posterior extension, a process called convergent extension. Convergent extension is regulated by the non- canonical Wnt/planar cell polarity (PCP) pathway. Accordingly, animal models with mutations in PCP components develop NTDs caused by aberrant convergent extension movements (Curtin et al., 2003; Goto and Keller, 2002; Greene et al., 1998; Paudyal et al., 2010; Wallingford and Harland, 2002; Wang et al., 2006a; Wang et al., 2006b; Ybot-Gonzalez et al., 2007b). Rare and novel missense mutations in PCP genes are risk factors for NTDs in humans (Allache et al., 2012; Bosoi et al., 2011; De Marco et al., 2013; De Marco et al., 2012; Kibar et al., 2009; Kibar et al., 2011; Lei et al., 2010; Robinson et al., 2012). Subsequent formation of a midline and

121

dorsolateral hinge points causes bending of the neural plate and elevation of the neural folds. Formation of the hinge points requires RhoA/ROCK-mediated actin-myosin contraction of the apical surface of neural plate cells (Escuin et al., 2015; Kinoshita et al., 2008; Rolo et al., 2009). Finally, the dorsal tips of the neural folds meet at the midline where they fuse to generate a closed neural tube. A defect in any of these phases of neural tube closure will results in a NTD.

We recently showed that claudins act upstream of the PCP and RhoA/ROCK signaling pathways to regulate cell shape changes and tissue movements during neural tube morphogenesis (Baumholtz et al., 2017b). Claudins comprise a family of integral tight junction proteins with ~24 unique members in vertebrates. They have four transmembrane domains, two extracellular loops, and cytoplasmic N- and C-termini. The first extracellular loop contains charged amino acids that play a critical role in determining the ion and size selectivity of the tight junction barrier (Hamazaki et al., 2002; Hughes, 2004; Itoh et al., 1999a; Piontek et al., 2017b). The second extracellular loop participates in trans-interactions with claudin molecules in apposing cells and in cis-interactions with other claudins in the same cell (Piontek et al., 2008; Suzuki et al., 2014). The transmembrane domains also participate in cis oligomerization (Rossa et al., 2014). Together, these interactions constitute the backbone of tight junction strands and prevent the mixing of apical and basolateral proteins (Krause et al., 2015). The claudin cytoplasmic C- terminal tail shows the most size and sequence heterogeneity between family members. It has a conserved PDZ-binding domain that interacts with adaptor proteins at the tight junction cytoplasmic plaque linking the tight junction to the actin cytoskeleton and intracellular signaling events (Fredriksson et al., 2015; Van Itallie et al., 2013; Voiculescu et al., 2008) and it contains numerous phosphorylation sites that can influence these interactions at the tight junction cytoplasmic plaque (Ikari et al., 2006; Tanaka et al., 2005a).

We discovered that simultaneous removal of Cldn3, -4, and -8 from tight junctions causes open NTDs in chick and mouse embryos (Baumholtz et al., 2017b). Using the chick animal model, we showed that claudins are required for PCP-dependent convergent extension and actin- myosin-dependent apical constriction during neural tube closure. Given the importance of claudins in neural tube closure in chick and mouse, we hypothesized that mutations in CLDN genes are risk factors for human NTDs. Therefore, we screened for rare and novel nonsynonymous CLDN variants in a cohort of 125 open spinal NTD cases. The functional

122

consequences of the CLDN variants were investigated in two ways. First, transfection of HEK293 cells was used to determine if the variant protein localizes to the tight junction and then an overexpression assay was performed in chick embryos to determine if the variants caused defects in neural tube closure. We found that CLDN19 I22T and E209G did not affect Cldn19 localization to tight junctions, but caused NTDs when overexpressed in chick embryos. Our data suggest that rare missense mutations in CLDN genes are risk factors for human NTDs.

4.3 MATERIALS AND METHODS

4.3.1 Human subjects The patient cohort consisted of 125 patients recruited at the Spina Bifida Center of the Gaslini Hospital in Genova, Italy. All patients included in this study are affected with myelomeningocele (open spina bifida), the most severe cases of neural tube defects recruited at this centre.

4.3.2 Next-generation DNA sequencing Genomic DNA was isolated from blood samples using the QIAamp DNA blood kit according to the manufacturer’s protocol (Qiagen, Toronto, Canada). DNA was then processed by microfluidic PCR using a custom Fluidigm Access Array, followed by next-generation sequencing on an Illlumina MiSeq by the Genome Quebec Innovation Center (Montreal, Quebec, Canada). Briefly, 92 custom pairs of primers were designed to amplify ~300 bp fragments of the coding exons of the 24 human claudin genes (Suppl. Table 4.1). Primer sets were validated using a standard PCR protocol to ensure they amplified a single fragment of the expected size and selected amplicons were validated by Sanger sequencing. The amplicons from the array were pooled and the resulting pools were multiplexed and sequenced using the Illumina MiSeq. Reads were processed and aligned to the human genome (UCSC hg19, NCBI build 37) by Genome Quebec. Aligned and processed BAM files were used to manually identify variants using the Integrative Genomics Viewer version 2.3. Variants with alternative allele frequencies of 30% or greater were selected for further analysis.

123

4.3.3 Bioinformatics The EVS (Exome Variant Server; http://evs.gs.washington.edu/EVS/) and ExAC Browser (Exome Aggregation Consortium; http://exac.broadinstitute.org/) public databases were queried for the presence and incidence of the variants identified in CLDN1-25. Rare variants were defined as having a minor allele frequency (MAF) ≤ 1% in these databases and novel variants were defined as those not present in these databases. The potential pathogenic effect of the identified mutations on protein function was predicted using three software programs: PolyPhen (Polymorphism Phenotyping; http://genetics.bwh.harvard.edu/pph2/), SIFT (Sorting Intolerant from Tolerant; http://sift.jcvi.org/) and MutationTaster (http://www.mutationtaster.org/). Multiple alignments of the CLDN proteins were done using the CLUSTAL W program. Accession numbers of claudin amino acid sequences used for alignment are listed in Suppl. Table 4.2. Predicted phosphorylation sites in CLDN proteins were identified using NetPhos 3.1 (Blom et al., 1999).

4.3.4 Sanger sequencing Rare and novel missense variants were validated by Sanger sequencing. Primers flanking the coding exons of CLDN3 (GenBank: AF007189.1), CLDN4 (NCBI: NG_012868.1), CLDN6 (GenBank: AJ249735.1), CLDN8 (GenBank: AJ250711.1), CLDN9 (Ensembl: ENSG00000213937), CLDN14 (NCBI: NM_144492.2), CLDN16 (Ensembl: ENSG00000113946), CLDN18 (Ensembl: ENSG00000066405), CLDN19 (ENSG00000164007), CLDN20 (Ensembl: 00000171217), CLDN23 (Ensembl: ENSG00000253958), CLDN24 (ENSG00000185758) were used to amplify the coding region by polymerase chain reaction (Suppl. Table 4.3). Amplicons were subjected to Sanger sequencing at the Genome Quebec Innovation Centre (Montreal, Quebec, Canada).

4.3.5 Generation of claudin wild-type and mutant constructs The full-length human wild-type and variant sequences for claudins encoded by a single exon (CLDN3, -4, -6, and -24) were amplified from genomic DNA of NTD patients by PCR and cloned into the pSCA vector using the Stratagene PCR cloning kit (Agilent Technologies, Santa Clara, CA). The full-length coding sequence of CLDN19 was obtained by RT-PCR from human fetal kidney (provided by Dr. Cindy Goodyer, McGill University, Montreal, Canada). Total RNA was isolated using the Strategene Absolutely RNA Miniprep Kit according to

124

manufacturer’s directions (Agilent Technologies, Santa Clara, CA). RT-PCR was performed using the OneStep RT-PCR kit (QIAGEN, Toronto, Canada) and amplicons were cloned into the pSCA vector. The primers used for RT-PCR are shown in Suppl. Table 4.4. pMXs-Puro- CLDN18 was a gift from Axel Hilllmer (Addgene plasmid #69466) (Yao et al., 2015) and used as a template to clone full-length CLDN18 into pSCA. Human variants for CLDN18 and -19 were introduced in the pSCA-claudin constructs by site-directed mutagenesis using the Stratagene QuikChange II Site-Directed Mutagenesis kit according to manufacturer’s directions (Agilent Technologies, Santa Clara, CA). Sequence identity of cDNA clones was confirmed by sequencing at the Genome Quebec Innovation Centre (Montreal, Quebec, Canada). Wild-type and variant constructs were then cloned into the pCanHA3 and pMES expression plasmids. Sequences of mutagenesis primers are shown in Suppl. Table 4.5

4.3.6 Immunolocalization of ectopically expressed claudins HEK293 cells grown on coverslips in a 24-well plate were transfected at 70-90% confluence with pMES or pCanHA3 expression vectors encoding wild-type or variant claudins using Lipofectamine LTX (Invitrogen) in a 1:1 DNA-reagent ratio. Transfected cells were incubated for 24-48 hours at 37°C. After rinsing with PBS, cells were fixed with 10% trichloroacetic acid or 4% PFA for 10 min at 4°C, blocked with 10% normal goat serum in 0.3% Triton X-100 in PBS and then incubated with primary antibody for 1 hour at room temperature. Primary antibodies used were Cldn3, -6, -9 (Abcam, Cambridge, UK, 1:100), Cldn4 (Invitrogen, Carlsbad, USA, 1:100), ZO-1 (Invitrogen, 1:100, Carlsbad, USA), and HA (Invitrogen, Carlsbad, USA, 1:100). Antibodies were detected with Alexa fluor 488 conjugated goat anti-rabbit IgG and Alexa fluor 594 conjugated goat anti-mouse IgG (Molecular Probes). Slides were coverslipped with SlowFade Gold Antifade kit (Molecular Probes), which contained DAPI to enable visualization of the nuclei and imaged using a laser scanning confocal microscope (LSM780, Leica, Germany).

4.3.7 Overexpression assay in the chick embryo pMES-claudin constructs (5µg/µl) were injected into the space between the vitelline membrane and the epiblast of HH4 embryos cultured on agar-albumen plates using a Narishige IM 300 microinjector and electroporated using a Protech CUY21SC square electroporator (five

125

3V/50ms pulses at 500ms intervals). Manipulated embryos were cultured until HH11-12 and scored for defects in neural tube closure and convergent extension.

4.4 RESULTS

4.4.1 Variants identified in CLDN genes among myelomeningocele patients Sequencing analysis of the coding region of the 24 human CLDN genes of 125 patients with neural tube defects (NTDs) identified a total of 39 synonymous single nucleotide polymorphisms (SNPs) and 31 nonsynonymous variants. Of the missense variants, 17 were common (MAF ≥ 1%, Table 4.1), 10 were rare (MAF < 1%, Table 4.2), and 4 were novel (absent in the EVS and ExAC databases, Table 4.2). The E161K rare nonsynonymous variant in CLDN24 could not be confirmed by Sanger sequencing as there was no genomic DNA remaining for this patient.

To identify mutations that cause or predispose to NTDs, we focused our analysis on the novel and rare nonsynonymous variants as these were more likely to have a negative impact on protein function. All the variants were found in the heterozygous state and most were private with the exception of CLDN3 p.A128T identified in 3 patients and CLDN6 p.R209Q found in two patients. No rare or novel missense mutations were identified in CLDN1, 2, 5, 7, 8, 10-15, 17, 20-22, 25 and no NTD patient was a carrier of a rare or novel missense mutation in more than one CLDN gene. Polyphen, SIFT, and MutationTaster were used to evaluate the potential pathogenic effect of each variant using the following criteria: 1) the degree of evolutionary conservation of the affected residue in the Cldn protein and 2) the nature of the amino acid replacement and its possible impact on protein function and localization. These are described below.

4.4.2 Rare and novel variants in CLDN3 Three nonsynonymous variants, p.A128T, p.P134L, and p.G207A were detected in CLDN3 (Fig. 4.1). The p.A128T and p.P134L variants affect residues predicted to localize to the third transmembrane domain, while the p.G207A variant is located in the cytoplasmic C-terminal tail (Fig. 4.2A). p.A128T, found in three NTD patients and absent from ExAC and EVS databases, affects a highly conserved residue among mammals (Fig. 4.2B) and is predicted to be deleterious by SIFT and MutationTaster. An alanine to threonine substitution results in reduced

126 hydrophobicity within the transmembrane domain and may affect the ability of the protein to insert into the membrane, but does not create a putative phosphorylation site as predicted by NetPhos3.1. p.P134L was found in one NTD patient (Fig. 4.1) and has a MAF of 0.45 and 0.48 in the ExAC and EVS databases, respectively. A proline to leucine substitution increases the hydrophobicity of this region and should not negatively impact the transmembrane domain. However, the p.P134L variant affects a residue that is conserved across evolution (Fig. 4.2B) and was predicted to be deleterious by in silico using the PolyPhen, SIFT, and MutationTaster programs. The p.G207A mutation was found in a single NTD patient (Fig. 4.1) and is a novel mutation as it is not present in the ExAC and EVS databases. The glycine residue is conserved across mammals but is absent in chick and Xenopus, which have shorter C-terminal domains (Fig. 4.2A). A glycine to alanine substitution introduces a hydrophobic side chain that could affect the structure of the cytoplasmic C-terminal tail, yet all three softwares predicted that this change would have no effect on protein function.

4.4.3 Rare variant in CLDN4 The CLDN4 rare missense mutation, p.V84, was reported in ExAC and EVS with a MAF of 0.002478 and 0.00002447, respectively (Fig. 4.1 and 4.2). The p.V84I variant affects a poorly conserved residue that is predicted to localize to the second transmembrane domain. The fact that this variant is poorly conserved across species and that the valine to leucine substitution is a conservative change argues against a pathogenic role for this variant. This variant was predicted to be benign by PolyPhen, SIFT and MutationTaster.

4.4.4 Rare variant in CLDN6 One rare missense variant, R209Q, was detected in two NTD patients in CLDN6 (Fig. 4.1 and 4.2). This variant was found in ExAC and EVS with a MAF of 0.15 in both databases. The R209 residue is predicted to localize to the C-terminal cytoplasmic tail of Cldn6. An arginine to glutamine substitution is not conservative as it results in the loss of a positively charged residue at this position. The fact that arginine at position 209 is replaced by glutamine in Xenopus Cldn6 argues against a pathogenic role for this variant, which is predicted to be benign by PolyPhen, SIFT, and MutationTaster programs.

127

4.4.5 Rare variant in CLDN9 The CLDN9 rare missense variant, p.S3L, was detected in one NTD patient (Fig. 4.1 and 4.2). This variant was detected in ExAC with a MAF of 0.1 and EVS with a MAF of 0.0077. It affects a highly conserved serine residue in the cytoplasmic N-terminus and is predicted to be damaging by PolyPhen, SIFT, and MutationTaster. Using NetPhos3.1, we found that S3 is predicted to be phosphorylated and, thus, p.S3L would results in loss of a putative phosphorylation site.

4.4.6 Rare variant in CLDN16 The rare missense variant, p.N223S, was detected in a single NTD patient in CLDN16 (Fig. 4.1). This variant was found in ExAC with a MAF of 0.03459 and EVS with a MAF of 0.0231. N223 is located in the second extracellular loop, which determines the ion specificity of the tight junction (Fig. 4.2 A). The asparagine to serine substitution is a conservative substitution with respect to charge, but NetPhos3.1 predicts that introducing a serine at amino acid 223 creates a putation phosphorylation site. Furthermore, the N223S substitution affects a highly conserved residue among all species analyzed (Fig 4.2 B). This mutation is predicted to be possibly damaging by Polyphen and disease causing by MutationTaster, but tolerated by SIFT.

4.4.7 Rare variant in CLDN18 The CLDN18 rare missense variant, p.V88I, was detected in one NTD patient (Fig. 4.1). This variant was detected in ExAC with a MAF of 0.23 and EVS with a MAF of 0.17. It is located in the second transmembrane domain and affects a highly conserved residue among all species analyzed (Fig. 4.2A, B). Both valine and isoleucine are hydrophobic amino acids; however, a substitution with isoleucine introduces a bulky side chain that may have structural consequences. The p.V88I variant was predicted to be damaging by Polyphen and MutationTaster and benign by SIFT.

4.4.8 Rare variants in CLDN19 Two rare mutations were identified in the CLDN19 gene (Fig. 4.1). p.E209G was identified in a single patient and was present in the ExAC database with a MAF of 0.006625, but absent in the EVS database. E209 is located in the C-terminal cytoplasmic tail of the CLDN19 protein immediately adjacent to the PDZ-binding domain (Fig. 4.2A). This residue is conserved among

128

all species analyzed (Fig. 4.2B). The glutamic acid to glycine substitution results in loss of a negatively charged amino acid at this position and is considered a non-conservative change. This rare variant is predicted to be probably damaging by MutationTaster. The p.I22T variant was found in one patient and has previously been reported in the ExAC database with a MAF of 0.11 and in the EVS database with a MAF of 0.0616. It is located in the first transmembrane domain (Fig. 4.2A). Isoleucine at position 22 is highly conserved among species analyzed except in chicken where it is replaced by leucine, which has similar amino acid characteristics (Fig. 4.2B). Replacement of isoleucine with threonine results in loss of hydrophobicity at this position and, thus, may affect the ability of the transmembrane domain to insert into the plasma membrane and it introduces a putative phosphorylation site as predicted by NetPhos3.1. This variant is predicted to be possibly damaging by PolyPhen and damaging by SIFT and MutationTaster.

4.4.9 Rare variant in CLDN23 The variant p.A90T was detected in a single patient (Fig. 4.1) and was present in the ExAC and EVS databases with a MAF of 0.52 and 0.46, respectively. This variant is located in the second transmembrane domain of the protein and affects a moderately conserved amino acid (Fig. 4.2A, B). An alanine to threonine substitution is not conservative resulting in reduced hydrophobicity; however, the fact that alanine is replaced by threonine at this position in mouse argues against a pathogenic role for this mutation. This variant does not create a potential phosphorylation site as determined by NetPhos3.1 and is predicted to be benign by PolyPhen, SIFT, and MutationTaster.

4.4.10 Rare and novel variants in CLDN24 We identified two novel missense variants in CLDN24 in two patients, p.G94R and p.L177M (Fig. 4.1), and the rare nonsynonymous p.E161K variant in a single NTD patient, which had a MAF of 0.002435 in the ExAC database. No CLDN24 variants have been reported in the EVS database. G94 is located in the second transmembrane domain and is highly conserved across species (Fig. 4.2A, B). The substitution of a glycine to arginine introduces a positively-charged bulky side chain that may have structural consequences. This variant is predicted to be probably damaging by PolyPhen and damaging by SIFT and MutationTaster. The variant p.L177M affects a moderately conserved amino acid being replaced by hydrophobic amino acids valine in

129

mouse and isoleucine in Xenopus (Fig. 4.2B). This variant is located in the fourth transmembrane domain (Fig. 4.2A). The substitution of a leucine for methionine maintains the hydrophobicity at this position and, thus, is unlikely to affect the ability of the CLDN24 protein to insert into the plasma membrane. This variant is predicted to be benign by all three prediction softwares. Finally, the rare variant p.E161K is located in the second extracellular loop, which participates in cis- and trans-interactions between claudin family members. The fact that a glutamic acid to lysine switch results in a change from a positive to a negatively charged residue, and that this amino acid is conserved across all species analyzed, suggests that it is pathogenic. In fact, it is predicted to be deleterious by Polphen, SIFT, and MutationTaster.

4.4.11 Patients with a mutation in a CLDN gene and a PCP gene 14 of the 19 patients with a novel or rare mutation in a CLDN gene were previously analyzed for mutations in planar cell polarity (PCP) genes (Allache et al., 2012; Bosoi et al., 2011; De Marco et al., 2013; De Marco et al., 2012; Kibar et al., 2009; Kibar et al., 2011; Robinson et al., 2012). Four of the patients with a mutation in a CLDN gene also had a mutation in a PCP gene (Table 4.3).

4.4.12 The substitution N223S in CLDN16 impedes formation of tight junction strands Claudins form tight junction strands when transfected in tight junction-free HEK293 (Milatz et al., 2015; Piontek et al., 2017a). To determine if the CLDN mutations identified in the NTD patients affected the ability of the claudin protein to localize to tight junctions, we transiently transfected HEK293 cells with constructs containing the wild-type or variant claudins. The enrichment of Cldn constructs at contacts between adjacent claudin-expressing cells and their colocalization with ZO-1, a classic tight junction protein, indicate the ability of the claudin constructs to form tight junction strands (Fig 4.3). CLDN3 wild-type, A128T, P143L and G207A, CLDN4 wild-type and V84I, CLDN6 wild-type and R209Q, CLDN16 wild-type, and Cldn19 wild-type, I22T and E209G were enriched at contacts between neighboring Cldn- expressing cells co-localizing with the tight junction marker ZO-1. Hence, the wild-type and variant Cldns indicated above participate in homophilic cis- and trans-interactions to form tight junction strands. In contrast, CLDN16 N223S was more homogeneously distributed at the plasma membrane and did not co-localize with ZO-1. This result suggests that the N223S

130

substitution does not affect the ability of Cldn16 to reach the plasma membrane and participate in cis-interactions, but impedes the trans-interaction of Cldn16 between adjacent cells.

4.4.13 Overexpressing CLDN19 variants causes open NTDs in chick embryos The chick embryo is an excellent model to study neural tube closure because the morphogenetic movements involved in neurulation have been well-characterized and are similar to those in human embryos. To further investigate the potential pathogenic role of the rare and novel CLDN variants identified in patients with NTDs, we used an overexpression assay in the chick embryo. We previously showed that simultaneous removal of Cldn3, -4 and -8 from tight junctions using the C-terminal domain of Clostridium perfringens enterotoxin (C-CPE) causes open NTDs in chick embryos due to defective convergent extension movements and apical contriction (Baumholtz et al., 2017b). We electroporated HH4 neural plate stage chick embryos with expression vectors encoding wild-type or variant human claudins. Embryos were collected and examined for defects in neural tube closure (Fig. 4.4 A). Overexpression of CLDN6 R209Q (n=21), CLDN16 N223S, (n=4) CLDN18 V88I (n=13), and CLDN24 G94R (n=5) and L177M (n=5) gave similar results as wild-type CLDN6 (n=11), CLDN16 (n=15) CLDN18 (n=7), and CLDN24 (n=4), respectively. A significant increase in open NTDs was observed in embryos electroporated with CLDN19 I22T (n=5/12, χ2=4.31, p = 0.0379) and E209G (n=8/12, χ2=8.21, p = 0.004) compared to wild-type CLDN19 (Fig. 4.4 A, B). Apical constriction occurred normally in embryos overexpressing CLDN19 E209G (Fig. 4.5A). However, these embryos were significantly shorter than embryos overexpressing wild-type CLDN19 (p<0.005) suggesting that convergent extension is responsible for the NTDs observed in these embryos (Fig. 4.5B). Neither apical constriction nor convergent extension (p>0.05) were affected in embryos overexpressing CLDN19 I22T suggesting that the NTDs are caused by defective fusion of the neural folds (Fig. 4.5A, B).

4.4.14 Variants in the Cldn8 cytoplasmic C-terminus disrupt neural tube closure No rare or novel missense variants were identified in CLDN8. This is expected as analysis of Cldn4 and -8-depleted-chick embryos suggests that deleterious mutations in CLDN8 would cause a more severe type of NTD that is embryonic lethal (Baumholtz et al., 2017b). To study the role of Cldn8 in neural tube closure we targeted the C-terminal tail of chicken Cldn8 for mutagenesis based on our hypothesis that direct or indirect interactions with the C-terminal tails of C-CPE-

131 sensitive claudins, Cldn4 and -8, are responsible for maintaining Vangl2, Par3 and Rho-GTPases localized to the apical membrane. We generated three variants at putative serine phosphorylation sites (Cldn8 S198A, Cldn8 S216A, Cldn8 S216I) and one that deleted the PDZ binding domain at the C-terminus of Cldn8 (Cldn8∆YV) (Fig. 4.6A). All variants localized to tight junctions in transfected HEK293 cells (Fig. 4.6B). Electroporation of neural plate stage embryos with wild- type Cldn8 expression vector did not significantly affect neural tube closure and convergent extension (Fig. 4.6C); 92% had closed neural tubes (n=45/49) and 96% were normal length (n=47/49). Similarly, ≥90% of the embryos electroporated with Cldn8∆YV (n=39/42) and Cldn8 S216A (n=45/50) had closed neural tubes. A significant increase in NTDs was observed in embryos electroporated with the Cldn8 S216I variant (n=21/65; χ2=9.5, p<0.002) and more than two-thirds of these embryos also had convergent extension defects (n=15/21) (Fig. 4.6C). Histological analysis of transverse sections through the neural tube of embryos overexpressing the Cldn8 S216I variant revealed a flat and broad midline indicating that apical constriction of midline neural ectoderm cells was also affected (Fig. 4.6D). Interestingly, the Cldn8 S198A variant also caused a significant increase in open NTDs (n=7/17; χ2=9.9, p<0.002) but did not affect convergent extension (n=2/17; χ2=1.31, p=0.253). These data suggest that residues within the Cldn8 C-terminus have distinct interaction partners that regulate different morphogenetic events during neural tube closure.

4.4.15 Residues in the Cldn8 cytoplasmic C-terminus regulate protein localization at the apical surface of neural ectoderm cells We showed that removal of Cldn4 and 8 from the neural ectoderm of C-CPE-treated embryos results in reduced localization of the RhoGTPases Cdc42 and RhoA at the apical membrane of neural ectoderm cells (Baumholtz et al., 2017b). We looked at whether specific residues in the Cldn8 C-terminal tail regulate localization of Cdc42 and RhoA. Cdc42 and RhoA partly co- localized with the tight junction marker ZO-1 at the apical of membrane of neural ectoderm cells in embryos overexpressing wild-type Cldn8 (n=3) (Fig. 4.6E,F). However, we observed the loss of Cdc42 from tight junctions in the neural ectoderm of embryos overexpressing the S198A variant (n=2/2) and increased Cdc42 localization to tight junctions in the neural ectoderm of 60% of embryos overexpressing the S216I variant (n=3/5) (Fig 4.6E). Furthermore, RhoA showed reduced localization to tight junctions in the neural ectoderm of 50% of embryos overexpressing

132

the S216I variant (n/3/6) (Fig. 4.6F). These data suggest that specific residues in the C-terminal tail of Cldn8 participate in unique interactions with signaling proteins at the tight junction cytoplasmique plaque to regulate the morphogenetic movements and cell shape changes required for neural tube closure.

4.5 DISCUSSION Recently, we showed that C-CPE removal of Cldn3, -4 and -8 from tight junctions in the neural (Cldn4 and-8) and non-neural ectoderm (Cldn3 and -4) results in NTDs in chick and mouse embryos (Baumholtz et al., 2017b). Morphological analysis of chick embryos revealed that claudins play a critical role in regulating two morphogenetic events that drive neural tube closure – convergent extension and apical constriction. Our data suggest that claudins play an evolutionarily conserved role in neural tube closure; however, the possible contribution of rare and novel mutations in CLDN genes to human NTDs has not yet been studied. In this study, we analyzed the human CLDN1-25 genes in a cohort of 125 NTD patients with myelomeningocele, an open spinal NTD. We identified 10 rare variants in 11 NTD patients and 4 novel variants in 6 NTD patients. Nine of the CLDN variants (CLDN3 p.A128T and p.P134L, CLDN9 p.S9L, CLDN16 p.N223S, CLDN18 p.V88I, CLDN19 p.I22T and p.E209G, and CLDN24 p.G94R and p.E161K) were predicted to be pathogenic by at least one prediction program.

Transfection experiments in HEK293 cells suggested that CLDN16 N223S impedes the trans-interaction of CLDN16 molecules between adjacent cells; however, functional validation using an overexpression assay in chick embryos showed this variant did not did not cause NTDs suggesting that adhesion interactions between CLDN16 are not required for neural tube closure. In contrast, overexpression of CLDN19 I22T and E209G caused NTDs due to defective neural fold fusion and convergent extension movements, respectively. CLDN19 p.I22T is located in the first transmembrane domain and introduces a putative phosphorylation site that could disrupt heteromeric cis interactions between claudins. CLDN19 p.E209G causes loss of a negatively- charged amino immediately adjacent to the cytoplasmic C-terminal PDZ-binding domain, which interacts with PDZ-domain adaptor proteins at the tight junction cytoplasmic plaque to form a signaling platform. These interactions also connect the tight junction to the actin cytoskeleton and are predicted to have critical roles in morphogenesis (Baumholtz et al., 2017a). Cldn19 null mice have closed neural tubes (Miyamoto et al., 2005) suggesting that a mutant allele may be

133

more detrimental than a null allele. This would not be surprising as knockout or knockdown of a single claudin often results in compensatory upregulation of another family member (Li et al., 2014). In contrast, a mutant allele would not affect the total levels of Cldn19 mRNA or protein but would result in production of a mutant protein that competes with wild-type protein for incorporation into tight junction strands but would be non-functional.

Components of the planar cell polarity (PCP) pathway have a well-established role in regulating convergent extension movements during neural tube closure (Curtin et al., 2003; Goto and Keller, 2002; Greene et al., 1998; Paudyal et al., 2010; Wallingford and Harland, 2002; Wang et al., 2006a; Wang et al., 2006b; Ybot-Gonzalez et al., 2007b) and mutations in PCP pathway genes have been recognized as risk factors for human NTDs (Allache et al., 2012; Bosoi et al., 2011; De Marco et al., 2013; De Marco et al., 2012; Kibar et al., 2009; Kibar et al., 2011; Lei et al., 2010; Robinson et al., 2012). We previously showed that claudins regulate convergent extension by acting upstream of PCP signaling during neural tube closure (Baumholtz et al., 2017b). A portion of the cohort analyzed in this study was also screened for variants in PCP genes. Interestingly, four patients with a rare or novel mutation in a CLDN gene also carried a mutation in a PCP gene. While these CLDN variants did not cause NTDs in our overexpression assay in the chick embryo, we could be looking at a two-hit model for NTDs where both variants act synergistically to confer a high risk for NTDs. This two-hit model for NTDs has been observed in mice carrying mutations in PCP genes where homozygous mice have a closed neural tube (e.g. Fzd3-/- and Fzd6-/-) (Guo et al., 2004; Wang et al., 2002), while double homozygous mice (e.g. Fzd3-/-;Fzd6-/-) develop NTDs (Wang et al., 2006b). Furthermore, heterozygous mice with spinal NTDs (e.g.Vangl2Lp/+) develop craniorachischisis, a severe NTD that affects the entire embryonic axis, when crossed to heterozygous mice carrying a mutations in a second PCP gene (e.g. Vangl2Lp/+; ScribCrc/+, Vangl2Lp/+;Celsr1Crsh/+, Vangl2Lp/+;Ptk7chz/+) (Murdoch et al., 2014). It would be interesting to test the hypothesis of genetic interactions among the CLDN and PCP variants detected in the same patient. In addition, the patients who carried the CLDN3 p.P134L and CLDN4 p.V84I variants and the second patient with the CLDN6 p.R209Q variant should be screened for mutations in PCP genes.

The mechanisms involved in spinal NTDs are thought to differ from those of cranial NTDs; apical constriction is more important for cranial neural tube closure, while convergent extension

134

plays a more critical role in spinal neural tube closure. This is consistent with the observation that overexpression of the CLDN19 E209G variant identified in a patient with a spinal NTD caused defective convergent extension but not apical constriction when overexpressed in the chick embryo. We have shown that claudins are required for apical constriction and convergent extension and thus, mutations in CLDN genes may be implicated in both cranial and spinal NTDs. In the future, it would be interesting to sequence genomic DNA from patients with cranial NTDs to compare the pathogenesis of rare and novel CLDN variants in the different types of NTDs.

We showed that the S198A and S216I Cldn8 variants in the C-terminal cytoplasmic tail disrupt interactions between Cldn8 and signaling proteins at the tight junction cytoplasmic plaque leading to NTDs. Overexpressing the S216I but not the S198A Cldn8 variant caused convergent extension defects suggesting that specific residues in the C-terminal domain of claudins have unique roles in regulating intracellular signaling events that affect neural tube morphogenesis. Furthermore, mutating S198 to a phosphorylation null variant (S198A) caused NTDs suggesting that phosphorylation can influence the interaction of Cldn8 with proteins at the tight junction cytoplasmic plaque. Suprisingly, deleting the PDZ-binding domain of Cldn8 (Cldn8ΔYV) did not affect neural tube closure suggesting that interactions between the Cldn8 C- terminus and signaling proteins at the tight junction cytoplasmic plaque is not mediated by cytoplasmic PDZ adaptor proteins.

We did not identifiy rare or novel missense mutations in CLDN8 in our cohort of patients with spinal NTDs. This is not surprising because removal of Cldn4 and 8 from tight junctions causes a more severe form of NTDs in chick embryos in which the neural tube remains open along the entire anterior-posterior axis called cranioraschischisis. Thus, patients with this embryonic lethal form of NTDs should be screened for CLDN8 mutations.

In conclusion, our study supports the hypothesis that rare deleterious variants in CLDN genes are significant risk factors for human spinal NTDs. To our knowledge, this is the first study to identify deleterious variants in CLDN genes in patients with NTDs, which further supports the concept that human NTDs are a group of disorders with high genetic heterogeneity, and both CLDN and PCP genomic variants contribute to their occurrence. These results enhance our

135 understanding of the pathogenesis of NTDs in humans, highlighting the importance of the claudin family as new candidates for NTDs in humans.

4.6 ACKNOWLEDGMENTS We would like to thank P. Lepage (Genome Quebec) for technical assistance.

AIB is the recipient of a doctoral fellowship from the Fonds recherché du Québec – Santé (FRQS, 29939). This work was supported by the Natural Sciences and Engineering Research Council of Canada (234319) (AKR) and a pilot project grant from the Montreal Children’s Hospital Foundation (AKR). AKR is a member of the RI-MUHC, which is supported in part by the FRQS.

136

Figure 4.1. Chromatograms of rare and novel nonsynonymous CLDN mutations. Mutations are shown in the sense (5’→3’) direction. The position of altered amino acids are indicated with an asterisk. WT, wild-type sequence.

137

Figure 4.2. Rare and novel CLDN mutations in NTD patients. (A) Schematic of a Claudin protein. The approximate positions of mutations identified are shown. Nonsynonymous mutations that are predicted to be deleterious by at least one prediction program by in silico analysis are boxed in red. (B) Partial ClustalW protein sequence alignment of human CLDNs with orthologues from other species. (accession numbers in Table 4.1) The amino acid position of CLDN variants found in NTD patients are indicated by arrows and residues conserved at this position in other species are highlighted in grey.

138

Figure 4.3. Subcellular localization of CLDN variants. Wild-type (WT) and variant claudins (green) co-localized with the tight junction adaptor protein ZO-1 (red) at the cell membrane in transiently transfected HEK293 cells (arrow). Blue indicates DAPI stain of the nucleus.

139

Figure 4 4. Open NTDs and convergent extension defects in chick embryos overexpressing CLDN variants. (A) Proportion of embryos overexpressing wild-type (WT) or variant claudin constructs with a closed neural tube (light grey bars) or open NTDs (dark grey bars). *p<0.05, ***p<0.005. (B) Dorsal view of chick embryos electroporated with Cldn-IRES-GFP constructs showing a closed neural tube and complete, cranial or spinal NTD; brightfield (left) and GFP expression (right). Dashed lines outline open neural tubes.

140

Figure 4.5. Embryos overexpressing CLDN19 E209G, but not CLDN19 I22T, exhibit convergent extension defects but apical constriction is not affected by CLDN19 variants. (A) Transverse sections showing Shh expression in embryos electroporated with wild-type (WT) CLDN19, CLDN19 I22T and CLDN19 E209G. Dashed line in higher magnficiation view marks Shh expression domain in floor plate. Scale bar, 50 µm. (B) Graph shows anterior-posterior (AP) length of embryos electroporated with wild-type CLDN19, CLDN19 I22T and CLDN19 E209G at the neural tube stage (***p<0.005, Mann-Whitney U-test).

141

Figure 4.6. S198 and S216, but not the PDZ-binding domain, are critical residues for Cldn8 function. (A) Schematic of Cldn8 indicating the residues (blue) that were mutated to evaluate critical residues in the Cldn8 cytoplasmic C-terminus. (B) HEK293 cells transfected with wild-type Cldn8 and Cldn8 variants show colocalization of Cldn8 staining with ZO-1, a tight junction marker. (C) Dorsal view of chick embryos electroporated with Cldn-IRES-GFP constructs showing a closed neural tube, an open NTD, a convergent exention defect (CED) and an open NTD and CED; brightfield (left) and GFP expression (right). Dashed lines outline open neural tubes. (D) H&E staining of transverse sections of the neural tube of embryos overexpressing wild-type (WT) Cldn8 and Cldn8 S216I. (E, F) Apical surface views of Cdc42 (E) and RhoA (F) in the neural ectoderm of embryos ovexpressing wild-type Cldn8 and Cldn8 variants.

142

Table 4.1. Common nonsynonymous variants in the coding sequence of CLDN1-25

Nucleotide Amino acid MAF (%) in MAF (%) in MAF (%) in Gene change change patients ExAC EVS CLDN6 c.388.C>G V130L 0.4 1.383 1.2388 c.427T>C I143V 41.5 36.14 38.5741 CLDN7 c.590T>C V197A 75.6 65.91 30.32 CLDN8 c.290T>C M97T 0.4 0.5126 1.7607 c.385A>G T129A 7.3 5.36 8.4807 c.451T>C S151P 27.6 29.74 38.4284 CLDN14 c.11C>T T4M 5.0 4.029 4.3946 CLDN16 cC164del A6LfsTer16 29.0 19,45 0 c.166C>G A56P 29.0 19.45 0 CLDN17 c.244G>A A82T 11.7 9.437 6.8584 CLDN18 c.445A>T M149L 5.9 11.48 10.7498 CLDN19 c.599G>A R200Q 0.9 2.681 1.1422 CLDN20 c.281G>A G94E 0.8 1.312 0.915 CLDN23 c.628G>A V210M 12.7 20.79 7.8521 CLDN24 c.52G>A L18F 80.8 76.73 N/A c.391T>C I131V 0.4 1.116 N/A c.621C>G Q207H 19.2 28.52 N/A

143

Table 4.2. Novel and rare nonsynonymous variants in the coding sequence of CLDN1-25

Nucleotide Amino acid MAF (%) in MAF (%) MAF (%) Mutation Gene change change patients in ExAC in EVS Polyphena SIFTa Tastera CLDN3 c.382G>A A128T 1.2 0 0 - + + c.401C>T P134L 0.4 0.45 0.48 + + + c.621G>C G207A 0.4 0 0 - - - CLDN4 c.250G>A V84I 0.4 0.002478 0.00002447 - - - CLDN6 c.626C>T R209Q 0.8 0.15 0.15 - - - CLDN9 c.8C>T S3L 0.4 0.1 0.0077 + + + CLDN16 c.668A>G N223S 0.4 0.03459 0.0231 + - + CLDN18 c.262G>A V88I 0.4 0.23 0.17 + - + CLDN19 c.43T>C I22T 0.4 0.11 0.0616 + + + c.6A>G E209G 0.4 0.006625 0 + - - CLDN23 c.268G>A A90T 0.4 0.52 0.46 - - - CLDN24 c.280G>C G94R 0.4 0 N/A + + + c.481G>A E161K 0.4 0.002435 N/A + + + c.529C>A L177M 0.4 0 N/A - - - aPolyphen: +, probably or possibly damaging; -,benign; SIFT: +,damaging; -, tolerated; MutationTaster: +,disease causing; -, polymorphism

144

Table 4. 3. Patients carrying mutations in a CLDN gene and core PCP gene

CLDN gene CLDN variant PCP gene PCP variant CLDN3 p.A128T CELSR1 p.G934R CLDN6 p.R209Q VANGL1 p.C145C CLDN16 p.N223S DVL2 p.T535I CLDN23 p.A90T DVL2 p.T535I

145

Supplemental Table 4.1. Fluidigm primers

Fragment Fragment Forward primer sequence Reverse primer sequence size (bp) CLDN1E01 CCATGGAATCACACAACAGAA AACTCTCCGCCTTCTGCAC 371 CLDN1E02 TGTTTGCAGTTTGCCTTAGA TTCCATTTTTCTCTTGTTGGTC 340 CLDN1E03 ATGGCACTAGCAGGACTTTG TGGACTTCTAATCTCCCTAATACC 315 CLDN1E04 TCCATTTTCGGTTTGTTTCA CAGAAATCTTAAAGTACTTCCCAAGG 358 CLDN2E02_001 GGTGTTCAAGGAGCAAGAGC GGAGGAGATTGCACTGGATG 352 CLDN2E02_002 GCATCACCCAGTGTGACATC ATGAGCAGGAAAAGCAGAGG 381 CLDN2E02_003 TTCTTCCCTGTTCTCCCTGA CATCCTTGGCAATTCAACCT 386 CLDN3E01_001 GGTGGTGGTGTTGGTGGT CCAAGGCCAAGATCACCAT 381 CLDN3E01_002 CACCACGGGGTTGTAGAAGT CTTCATCGGCAGCAACATC 355 CLDN3E01_003 AGCAGCGAGTCGTACACCTT CCGTCGGTGAGTCAGTCC 340 CLDN4E01_001 AGCCTTCCAGGTCCTCAACT CCTCCAGGCAGTTGGTACAC 365 CLDN4E01_002 CATCATCGTGGCTGCTCTG AGCAGAGAGGAACAGAGTGGA 402 CLDN5E02_001 CCCCAGGCTTATCCAACG GAGGCGTGCTCTACCTGTTT 361 CLDN5E02_002 GACAATGTTGGCGAACCAG GTCCGCAGCGTTGGAGAT 424 CLDN5E02_003 GCACGCCAGGATCAGACC CCCTAACTTCAGCTGCCAGA 374 CLDN6E01_001 GCCTCCGCATTAGTTCCATA CATTACATGGCCCGCTACTC 372 CLDN6E01_002 TTCTTGGTAGGGTACTCAGAGG CACGTGCCCTCTGTGTCAT 415 CLDN6E01_003 TCCTCCACACAGGTGGTACA TGCTTCTGTCCCAAACACAG 370 CLDN7E01 CCAGCCGACCACTTCCTC CGTTTGTTTTACTGTAGGGTCTCC 417 CLDN7E02 TCAGTATAGTGAGGCCCCAAA GGCCCAGGTCTTGGACAC 393 CLDN7E03 GGACAGGAACAGGAGAGCAG CCATCTGGGAGGAGCAAG 353 CLDN7E04 AGGCCCTTTCAGGCATCTA CCCTTTGATCCCTACCAACA 354 CLDN8E01_001 CAAAGTTTCTTTGGGGTCCA GTGGTGCTCATCCCTGTGA 374 CLDN8E01_002 AGAGCTTCTCCAAGCTCACG ACTGTGGATGAATTGCGTGA 348 CLDN8E01_003 AGCCAGCAGGGAATCATAGA CTGGCCAGAAGTAGCAAAGC 376 CLDN9E01_001 GGGGCTGAGAAGACCTAACC GGCGAGGAGGAGGATGAC 414 CLDN9E01_002 TGTACCACGTGTGTGGAGGA ATCAGGCCAAGGTCGAAAG 409 CLDN9E01_002rd AAGGCCCGTATCGTGCTC ATCTGGTCATCAGGCCAAG 387 CLDN10E01_001 ACAGGGCATGGGTGTGAG AGGGAAGGAGGGCTGAGG 437 CLDN10E02 TTTTTGGAAAACAAACATCCA TGAGCACAGCCCTAACAAAT 361 CLDN10E03 TTTGGCTGGGATTGTATTCAT TGCATATTTTGCGTATGTGG 375 CLDN10E04 TTGGGATGGTCTAATGGCTA CAGGTCATTTTTGTCTCTTTTTC 334 CLDN10E05 ACTTCTTGGGGCAAGAGGAG AATTATGGGAGGGCCTTGAT 352 CLDN11E01 GTACCTGGGCAGGCACTGT TATCCCGCTCTCTACCCAGA 366

146

CLDN11E02 AGAAGGAGGAAGGGAGATGG CTTCACTTGGCTTCCTTTCA 379 CLDN11E03 TGGAAGCCACAAGTGTGTGTA CACACTGTGACGAGCAAACC 405 CLDN12E04_001 TCTGACTGACAGTACTCCACAAG CTGAAGGCAGTGTTGCACAT 374 CLDN12E04_002 AGTTTGCCCTACCCCTCAG TGGGTGGATGGGAGTACAAT 387 CLDN12E04_003 TTTGAGCCAGTCTTTTCATTTG TTGGCTTCATTGATTGGTCA 360 CLDN14E01_001 CTGCCTCCATTGACAGTCC TCCTGGACCACCAACGAC 382 CLDN14E01_002 GGAGATGAAGCCCAGGTACA TCCTACCTGAAAGGGCTCTG 378 CLDN14E01_003 ATCGGTAGATCTGGCACTGG TCGGTGACAGAAATAAGTGCAT 368 CLDN15E01 TGCATCCTCACGGAAGTACC TTCCAGTTCCCTAGGGGTTC 376 CLDN15E02 TCTGGGGAGTACAGATGAGG GATGGGAAAGGCTGACAAC 361 CLDN15E03 CAGGATGGAGATCAGTGAGG CTGCTCTGGGACTTGGTG 372 CLDN15E04 TAGGCGTTTCTGCCGTATTT GGCTGGCGTCTCACTTGT 352 CLDN15E05 CGGCCCCTGAGGTTACTA CCTCACTGATCTCCATCCTG 366 CLDN16E01 CCACCCGAAACACACTCAG GGCCTGGATCATGAAAAGAA 426 CLDN16E02 GGCTTCAATTGTCAGTGCTT TTTTCTGTCCCTTTCCCTTC 342 CLDN16E03 AGGGGTACTTATGTTCAAGTTCAT AGCAGCTTCAGCACAACTCT 403 CLDN16E04 TGTAGCATCCTCCCTTTCTTT TGCCCTTGAACAATTGGA 359 CLDN16E05 TTTCCCAAGTTCACTGAGTTCT AAAGAAAAAGTATAGGAGAATCAAACA 343 CLDN17E01_001 GCAAGTTCTCCTGCCTCAC GTGAGCTGGACAGCCAATA 358 CLDN17E01_002 GATGGCTGGGTTGTAGAAATC TTTGAGAGGCTCTGGGAAG 333 CLDN17E01_003 ACAAGGAGCTATAGAACTTGCATT GTTAGGCCAAGTTCAGTCACA 375 CLDN18E01 CACCAGCCTCTCAGAGAAAA GTTTCCTCTCCACCTCCAAT 373 CLDN18E02 TTTGTCTGCTTGTGTCTTGC TTGGACCTCCACACTCAGAT 341 CLDN18E03 CAATATTCTGCAGCCTACTCATC ATGGCATGGTTGTCTCTGAT 354 CLDN18E04 ACCATATTGACAGCCACCAT GGCTGAAATATTCCCATTCTG 323 CLDN18E05 CAAAGACATCTACAATCATGGAAT AGACTGAGGCTAAGACCATTTG 375 CLDN19E01 CTGTTCCCACCTCCCATCT CTGCCTCTGACCCTCCTTCT 396 CLDN19E02 GTGCAGAGGCCTAAAGACAA CTCTCAAGCTGGGCTCCT 372 CLDN19E03 TCCAGTGGACAAAGGTCAGT GGAGACAGCAACCCCATT 324 CLDN19E04 CCTGCCTCTGGTGTCTCTCT CCTGATGCCACTCTCCCTAC 380 CLDN19E05 GACAGACCGAATGATACCATGA GCCACCTACACAGATGGTGA 352 CLDN20E02_001 AGAATTCTGACAGCCATCATTG TGTCCCCTCCTAAGCGAGTA 398 CLDN20E02_002 GCTTTGGGGATCTGCACTT GCACCCAAAAGTTCCATTTC 426 CLDN20E02_004 TCATTTCTGCAATGCTGTTG AAACATGGAAACAATAATGGAAA 375 CLDN22E01_001 GGGACATCCTACCAAATCCA TGATCCTGGGAGGAATTCTG 375 CLDN22E01_002 TCTCATCCCAGAACTCCTGAA CTTTGCTGGGATGGGTTTTA 405 CLDN22E01_003 ACACAGGTTTGCCAGAGTCC ACCTTGGCATTACTGGTTGG 379 CLDN23E01_001 CGACAGCGGAGAAGGAAG CACGAAGTTGGGCTCGTC 391 CLDN23E01_002 GTCCTGGGGCTTCTGCTG GGCCGTCGCTGTAGTACTTG 409 CLDN23E01_003 AGCAGCGTCAGCACCATC GAAAGGCAGATTTCCATCCA 394

147

CLDN24E01_001 ACATCTCTGGCCTTGTCG TTTGAGAATTGGAGAGAGTCAGA 375 CLDN24E01_002 CAGAATTCCTCCCAGAATGA TGGCTTTAATCTTTAGAACAGCA 374 CLDN24E01_003 TGGCAAATAAGTTGTAATAATGGA AGCCCTCTTTGTTCATCCTT 393 CLDN25E01_001 TTGGACACACCCTCTAAACC CTCCCAAATCCCCATGAT 349 CLDN25E01_002 TGGGTCTGCTCCTGTGTTAC GGAGACTGGAAGGAGGGTAG 354 CLDN25E01_003 GTCTCCTGGGAAGGACTTTG GGCAGAAGCCAGTCCTAGAT 349 CLDN25E01_004 TATTTTCCTGGCTCTTGGTG AAAAATACATTTGTTAGACACTGTGC 363 CLDND1E01 AGGGAAGTGCGTCAGAGGAG CTGCAGCAGCCACCTCCT 346 CLDND1E02 TTATTCCAAATGCACCCTTT TCAAACCCAGAACACTTAATCC 341 CLDND1E03_001 GGCAATCATTACGTCGTTTT CAGCATTTGTAATTGCTTGTG 343 CLDND1E03_002 TGAATTCATCCCAGATGCTT AGCTTAATTAAAAACATGCTTGC 343 CLDND1E04 ATCTTTGTTCCCCCAACTGT GAGAACCAGTGGCATCCAT 328 CLDND1E05 CAGGAAGAAGCTAAAAGGTAAGTATG GCATCCTTTTGAATAATGTGCT 339 CLDND1E06 TTTCATAATAAAATGGTATATCCACAA TTTTGTTTTGTTTTCTAAATTTTCC 370

148

Supplemental Table 4.2. Accession numbers of claudin amino acid sequences used for ortholog alignment

Species Gene Accession Number Human CLDN3 NP_001297.1 CLDN4 NP_001296.1 CLDN6 NP_067018.2 CLDN9 NP_066192.1 CLDN16 NP_006571.1 CLDN18 AAF26448.1 CLDN19 AAQ07256.1 CLDN20 AAH20838.1 CLDN23 NP_919260.2 CLDN24 NP_001172078.1 Chimpanzee CLDN3 XP_016813082.1 CLDN4 XP_016800774.1 CLDN6 XP_016802155.1 CLDN9 XP_016784715.1 CLDN16 XP_001161102.2 CLDN18 XP_526318.3 CLDN19 XP_003949437.2 CLDN20 XP_009450550.1 CLDN23 XP_528065.3 CLDN24 XP_003310625.1 Dog CLDN3 NP_001003088.1 CLDN4 XP_005621019.1 CLDN6 XP_547168.3 CLDN9 XP_005621712.1 CLDN16 XP_850268.2 CLDN18 XP_534274.2 CLDN19 XP_848612.2 CLDN20 XP_005615593.1 CLDN23 XP_005629968 CLDN24 XM_540033.4 Mouse Cldn3 NP_034032.1 Cldn4 NP_034033.1 Cldn6 NP_061247.1 Cldn9 NP_064689.2 Cldn16 NP_444471.1 Cldn18 AAF26447.1 Cldn19 NP_694745.1 Cldn20 NP_001095030.1 Cldn23 NP_082274.1 Cldn24 NP_001104788.1 Chicken CLDN3 NP_989533.1 CLDN4 XP_003642430.1

149

CLDN16 XP_426702.1 CLDN18 XP_003641778.1 CLDN19 ENSGALT00000039213.2 CLDN20 XP_004935700.1 CLDN23 XP_004941217.2 Xenopus Cldn3 NP_001087400.1 Cldn4 AAY16547.1 Cldn6 NP_001086756.1 Cldn16 XP_002934087.1 Cldn18 NP_001083443.1 Cldn19 NP_001087995.1 Cldn23 XP_004911118.1 Cldn24 XP_018098209.1

150

Supplemental Table 4.3. Primer sequences (5’->3’) used for PCR

Gene Forward primer Reverse primer CLDN3 CGCATATGTCCATGGGCCTGGAGATC CGTTAGACGTAGTCCTTGCGGTC CLDN4 CGCTTGGAATCCTACGGCCC GCGCTGAGCTCAGTCCAGG CLDN6 CCTGTCCACATGTGGCCTGA GTTGGGCACTGCCACTTCT CLDN8 GGTTCCGAGTTCATTACTACAG GATTAGGCAGTTAAGAACAGTA CLDN9 CGCTCCTGCTGGACACAGAGAC GCGGCATCTGGTCATCAGGCC CLDN14 GCTTCATTAGGGCTCCGGCTG TGTGCTGGAACCCCTGCCTC CLDN16 exon 4 GCCTCTGAATCACGCCAGCC GCTTGTTTCCACTGGATTGGGA CLDN18 exon 2 CGGCTGTCTCTCAGAGGTTTGG GCGAACCACATTCAGCAGTAGG CLDN19 exon 1 CCAGCTGCTCCTCCCACCTG GCCTACCGTCCAGGGCGAG CLDN19 exon 4 GCGGCTCCAGCCTCCAGCTCC GCGGGTGCTGCGAGGCTGGC CLDN23 CGGGAAGGCAGGCTGCAGGG GCCCAGGCTCTACAAGCGTCTA CLDN24 CGCCTGTCGCAATGGCTTTAATC CGTTACACTTGAGGATCTGCTG

151

Supplemental Table 4.4. Primer sequences (5’->3’) used for RT-PCR

Gene Forward primer Reverse primer CLDN19 GCCATATGGCCAACTCAGGCCTGG GCTCAGACGTACTCTCGGGCAG

152

Supplemental Table 4.5. Primer sequences used for mutagenesis

Cldn16N223S-5’ ggaacgttctactttggttttgcacagtatatttcttggtatcc Cldn16N223S-3’ ggataccaagaaatatactgtgcaaaaccaaagtagaacgttcc CLDN18V88I-5’ gatgatcgtaggcatcatcctgggtgccattg CLDN18V88I-3’ caatggcacccaggatgatgcctacgatcatc

153

5 CHAPTER V: DISCUSSION

154

5.1 OVERVIEW AND MAJOR FINDINGS Vertebrate neurulation involves a series of complex morphogenetic processes that leads to formation of a closed neural tube. Primary neurulation, the process during which the flat neural plate rolls up and closes to form the neural tube, occurs in four distinct steps: the neural plate is induced to differentiate, it then undergoes convergent extension movements resulting in its anterior-posterior lengthening and mediolateral narrowing, subsequent bending of the neural plate creates the neural folds, and finally the dorsal tips of the neural folds fuse to complete neural tube closure. These morphogenetic processes involve both the neural ectoderm and the surrounding non-neural ectoderm. Failure of any of the morphogenetic events of neurulation leads to perturbations in neural tube closure, resulting in neural tube defects (NTDs). Despite the fact that mandatory folic acid supplementation in North America has reduced the incidence of NTDs by 50-70%, they still remain the second most common birth defect affecting 0.5-1 per 1000 births annually. Thus, there remains an urgent need to understand the mechanisms underlying NTDs not prevented by folic acid supplementation. The cellular and molecular mechanisms that regulate neural tube closure need to be coordinated in a spatiotemporal manner. Neural plate induction requires binding of inhibitory molecules to BMPs blocking their signaling and activation of FGF signaling (Hemmati- Brivanlou and Melton, 1997; Streit et al., 2000; Streit et al., 1998). The process of convergent extension, which is responsible for neural plate shaping, is linked to the non-canonical Wnt/planar cell polarity (PCP) pathway (Goto and Keller, 2002; Wallingford and Harland, 2002; Wang et al., 2006a). Actin-myosin dynamics play a key role in neural plate bending by regulating formation of the median and paired dorsolateral hinge points. Apical constriction of neural ectoderm cells depends upon the actin-myosin contractile force generated at the apical surface of neuroepithelial cells by localized myosin light chain phosphorylation downstream of RhoA-ROCK signaling (Escuin et al., 2015; Kinoshita et al., 2008; Rolo et al., 2009). RhoA- ROCK-dependent disassembly of actin filaments is also critical for neural plate bending as apical actin accumulation results in a stiff neural plate that resists bending (Escuin et al., 2015). Little is understood about the mechanisms that regulate fusion of the neural folds which is divided into three steps: initial contact, epithelial adhesion, and remodeling to generate a closed neural tube overlain by a continuous layer of non-neural ectoderm. In the mouse, initial contact between the

155

apposed neural folds is mediated by cellular protrusions emanating from the neural and non- neural ectoderm under the control of the RhoGTPases Rac1 and Cdc42 (Rolo et al., 2016). In the chick, few cellular protrusions are observed between the apposed neural folds (Bancroft and Bellairs, 1975; Lee et al., 1976; Lee et al., 1978). Instead, a network of fibrils consisting of extracellular matrix proteins connects the apposed neural folds. Ephrin-Eph interactions have been implicated in neural fold adhesion. Our previous analysis of the claudin mRNA expression patterns in chicken embryos revealed that 14 claudins are expressed during neurulation and that a subset of these claudins were differentially expressed in the neural and non-neural ectoderm: Cldn8, -14 and -17 were enriched in the neural ectoderm, while Cldn3, -11, -12 and -22 were more highly expressed in the non-neural ectoderm (Collins et al., 2013; Simard et al., 2005). This observation suggests that the combination of claudins expressed in the ectoderm creates signaling compartments that regulate differentiation and morphogenesis of the neural and non-neural ectoderm. NTDs have not been reported in any of the single claudin knockout mouse lines (Anderson et al., 2008; Ben- Yosef et al., 2003; Fujita et al., 2012; Furuse et al., 2002; Gow et al., 1999; Kage et al., 2014; Li et al., 2014; Miyamoto et al., 2005; Morita et al., 1999; Tamura et al., 2008). However, Cldn4, - 6 and -7 are downregulated in Grhl2 mutant mouse lines, which exhibit open NTDs (Pyrgaki et al., 2011; Senga et al., 2012; Werth et al., 2010). These data suggest that claudins may have functionally redundant roles during neural tube closure. The first objective of this thesis was to determine if claudins are required for neural tube closure. In Chapter II, I used the C-terminal domain of Clostridium perfringens enterotoxin (C- CPE) as a tool for studying the functional redundant roles of claudins in neural tube closure. I showed that C-CPE removal of Cldn3 and -4 from the non-neural ectoderm and Cldn4 and -8 from the neural ectoderm of neural plate stage chick embryos caused folate-resistant open NTDs. Analysis of the different steps of neural tube closure revealed that differentiation of the ectoderm into neural and non-neural progenitors occurred normally, but that C-CPE-sensitive claudins are required for convergent extension movements and apical constriction of cells at the median hinge point (MHP). Furthermore, I showed that Vangl2, a core planar cell polarity (PCP) protein, showed reduced membrane localization in neural ectoderm cells of C-CPE-treated embryos suggesting that claudins intersect with the PCP pathway to regulate convergent extension. In addition, I showed that junctional localization of RhoA was reduced leading to reduced levels of

156

pMLC at the apical surface of neural ectoderm of C-CPE-treated embryos. Localization of Cldn3, -4 and -8 to the tight junction was not affected in embryos treated with the ROCK inhibitors Y27632 or GSK269962. These data position claudins upstream of RhoA-ROCK signaling in regulating apical constriction of MHP cells. These novel findings suggest a role for Cldn4 and -8 in regulating morphogenesis of the neural ectoderm during neural tube closure. The role of claudins in fusion of the dorsal tips of the neural folds to form a closed neural tube could not be evaluated in C-CPE-treated embryos due to defects in earlier phases of neural tube closure. In Chapter III, I explore the role of Cldn3 in the non-neural ectoderm in regulating neural fold fusion. I showed that neural plate stage chick embryos treated with a C-CPE variant that specifically targets Cldn3, C-CPELDR, develop folate-resistant NTDs. Unlike embryos treated with wild-type C-CPE, convergent extension and apical constriction are not affected in C- CPELDR-treated embryos. Furthermore, dorsolateral hinge point formation occurs normally. Collectively, these data suggest that the NTDs observed in CPELDR-treated embryos are caused by defective fusion of the neural folds. Scanning electron microscopy revealed an extracellular matrix-based meshwork of fibrils connecting the apposed neural folds of control GST-treated embryos that was absent in C-CPELDR-treated embryos. This observation suggests that Cldn3 mediates the initial contact between the dorsal tips of the apposed neural folds by regulating the secretion of extracellular matrix proteins into the extracellular space. The data presented in Chapter II and III indicate that claudins regulate morphogenesis of the neural and non-neural ectoderm during neural tube closure in the chick embryo. We also showed that mouse embryos treated with C-CPE exhibit open NTDs. These data indicate that claudins play an evolutionarily conserved role in neural tube closure and potentially implicate claudins in human NTDs. In Chapter IV, I explored the possibility that rare and novel missense mutations in the coding sequence of the 24 human CLDN genes contribute to the etiology of human NTDs. Using a target enrichment next-generation sequencing approach combined with Sanger sequencing, I identified 10 rare (MAF≤1% in the ExAC and EVS databases) and 4 novel missense mutations in nine CLDN genes in a cohort of 125 patients with open spinal NTDs: 3 in CLDN3, 1 in CLDN4, 1 in CLDN6, 1 in CLDN9, 1 in CLDN16, 1 in CLDN18, 2 in CLDN19, 1 in CLDN23 and 3 in CLDN24. Nine of these variants were predicted to be deleterious by in silico analysis. All patients were heterozygous for the missense mutations and most were private mutations with

157 the exception of CLDN3 p.A128T found in 3 patients and CLDN6 p.R209Q observed in 2 patients. Of these variants, only CLDN19 I22T and E209G affected neural tube closure when overexpressed in chick embryos. The NTDs caused by overexpressing CLDN19 E209G are due to defective convergent extension movements while defective fusion of the neural folds is observed in embryos overexpressing CLDN19 I22T. These novel findings suggest that rare and novel variants in CLDN genes contribute to increased susceptibility to human NTDs and that the phase of neural tube closure affected depends on the location of the variant within the claudin protein. Furthermore, site-directed mutagenesis of residues in the cytoplasmic C-terminal tail of chicken Cldn8 identified residues that are required for Cldn8 function in neural tube closure: S198 and S216. Overexpression studies in the chick embryo revealed that S216, but not S198, regulates convergent extrension suggesting that these residues regulate different morphogenetic events during neural tube. Together, these data suggest that individual claudins have unique roles in regulating intracellular signaling events that affect cell and tissue behaviors during neural tube closure.

158

5.2 GENERAL DISCUSSION Each claudin is unique with respect to which other claudin family members and PDZ domain-containing proteins it can interact with and these interactions have not been fully investigated. Here, I will discuss how claudin-based interactomes form signaling compartments that regulate morphogenesis of the neural and non-neural ectoderm during neural tube closure.

5.2.1 The role of claudins in the neural ectoderm during neural tube closure Genetic knockout of a single Cldn in mouse lines does not lead to neural tube defects (NTDs) and this is likely due to complex compensatory mechanisms (Anderson et al., 2008; Ben-Yosef et al., 2003; Fujita et al., 2012; Furuse et al., 2002; Gow et al., 1999; Kage et al., 2014; Li et al., 2014; Miyamoto et al., 2005; Morita et al., 1999; Tamura et al., 2008). To investigate the possibility that claudins have functionally redundant roles during neural tube closure in chick embryos, I used the C-terminal domain of Clostridium perfringens enterotoxin (C-CPE) as a claudin inhibitor. Cldn3, -4, -8 and -14 are the only claudins present during chick neural tube closure that interact with C-CPE. The action of C-CPE seems to be specific to interacting claudins: Cldn1 localization to tight junctions was not affected, while localization of Cldn3, -4 and -8 at tight junctions was disrupted (Baumholtz et al, 2017b). In addition, localization of the tight junction scaffolding protein ZO-1 appeared normal. Furthermore, C- CPEYL, which does not bind to claudins, did not affect claudin localization to tight junctions. These results confirm the suitability of C-CPE as a specific inhibitor of a subset of claudins. Thus, any phenotypes resulting from C-CPE treatment are the direct result of its interaction with claudins. C-CPE-treated chick embryos exhibit open NTDs caused by defective convergent extension and apical constriction of midline neural ectoderm cells (Baumholtz et al, 2017b). These morphogenetic movements are intrinsic to the neural ectoderm suggesting that C-CPE- sensitive claudins expressed in the neural ectoderm are required for neural tube closure. In chick, Cldn4 and -8 are the only claudins present in the neural ectoderm that interact with C- CPE. Treating chick embryos with C-CPELSID, which specifically removes Cldn4 from the neural and non-neural ectoderm, did not affect neural tube closure suggesting that Cldn8 plays a critical role in regulating morphogenesis of the neural ectoderm. However, Cldn8 is upregulated in C-CPELSID-treated chick embryos and may compensate for the loss of Cldn4. A C-CPE

159

variant that specifically interacts with Cldn8 is not available; therefore, mutagenesis studies were performed to further investigate the role of Cldn8 in morphogenesis of the neural ectoderm. Mutagenesis studies identified critical residues in the C-terminal domain of Cldn8 that regulate neural tube closure: S198 and S216. These data suggest that residues in the cytoplasmic C- terminal domain of Cldn8 interact with signaling proteins at the tight junction cytoplasmic plaque to regulate apical constriction and convergent extension during neural tube closure. It should be noted that Cldn17 is the only Cldn family member that has not been tested for its ability to bind to C-CPE (Lohrberg et al., 2009; Moriwaki et al., 2007; Sonoda et al., 1999). Cldn17 is enriched in the neural ectoderm during neural tube closure in chick embryos (Collins et al., 2013) and, therefore, it will be important to determine if Cldn17 interacts with C-CPE and whether loss of Cldn17 from tight junctions contributes to the NTDs observed in C-CPE-treated chick embryos.

5.2.1.1 Claudin function in convergent extension Our data suggest that claudins interact with components of the planar cell polarity (PCP) pathway to regulate convergent extension during neural tube closure. Convergent extension movements lead to anterior-posterior extension and medial-lateral narrowing of the neural plate. C-CPE-treated embryos are shorter and wider than GST-treated control embryos consistent with defects in convergent extension. Vangl2, a core PCP protein, showed reduced localization to the membrane of neural ectoderm cells after 5h of C-CPE-treatment. The role for PCP components in regulating convergent extension movements during neural tube closure is well-defined owing to the number of animal models of NTDs in which PCP signaling is disrupted (Curtin et al., 2003; Greene et al., 1998; Hamblet et al., 2002; Wallingford and Harland, 2002; Wang et al., 2006a; Wang et al., 2006b). Our data suggest that claudins are required to localize PCP proteins to the apical membrane.

Claudins are localized apical to components of the PCP pathway at the apical membrane of neural ectoderm cells (unpublished observation), therefore, the association between claudins and PCP proteins is likely to be indirect. Cytoplasmic scaffolding proteins that localize to tight junctions and the more basolateral adherens junctions are good candidates to mediate this interaction. Indeed, a proximity ligation assay showed that members of the PCP pathway, Dishevelled-1 and -2 and Vangl-1 and -2, are located near ZO-1 (Van Itallie et al., 2013).

160

Scribble, a PDZ-domain containing cytoplasmic protein that localizes to both tight and adherens junctions, interacts with PCP and tight junction proteins to regulate branching morphogenesis in the mouse lung: the PCP proteins Celsr1 and Vangl2 and the tight junction proteins ZO-2 and Cldn18 are mislocalized in the lungs of the Scribble mouse mutant Circletail (Yates et al., 2013). The localization of claudins in the neural ectoderm of the Circletail mouse mutant, which exhibits open NTDs, has not been examined. However, localization of Vangl2 is disrupted in the neuroepithelium of the Circletail mouse mutant (Kharfallah et al., 2017) suggesting that Scribble interactions are conserved between the mouse lung and neural tube. These data support the hypothesis that PCP signaling proteins and claudins interact via tight junction scaffolding proteins.

5.2.1.2 Claudin function in apical constriction Apical constriction of midline neural plate cells results in a change in cell shape from cuboidal to wedge-shaped and this cell shape change generates a physical force that contributes to bending of the neural plate (Burnside, 1971; Schoenwolf and Franks, 1984; Smith et al., 1994). Apical constriction is regulated, in part, by RhoA/ROCK-mediated actin-myosin contraction. RhoA accumulates in the apical region of the bending neural plate where it activates its effector Rho-associated kinase (ROCK) which, in turn, phosphorylates the regulatory light chain of myosin (MLC) (Kinoshita et al., 2008). This phosphorylation increases the ATPase activity of MLC, which moves along the apical actin filaments to generate a contractile force that causes apical constriction (Lee et al., 1983; Lee and Nagele, 1985; Rolo et al., 2009). The midline neural ectoderm cells of neural plate stage chick embryos cultured in the presence of C- CPE for 5 or 20 hours remain columnar and do not constrict at their apices. Our data suggest that claudins interact with RhoA-ROCK signaling to regulate actin-myosin-dependent apical constriction. RhoA, which shuttles between the cytoplasm and the membrane, showed reduced localization to the membrane and there was a significant reduction in phosphorylated MLC at the apical surface of neural ectoderm cells in C-CPE-treated embryos. The localization of Cldn4 and -8, the two C-CPE-sensitive claudins expressed in the neural ectoderm, was not affected in embryos treated with the ROCK inhibitors Y-27632 or GSK 269962. These data position Cldn4 and -8 upstream of RhoA-ROCK signaling in regulating apical constriction of neural ectoderm cells.

161

Our data suggest that Cldn4 and -8 are required for active RhoA/ROCK signaling at the apical membrane, which causes neural ectoderm cells to constrict at their apices. Claudins may be required to anchor RhoA to the membrane, where RhoA is activated by Rho guanine nucleotide exchange factors (RhoGEFs). Alternatively, claudins may recruit RhoGEFs to the membrane where they activate RhoA. These interactions could be direct or indirect: claudins may directly interact with RhoA or RhoGEFs through their cytoplasmic C-terminal tails or indirectly through tight junction scaffolding proteins. In support of the latter, a number of tight junction cytoplasmic proteins can bind to RhoA GEFs. Cingulin and paracingulin bind to and inactivate GEFH1 (Guillemot et al., 2008a). Cingulin recruits p114RhoGEF to tight junctions in epithelial cell lines where it activates RhoA to promote tight junction formation (Terry et al., 2011). A good candidate in the neural tube is PDZ-RhoGEF (ARHGEF11), which associates with tight junctions by binding to ZO-1 (Itoh et al., 2012). PDZ-RhoGEF localizes to the apical surface of neural ectoderm cells in chick embryos and knockdown of PDZ-RhoGEF causes NTDs due to defective apical constriction and convergent extension (Nishimura et al., 2012). Localization of ZO-1 to tight junctions is not affected in C-CPE-treated embryos, however, higher resolution techniques such as immuno-gold freeze-fracture electron microscopy and Structured Illumination Microscopy (SIM) may reveal subtle differences in ZO-1 localization that would be missed by standard confocal microscopy such as small breaks in ZO-1 tight junction strands or in the distribution of ZO-1 in apical versus more basal tight junction strands. Additionally, it is possible that changing the combination of claudin family members that interacts with ZO-1 could alter the affinity of ZO-1 for PDZ-RhoGEF. While there is evidence to support the hypothesis that claudins interact with RhoGEFs through scaffolding proteins at the tight junction cytoplasmic plaque, deleting the PDZ binding domain of Cldn8 did not cause NTDs. Furthermore, we observed the loss of RhoA from tight junctions in 50% of chick embryos overexpressing the Cldn8 S216I variant and this was associated with defective apical constriction. Thus, the possibility that the cytoplasmic C-terminal tail of Cldn8 directly interacts with RhoA or its effector RhoGEFs should be further explored.

Apical constriction is also regulated by microtubules. In neuroepithelial cells, microtubules are apically-basally aligned with their minus ends oriented towards the cell apex and the plus ends towards the cell base. Treating chick embryos with colchicine, which disrupts microtubules, causes defects in apical constriction and rounding of the apical surface of

162

neuroepithelial cells (Fernandez et al., 1987; Karfunkel, 1972) suggesting that microtubules have a yet undefined role in apical constriction of neural ectoderm cells. Microtubules are known to play an important role in protein trafficking during morphogenesis (Jesuthasan and Stahle, 1997; Rodriguez-Boulan et al., 2005). In Xenopus embryos, microtubules are required for apical constriction of bottle cells and endocytosis acts downstream of actin-myosin contraction to remove excess membrane from the apical surface; inhibiting microtubules using the drug nocodazole or endocytosis leads to defective apical constriction (Lee and Harland, 2007; Lee and Harland, 2010). Thus, it is possible that microtubules function to transport excess apical membrane endocytosed during apical constriction away from the apical surface. Another possibility is that microtubules transport proteins required for apical constriction to the apical surface. In C-CPE-treated embryos, microtubules appear disorganized, shorter and do not extend to the apical surface of neural ectoderm cells compared to control embryos. Furthermore, the apical surface of neural ectoderm cells in C-CPE-treated embryos is round, smooth and devoid of microvilli, finger-like projections of the plasma membrane with a microtubule-based core, compared to GST-treated control embryos. These data suggest that C-CPE-sensitive claudins expressed in the neural ectoderm interact with the microtubule cytoskeleton to regulate apical constriction and the shape of the apical of surface of neuroepithelial cells.

The mechanism by which claudins affect the organization of microtubules is not clear. However, a recent study conducted showed that the tubulin β-2A chain is located near the N- terminus of Cldn4 in MDCK cells (Fredriksson et al., 2015). Tubulin is the major constituent of microtubules. Therefore, C-CPE-sensitive claudins expressed in neuroepithelial cells may tether microtubules to the apical surface of neural ectoderm cells to stabilize their minus ends and promote microtubule polymerization. Another possibility is that the effects on the microtubule network are caused by mislocalization of the RhoGTPases RhoA and Cdc42 in neuroepithelial cells of C-CPE-treated embryos. Cdc42 and RhoA are critical regulators of the microtubule cytoskeleton (Cook et al., 1998; Palazzo et al., 2001a; Palazzo et al., 2001b). Cdc42 plays a key role in orienting the microtubule organizing center in fibroblasts (Palazzo 2001) (Palazzo et al., 2001b), while RhoA acts to stabilize microtubules (Cook; Palazzo 2001). Thus, claudins may be required to anchor RhoGTPases to the apical membrane of neural ectoderm cells where the RhoGTPases ensure that microtubule bundles are apical-basally aligned.

163

Our data indicate that C-CPE removal of claudins affects the localization of RhoA and Vangl2, proteins that regulate apical constriction and convergent extension, respectively. Both of these events are intrinsic to the neural ectoderm, where Cldn4 and -8 are the only C-CPE- sensitive claudins. Mutagenesis studies revealed that the S216 residue in the C-terminal tail of Cldn8 is critical for Cldn8 function in apical constriction and convergent extension and that this residue is required to anchor RhoA to the apical membrane of neural ectoderm cells. Future studies are needed to determine if S216 interacts with components of the PCP pathway to regulate convergent extension during neural tube closure. These data suggest that the cytoplasmic C-terminal domain of Cldn8 interacts with signaling proteins at the tight junction cytoplasmic plaque to regulate cell and tissue behaviors during neural tube closure.

5.2.2 The role of Cldn3 in the non-neural ectoderm during neural tube closure

5.2.2.1 Cldn3 functions in epithelial fusion Removing only Cldn3 from the non-neural ectoderm of chick embryos using C-CPELDR prevents fusion of the dorsal tips of the neural folds to form a closed neural tube. Scanning electron microscopy revealed a meshwork of fibrils bridging the apposed neural folds in control GST-treated embryos that is absent in Cldn3-depleted embryos. Identified in the 1970s, this meshwork was proposed to act as a temporary adhesive (Bancroft and Bellairs, 1975; Lee et al., 1983; Lee and Nagele, 1985). Lanthanum staining, which is used to identify cell surface glycoproteins, its ability to bind to concanavalin A, which binds to cell surface carbohydrate residues, and its removal by trypsin suggest that this meshwork is a mix of two types of extracellular matrix proteins: proteoglycans and glycoproteins. Proteoglycans consist of a core protein bearing covalently linked glycosaminoglycan chains (Poulain and Yost, 2015). In embryonic tissues, the glycosaminoglycan chains are usually heparan or chondroitin sulfates. Glycoproteins are proteins that contain carbohydrate groups attached to polypeptide side chains and include laminin and fibronectin. I hypothesize that this meshwork is functionally equivalent to the cellular protrusions observed in the mouse embryo in mediating the initial contact between the apposed neural folds. A clear link between extracellular matrix proteins and tight junction proteins has not been established and it is unclear how claudins would regulate the secretion of extracellular matrix proteins into the extracellular space. However, a proximity ligation assay in MDCK cells revealed that the heparan sulfate proteoglycan syndecan-1 and the receptors for

164

extracellular matrix proteins, integrin α2 and β1, localize near the N-terminus of Cldn4 (Fredriksson et al., 2015) but not near the N- or C-terminus of ZO-1 (Van Itallie et al., 2013). Therefore, it is possible that Cldn3 promotes the translocation of extracellular matrix proteins and their receptors to the apical surface of non-neural ectoderm cells where extracellular matrix proteins are secreted into the extracellular space between the apposed neural folds. In support of this hypothesis, Cldn3 is expressed at the apical membrane of ameloblasts in the developing teeth of mice where it is required for secretion of extracellular matrix proteins (Bardet et al., 2017). Once the initial contact is made between the neural folds, the attachment must be stabilized by direct protein-protein interactions and junctional complexes must be remodeled between newly adjacent cells. In mice, direct interactions between Eph receptors and their ephrin ligands stabilize the adhesion between the apposed neural folds (Abdul-Aziz et al., 2009; Pai et al., 2012; Ray and Niswander, 2012). Claudins may also be involved in this step of neural fold fusion as claudin interactions with Eph receptors and ephrin ligands have previously been reported. In epithelial cells, Cldn1 and -4 interact with ephrinB1 via their first extracellular loops and this association leads to outside-in signaling resulting in tyrosine phosphorylation of the cytoplasmic tail of ephrin-B1 and reduced cell-cell adhesion (Tanaka et al., 2005b). EphA2 and Cldn4 associate via their extracellular loops and this association leads to reverse signaling through Cldn4 resulting in phosphorylation of the C-terminal domain of Cldn4 at tyrosine 208 (Tanaka et al., 2005a). Tyrosine phosphorylation of Cldn4 disrupts the interaction between Cldn4 and ZO-1 resulting in reduced Cldn4 localization to tight junctions. Signaling between Cldn4 and EphA2 is bidirectional as Cldn4 and EphA2 association inhibits EphA2 phosphorylation. Furthermore, the interaction between Cldn4 and EphA2 is required for EphA2 to localize to the cell membrane. Therefore, it is possible that Cldn3 interacts with Eph receptors and ephrin ligands to stabilize the adhesion between the apposed neural folds and facilitate the remodeling of tight junctions that occurs during neural fold fusion. Cldn3 null mice do not exhibit NTDs (Ahmad et al., 2017; Kooij et al., 2014), thus I hypothesize that other claudins compensate for the loss of Cldn3 in the non-neural ectoderm of mouse embryos. Good candidates for other claudins that might fulfill this role are Cldn6 and -7, which are not present in the chick genome. Cldn3, -4, -6 and -7 are downstream targets of grainyhead-like 2 (Grhl2), a transcription factor that is expressed in the non-neural ectoderm of

165

mice throughout neural tube closure and is required for neural fold fusion (Pyrgaki et al., 2011; Rifat et al., 2010; Senga et al., 2012; Werth et al., 2010).

5.2.3 The role of CLDN rare and novel missense variants in the etiology of human NTDs

5.2.3.1 Rare mutations in CLDN19 confer increased susceptibility to NTDs Sequencing analysis of the coding region of CLDN1-25 in 125 NTD patients identified 10 rare and 4 novel missense variants that were present in the heterozygous state. To identify deleterious mutations, plasmids encoding human wild-type or variant Cldns were overexpressed in chick embryos. Overexpression of wild-type CLDN6, 16, 18, 19 and 24 did not cause NTDs suggesting that any phenotypes observed are due to the mutations themselves and not a consequence of overexpressing a claudin. I found that CLDN19 I22T and E209G do not affect the ability of Cldn19 to localize to the tight junction but cause open NTDs when overexpressed in chick embryos. Convergent extension and apical constriction are not affected in embryos overexpressing CLDN19 I22T suggesting that defective fusion of the neural folds is responsible for the NTDs observed in these embryos. I22 is located in the first transmembrane domain of Cldn19. Replacement of isoleucine with threonine introduces a putative phosphorylate site that could cause a conformational change in Cldn19. Thus, CLDN19 I22T may affect the ability of Cldn19 to participate in heteromeric cis interactions with other family members within tight junction strands. Embryos overexpressing CLDN19 E209G have a shortened anterior-posterior axis suggesting that defective convergent extension movements are responsible for the NTDs observed in these embryos. E209 is located in the C-terminal cytoplasmic tail of the CLDN19 protein immediately adjacent to the PDZ-binding domain. The glutamic acid to glycine substitution results in loss of a negatively charged amino acid that could cause a conformation change in the adjacent PDZ-binding domain, thus disrupting the interaction of Cldn19 with scaffolding proteins at the tight junction cytoplasmic plaque that link Cldn19 to PCP proteins. Convergent extension movements and neural fold fusion are morphogenetic events that occur in both the neural and non-neural ectoderm. The mRNA and protein expression patterns of Cldn19 during neural tube closure in chick and mouse embryos have not been assessed. It will be important to characterize the expression pattern of Cldn19 during neurulation to determine if

166

the NTDs observed in embryos overexpressing the CLDN19 variants are caused by defective morphogenesis of the neural or non-neural ectoderm or both. I found that overexpressing CLDN19 I22T and E209G in chick embryos causes NTDs. However, Cldn19-depleted mice do not exhibit NTDs (Hou et al., 2009). I hypothesize that this is due to a difference in the experimental approach. Knockdown of claudins often results in compensatory upregulation of other family members. For example, compensatory upregulation of Cldn3 is observed in the kidneys of Cldn4 null mice (Fujita et al., 2012). I, therefore, hypothesize that Cldn19 knockdown mice do not exhibit NTDs due to compensatory upregulation of other family members in the neural tube. I hypothesize that the CLDN19 I22T and E209G variants are able to incorporate into tight junction strands but are unable to interact with other claudin family members (I22T) or with proteins at the cytoplasmic plaque (E209G) required for regulating neural tube morphogenesis. Compensatory upregulation of other claudin family members should not occur because CLDN19 I22T and E209G correctly localize to the tight junction. Thus, overexpression of variant claudins mimics what is seen in a heterozygous patient where both wild-type and variant CLDN genes are present. Overexpressing CLDN19 I22T and E209G causes a 23% and 48% increase in the incidence of NTDs in chick embryos, respectively, suggesting that these mutations represent low penetrant variants that interact with other genes and/or environmental factors to modulate the incidence and severity of NTDs. Mutations in CLDN19 are associated with FHHNC, an autosomal recessive kidney disorder characterized by excessive loss of Ca2+ and Mg2+. Patients with these mutations do not exhibit NTDs. Examination of the location of the CLDN19 mutations observed in these patients revealed that the mutations are concentrated in the first extracellular loop (Al-Shibli et al., 2013; Claverie-Martin et al., 2013; Sharma et al., 2016; Yuan et al., 2015) and third and fourth transmembrane domains (Claverie-Martin et al., 2013; Ekinci et al., 2012; Naeem et al., 2011). Charged residues in the first extracellular loop of claudins determine the ion selectivity of the tight junction, therefore, mutations in the first extracellular could disrupt the ability of CLDN19 to form cation selective channels in the kidney leading to FHHNC. Cldn16 and -19 form heterodimers through cis interactions of the third and fourth transmembrane domains and these interactions are required for their assembly into tight junction strands (Gong et al., 2015b). Thus, mutations in the third and fourth transmembrane domains could disrupt the heteromeric interactions between Cldn16 and -19 that are required to form cation selective channels leading

167

to FHHNC. The CLDN19 mutations observed in NTD patients are located in the first transmembrane domain and the C-terminal cytoplasmic tail and, therefore, should not affect the ability of Cldn19 to form a cation selective channel or disrupt heteromeric cis interactions between Cldn19 and -16. Thus, I propose a model where the location of the mutation within the Cldn19 protein affects the phenotype. In support of this hypothesis, the NTD patients with mutations in CLDN19 do not have kidney defects. Futhermore, the different Cldn19 variants caused different types of NTDs: overexpressing the C-terminal CLDN19 E209G variant caused convergent extension defects, while only neural fold fusion was affected in embryos overexpressing the CLDN19 I22T variant in the first transmembrane domain.

5.2.3.2 CLDN-gene interactions NTDs are a complex disease involving gene-gene and gene-environment interactions. In humans, NTDs are mostly sporadic and putative mutations reported to date are predominantly heterozygous, inherited from an unaffected parent (Allache et al., 2012; Bosoi et al., 2011; De Marco et al., 2013; De Marco et al., 2012; Kibar et al., 2009; Kibar et al., 2011; Lei et al., 2010; Robinson et al., 2012). Therefore, it has been suggested that human NTDs exhibit a polygenic pattern of inheritance involving multiple heterozygous gene mutations. Such genetic interaction is observed in NTD mouse models where doubly heterozygous mice exhibit an increased incidence and/or more severe NTDs than mice heterozygous for a single gene. For example, Vangl2Lp genetically interacts with other PCP genes including Vangl1, Dvl1, Dvl2, Celsr1, Fzd3 and Fzd6 (Hamblet et al., 2002; Curtin et al. 2003; Wang et al., 2006a,b; Torban et al., 2008) and mutations in genes not previously implicated in the PCP pathway such as Grhl3ct, Bardet-Biedle syndrome 1 (BBS1), BBS4, BBS6, and Sec24b (Merte et al., 2010; Ross et al., 2005; Stiefel et al., 2003) to cause NTDs. In the present study, rare and novel missense variants were identified in CLDN3, 4, 6, 9, 16, 18, 19, 23 and 24 in a cohort of 125 patients. Of the variants tested so far, only CLDN19 p.I22T and p.E209G caused NTDs in the chick overexpression assay. Despite the fact that CLDN16 p.N223S, CLDN18 p.V88I and CLDN24 p.G94R were predicted to be deleterious by in silico analysis, overexpressing these claudin variants did not cause NTDs in chick embryos. These results indicate that alone, these mutations do not cause NTDs. However, given the polygenic mode of inheritance in human NTDs, we cannot rule out the possibility that these

168

CLDN mutations interact with a non-CLDN genetic variant or an environment factor to cause a NTD. In fact, the patient with the CLDN16 p.N223S variant also has a common variant in the PCP gene DVL2. The patients with the CLDN18 p.V88I and CLDN24 p.G94R variants do not harbor a mutation in a PCP gene, but the genomic DNA of these patients has not been sequenced for mutations in other genes implicated in NTDs including those that interact with claudins to regulate neural tube closure (e.g. RHOA, CDC42, PAR3).

5.2.3.3 CLDN-environment interactions and folic acid Periconceptional maternal folic acid supplementation and fortification reduces the incidence of NTDs in humans from 50-70% (Atta et al., 2016; De Wals et al., 2007; Osterhues et al., 2013). To test the folate responsiveness of claudin-depleted chick embryos, the culture medium of C-CPE or C-CPELDR-treated chick embryos was supplemented with 100µM or 10mM of folic acid. Exposure to folic acid did not reduce the incidence of NTDs, which were of similar severity and position along the anterior-posterior axis as chick embryos cultured without folic acid. These data suggest that NTDs caused by claudin depletion are models of folate- nonresponsive NTDs. The concentrations of folic acid used are in the range of previous studies showing the effectiveness of folic acid in preventing NTDs in chick embryos (Guney et al, 2003; Weil et al., 2004). Previous studies testing folate responsiveness in chick embryos used embryos in which the neural folds were elevated and began fusing (HH8-10), whereas my studies began folic acid supplementation at the neural plate stage (HH4). Thus, my studies involved exposure to folic acid throughout neurulation, which is more representative of the length of exposure to folate experienced by human embryos in utero. The observation that NTDs caused by claudin- depletion are folate-nonresponsive is consistent with a previous study which showed that folic acid supplementation does not prevent NTDs in Grhl2 mutant mice, which exhibit reduced expression of Cldn4, -6 and -7 (Marean et al., 2001). Together, these data suggest that CLDN mutations may contribute to a subset of the ~30% of human NTDs that cannot be prevented by folic acid supplementation. In order to fully investigate the relationship between folate status and CLDN mutations in susceptibility to NTDs, the effect of folate deficiency on the incidence and severity of NTDs in claudin-depleted embryos should also be investigated. While it is not possible to create a folate- depleted culture medium for chick embryos, these experiments could be performed in Cldn

169

mutant mouse lines. Although suboptimal maternal folate levels may not be sufficient to cause a NTD, they may contribute to a NTD in individuals that are genetically susceptible such as those with a CLDN mutation. Such a gene-environment interaction has been observed in mice, where folate deficiency alone does not cause a NTD unless present in combination with a mutation in a predisposing gene (Burgoon et al., 2002; Burren et al., 2008; Heid et al., 1992).

5.2.4 Individual claudins have unique roles in regulating intracellular signaling events that affect epithelial morphogenesis Overexpressing Cldn8 S216I in chick embryos caused NTDs due to defective apical constriction and convergent extension of the neural ectoderm. While CLDN19 E209G caused NTDs associated with convergent extension defects, apical constriction was not affected. Removal of Cldn3 from the non-neural ectoderm caused NTDs due to defective fusion of the neural folds. Together, these data support the hypothesis that individual claudins have unique roles in regulating intracellular signaling events that affect epithelial morphogenesis. While the expression pattern of Cldn19 during neural tube closure is unknown, Cldn8 is exclusively expressed in the neural ectoderm and Cldn3 is restricted to the non-neural ectoderm. Therefore, my studies support a model where the combination of claudins expressed in an epithelial tissue creates signaling compartments that uniquely regulate morphogenesis of that tissue. Furthermore, my studies highlight the importance of the C-terminal domains of claudins in regulating cell and tissue behaviors as both S216I and E209G are located in the C-terminus of Cldn8 and CLDN19, respectively. My work has provided further insight into our understanding of how different combinations of claudins interact with signaling complexes at the tight junction cytoplasmic plaque to affect cellular behaviors and tissue morphogenesis.

5.3 FUTURE DIRECTIONS My data suggest that Cldn8 in the neural ectoderm and Cldn3 in the non-neural ectoderm are required to regulate neural tube morphogenesis. However, a complete Cldn8 knockout mouse has not been generated. Furthermore, the mechansisms by which Cldn3 and 8 regulate neural tube morphogenesis have not been fully elucidated. I have also shown that rare deleterious mutations in CLDN19 contribute to increased susceptibility to human NTDs. However, not all CLDN variants that were predicted to be deleterious by in silico analysis caused NTDs when overexpressed in the chick embryo. NTDs are a multigenic disorder and the possibility that NTD

170

patients harbor additional non-CLDN genetic variants that interact with CLDN variants to generate NTDs still exists. These outstanding questions would be addressed by the experiments proposed below.

5.3.1 Defining the role of Cldn8 in neural tube closure Removing Cldn4 and -8 from the neural ectoderm using C-CPE causes NTDs due to defective apical constriction and convergent extension movements in chick embryos. Site- directed mutagenesis studies showed that the S198 and S216 residues in the C-terminal tail of chicken Cldn8, but not the PDZ binding domain, are critical for Cldn8 to regulate neural tube closure. The role of Cldn8 in neural tube closure could be futher explored in the experiments described below.

5.3.1.1 Determining if Cldn8 mutant mice develop NTDs Conditional deletion of the Cldn8 locus in the kidney causes salt wasting defects and localization of Cldn4 to tight junctions is reduced in these mice (Gong et al., 2015b). A complete knockout mouse for Cldn8 has not been generated, therefore, we are collaborating with Dr. Hou’s research group (Washington University, St. Louis, MO), who has generated the kidney-specific Cldn8 knockout mice to generate a global Cldn8 knockout mouse. Embryos will be collected from E8 until E10 and analyzed for defects in neural tube closure. I hypothesize that Cldn8-/- mice will have an embryonic lethal NTD phenotype, particularly if the absence of Cldn8 in the neural tube disrupts Cldn4 localization to tight junctions as was observed in the kidney. In addition to genetic risk factors, NTDs are caused by environmental agents. Therefore, if Cldn8-/- mice do not develop NTDs, pregnant dams could be subjected to environmental agents that are known to cause NTDs such as high temperatures, high sugar, and low folate diets to determine if environmental factors increase the incidence of NTDs in Cldn8 mutant mice. Overexpression of the S198A and S216I variants caused NTDs in chick embryos. To determine if these variants play a conserved role in neural tube closure in mouse embryos, the CRISPR/Cas9 gene editing technology could be used to generate the corresponding point mutations in mouse embryos which could then be analyzed for NTDs as described above.

171

5.3.1.2 Determining if Cldn8 genetically interacts with PCP proteins to regulate neural tube closure The core PCP protein Vangl2 shows reduced localization to the membrane in C-CPE- treated chick embryos suggesting that claudins intersect with the PCP pathway to regulate neural tube closure. To determine if Cldn8 genetically interacts with PCP proteins, Cldn8+/- mice could be crossed to heterozygous PCP mouse mutants to generate double heterozygotes. I hypothesize that Cldn8 null mice, like mice homozygous for mutations in PCP genes, will develop embryonic lethal NTDs, therefore, these experiments should be performed using heterozygous mice. PCP mouse mutants will be prioritized based on the interaction partners identified in experiments described below. This approach has been used to show that Vangl2 genetically interacts with Syndecan 4 (Escobedo et al., 2013) and that the PCP genes Vangl2, Scribble and Celsr1 genetically interact with each other in mice to regulate neural tube closure (Murdoch et al., 2014).

5.3.1.3 Determining how Cldn8 regulates epithelial morphogenesis during neural tube closure I showed that overexpression of Cldn8 S216I and S198A, but not S216A and ΔYV, cause NTDs. Furthermore, Cldn8 S216I causes apical constriction and convergent extension defects and RhoA is mislocalized at the apical surface of neural ectoderm of embryos overexpressing S216I. These data suggest that residues in the C-terminus of Cldn8 interact with signaling proteins at the tight junction cytoplasmic plaque to regulate morphogenetic events during neural tube closure. To further define how these residues interact with signaling proteins at the tight junction cytoplasmic plaque, GST-pull down experiments could be performed using GST fused to the C-terminal tail of wild-type Cldn8 and Cldn8 S198A, S216A, S216I, and ΔYV and protein lysate from HH8 neural fold stage chick embryos. Mass spectrometry analysis could be used to identify proteins that interact with wild-type Cldn8, S216A and ΔYV, but not S198A and S216I as these proteins are critical for Cldn8 function in neural tube closure. Alternatively, a proximity ligation assay could be conducted using biotin ligase fused to the C-terminus of the Cldn8 variants to biotinylate proximal proteins in cultured MDCK II epithelial cells as has previously been done for Cldn4 (Fredriksson et al., 2015). Biotinylated proteins should then be purified on streptavidin resin and identified by mass spectrometry.

172

5.3.2 Characterization of Cldn3-dependent epithelial remodeling events required for neural fold fusion

5.3.2.1 Characterization of the meshwork that links the apposed neural folds Scanning electron microscopy revealed that a meshwork of fibrils connects the apposed neural folds of control GST-treated embryos and that this meshwork is absent in Cldn3-depleted embryos. This mesh-like network bridging the apposed neural folds was first described in the 1970s and was shown to consist of glycoproteins and proteoglycans (Bancroft and Bellairs, 1975; Lee et al., 1976; Lee et al., 1978). Concavalin A staining of glycoproteins will be used to determine if the distribution of glycoproteins differs in control and Cldn3-depleted embryos (Inagaki et al., 1996; Lee et al., 1978). As an alternative approach, periodic acid Schiff and alcian blue staining could be done to confirm that the meshwork consists of glycoproteins and to determine if the staining pattern is different in Cldn3-depleted embryos. Embryos should be analyzed at HH9-10, which corresponds to the developmental stages when the cranial and spinal neural folds are apposed and when I observed the meshwork by scanning electron microscopy. Wild-type embryos could be cultured with and without addition of chlorate, heparitinase or chondroitinase, enzymes to cleave heparan/chondroitin sulfate side chains, and analyzed for defects in neural tube closure (Escobedo et al., 2013; Yip et al., 2002). If NTDs are observed in treated embryos, scanning electron microscopy will be performed to determine if the meshwork is absent in treated embryos. Absence of the meshwork would indicate that proteoglycans are a major component of the meshwork. Embryos will be treated at HH8, once convergent extension is underway, as inhibiting proteoglycan sulfation causes NTDs due to defective convergent extension in cultured mouse embryos (Escobedo, 2013). Cldn3-depleted embryos will be cultured in the presence of exogenous sulfate to determine if this rescues the NTDs observed in these embryos.

5.3.2.2 Defining the interaction between Cldn3 and Eph receptors and ephrin ligands during neural fold adhesion Following the initial contact between the apposed neural folds, the attachment is stabilized by direct interactions between Eph receptors and their ephrin ligands. Claudin interactions with Eph receptors and their ephrin ligands have been observed in epithelial cell lines. I hypothesize that Cldn3 will interact with Eph receptors and/or ephrin ligands in the

173

dorsal tips of the neural folds to stabilize their adhesion. Analysis of EphA2, which is expressed in the dorsal tips of the neural folds of mouse embryos (Abdul-Aziz et al., 2009) and interacts with Cldn4 should be prioritized (Tanaka et al., 2005a). The expression pattern of EphA2 has not previously been reported in chick embryos, therefore, in situ hybridization could be performed to examine the mRNA expression pattern of EphA2 during neurulation in chick embryos. If EphA2 is expressed in the dorsal tips of the neural folds of chick embryos, immunofluorescence could be performed to determine if it is mislocalized in Cldn3-depleted chick embryos. If changes in its localization or expression are observed, co-immunoprecipitation assays could be done to determine if EphA2 interacts with Cldn3. If Cldn3 interacts with EphA2, it could be determined if this interaction affects phosphorylation of the Eph or Cldn3. If EphA2 is not mislocalized in Cldn3-depleted chick embryos, the localization of other Eph receptors and ephrin ligands that are expressed during the fusion phase in chick embryos and are required for neural fold fusion in mouse embryos such as ephrinA5 and EphA1, A3 and A7 should be examined (Abdul-Aziz et al., 2009; Baker and Antin, 2003; Holmberg et al., 2000).

5.3.3 Determining the contribution of CLDN mutations to the etiology of human NTDs

5.3.3.1 Identifying how CLDN19 mutations cause NTDs Chick embryos overexpressing CLDN19 E209G, but not wild-type CLDN19 or CLDN19 I22T, have a shortened anterior-posterior axis suggesting that defective convergent extension movements, which are dependent on the planar cell polarity pathway, are responsible for the NTDs. To determine if disrupted planar cell polarity signaling is responsible for the NTDs observed in embryos overexpressing CLDN19 E209G, immunofluorescence could be performed on these embryos to look at localization of the core PCP proteins Vangl, Dishevelled, Celsr, Frizzled, Prickle, and Diversin and compared to embryos overexpressing wild-type CLDN19 and CLDN19 I22T. Apical constriction and convergent extension were not affected in embryos overexpressing CLDN19 I22T suggesting that defective fusion of the neural folds is responsible for the NTDs. Scanning electron microscopy could be performed on embryos overexpressing wild-type CLDN19 and CLDN19 I22T to determine if formation of the meshwork that connects the apposed neural folds is affected in embryos overexpressing CLDN19 I22T indicating that the initial contact between the neural folds is disrupted. If an effect is observed, rescue experiments

174

could be performed on embryos overexpressing CLDN19 I22T using exogenous sulfate. If meshwork formation is not disrupted, the possibility that CLDN19 interacts with Eph receptors and ephrin ligands to regulate neural fold fusion and that overexpressing the CLDN19 I22T variant disrupt these interactions could be explored. Immunofluorescence could be performed to determine if localization of Eph receptors and ephrin ligands is dirsrupted in embryos overexpressing CLDN19 I22T compared to wild-type CLDN19. Eph receptors and ephrin ligands that are expressed during the fusion phase in chick embryos and are required for neural fold fusion in mouse embryos such as ephrinA5 and EphA1, A2, A3 and A7 (Abduz-Aziz, 2009; Holmberg, 2000; Baker, 2003) should be prioritized. If no effect is observed, GST-pull down experiments could be performed using GST fused to the N-terminal tail of CLDN19 and CLDN19 I22T and protein lysate from HH9-10 chick embryos, the developmental stages when the neural folds are apposed. Mass spectrometry could be performed to identify proteins that interact with wild-type Cldn19 but not Cldn19 I22T.

5.3.3.2 Determining if CLDN variants interact with other genes to increase susceptibility to NTDs Overexpression of wild-type CLDN6, 16 and 24 and CLDN6 R209G, CLDN16 N223S, and CLDN24 G94R and L177M did not cause NTDs. However, NTDs are a complex disorder involving gene-gene interactions. Thus, humans with CLDN mutations may also harbor additional non-CLDN genetic variants that interact with CLDN mutations to generate NTDs. In fact, one patient with the CLDN6 p.R209Q variant also has a rare synonymous variant p.C145C in the core PCP protein VANGL1 and the patient with the CLDN16 p.N223S variant has a common variant p.T535I in the core PCP gene DVL2. To determine if these PCP variants interact with the CLDN variants to confer increased risk to NTDs, chick embryos can be electroporated with both the CLDN and PCP variants and assessed for NTDs. The patients with the CLDN24 p.G94R and p.L177M variants did not have mutations in the core PCP genes, therefore, these patients should be sequenced for mutations in genes associated with NTDs such as PAR3, SHROOM3, GRHL3, and genes in the folate pathway.

175

5.4 CONCLUDING SUMMARY I have shown that claudins interact with signaling and polarity complex proteins at the apical membrane of epithelial cells to regulate morphogenetic movements and cell shape changes required for neural tube closure: Cldn8 collaborates with Cldn4 to regulate convergent extension movements and apical constriction of neural ectoderm cells, while Cldn3 in the non-neural ectoderm regulates fusion of the neural folds. My data suggest that the combination of claudins expressed in an epitheial cell layer creates functional compartments that influence cell and tissue behaviors.

176

5.5 ORIGINAL CONTRIBUTIONS The following are original contributions that I have made to understanding the role of claudins during neural tube closure.

Chapter II: In this chapter, I describe a role for claudins in regulating neural tube closure. I showed that C-CPE removal of Cldn4 and -8 from the neural ectoderm and Cldn3 and -4 from the non-neural ectoderm causes dose-dependent folate resistant neural tube defects (NTDs) in chick embryos. I showed that Cldn4 and -8 act upstream of planar cell polarity and RhoA- ROCK signaling to regulate convergent extension movements and apical constriction of the neural ectoderm, respectively, during neural tube closure. Additionally, I showed that Cldn4 and -8 are required for the correct localization of Par3 and the RhoGTPases RhoA and Cdc42 at the apical surface of the neural ectoderm. This work has shown for the first time that claudins play a direct role in neural tube closure and discusses the pathways that claudins intersect with to regulate neural tube morphogenesis.

Chapter III: In this chapter, I showed that Cldn3 is required in the non-neural ectoderm for fusion of the neural folds to form a closed neural tube. My data suggests that Cldn3-depleted embryos are lacking the extracellular matrix-like meshwork emanating from the tips of the apposed neural folds to mediate their initial contact. These data suggest that Cldn3 is required to traffic or secrete extracellular matrix proteins at the apical surface of non-neural ectoderm cells during neural tube closure.

Chapter IV: From the data presented in Chapters II and III, I observed that claudins are required for neural tube closure. I sequenced a cohort of 125 patients with open spinal NTDs to identify mutations in the coding sequence of the 24 human CLDN genes and found ten rare and four novel missense mutations, nine of which were predicted to be deleterious by in silico analysis using Polyphen, SIFT, and MutationTaster. Using an overexpression assay, I showed that CLDN19 I22T and E209G but not wild-type CLDN19 caused NTDs in chick embryos confirming that these mutations are deleterious. This is the first report of CLDN mutations in a cohort of patients with NTDs. These data suggest that CLDN mutations contribute to increased

177

susceptibility to human NTDs. Furthermore, I showed that residues in the C-terminal cytoplasmic tail of Cldn8 mediate the function of Cldn8 in neural tube closure.

178

REFERENCES . Abdul-Aziz, N. M., Turmaine, M., Greene, N. D. and Copp, A. J. (2009). EphrinA-EphA receptor interactions in mouse spinal neurulation: implications for neural fold fusion. Intl J Dev Biol 53, 559-568. Adachi, M., Hamazaki, Y., Kobayashi, Y., Itoh, M., Tsukita, S., Furuse, M. and Tsukita, S. (2009). Similar and distinct properties of MUPP1 and Patj, two homologous PDZ domain-containing tight-junction proteins. Mol Cell Biol 29, 2372-2389. Adachi, M., Inoko, A., Hata, M., Furuse, K., Umeda, K., Itoh, M. and Tsukita, S. (2006). Normal establishment of epithelial tight junctions in mice and cultured cells lacking expression of ZO-3, a tight-junction MAGUK protein. Mol Cell Biol 26, 9003-9015. Agopian, A. J., Bhalla, A. D., Boerwinkle, E., Finnell, R. H., Grove, M. L., Hixson, J. E., Shimmin, L. C., Sewda, A., Stuart, C., Zhong, Y., et al. (2013). Exon sequencing of PAX3 and T (brachyury) in cases with spina bifida. Birth Defects Res A Clin Mol Teratol 97, 597-601. Ahmad, R., Kumar, B., Chen, Z., Chen, X., Muller, D., Lele, S. M., Washington, M. K., Batra, S. K., Dhawan, P. and Singh, A. B. (2017). Loss of claudin-3 expression induces IL6/gp130/Stat3 signaling to promote colon cancer malignancy by hyperactivating Wnt/beta-catenin signaling. Oncogene. Aijaz, S., D'Atri, F., Citi, S., Balda, M. S. and Matter, K. (2005). Binding of GEF-H1 to the tight junction-associated adaptor cingulin results in inhibition of Rho signaling and G1/S phase transition. Dev Cell 8, 777-786. Al-Haggar, M., Bakr, A., Tajima, T., Fujieda, K., Hammad, A., Soliman, O., Darwish, A., Al-Said, A., Yahia, S. and Abdel-Hady, D. (2009). Familial hypomagnesemia with hypercalciuria and nephrocalcinosis: unusual clinical associations and novel claudin16 mutation in an Egyptian family. Clin Exp Nephrol 13, 288-294. Al-Shibli, A., Konrad, M., Altay, W., Al Masri, O., Al-Gazali, L. and Al Attrach, I. (2013). Familial hypomagnesemia with hypercalciuria and nephrocalcinosis (FHHNC): report of three cases with a novel mutation in CLDN19 gene. Saudi J Kidney Dis Transpl 24, 338- 344. Alberini, G., Benfenati, F. and Maragliano, L. (2017). A refined model of claudin-15 tight junction paracellular architecture by molecular dynamics simulations. PloS one 12, e0184190. Allache, R., De Marco, P., Merello, E., Capra, V. and Kibar, Z. (2012). Role of the planar cell polarity gene CELSR1 in neural tube defects and caudal agenesis. Birth Defects Res A Clin Mol Teratol 94, 176-181. Anderson, W. J., Zhou, Q., Alcalde, V., Kaneko, O. F., Blank, L. J., Sherwood, R. I., Guseh, J. S., Rajagopal, J. and Melton, D. A. (2008). Genetic targeting of the endoderm with claudin-6CreER. Dev Dyn 237, 504-512. Angelow, S., Kim, K. J. and Yu, A. S. (2006). Claudin-8 modulates paracellular permeability to acidic and basic ions in MDCK II cells. J Physiol 571, 15-26. Arrate, M. P., Rodriguez, J. M., Tran, T. M., Brock, T. A. and Cunningham, S. A. (2001). Cloning of human junctional adhesion molecule 3 (JAM3) and its identification as the JAM2 counter-receptor. J Biol Chem 276, 45826-45832.

179

Atta, C. A., Fiest, K. M., Frolkis, A. D., Jette, N., Pringsheim, T., St Germaine-Smith, C., Rajapakse, T., Kaplan, G. G. and Metcalfe, A. (2016). Global Birth Prevalence of Spina Bifida by Folic Acid Fortification Status: A Systematic Review and Meta-Analysis. Am J Public Health 106, e24-34. Aue, A., Hinze, C., Walentin, K., Ruffert, J., Yurtdas, Y., Werth, M., Chen, W., Rabien, A., Kilic, E., Schulzke, J. D., et al. (2015). A Grainyhead-Like 2/Ovo-Like 2 Pathway Regulates Renal Epithelial Barrier Function and Lumen Expansion. J Am Soc Nephrol 26, 2704-2715. Baker, R. K. and Antin, P. B. (2003). Ephs and ephrins during early stages of chick embryogenesis. Dev Dyn 228, 128-142. Balda, M. S., Garrett, M. D. and Matter, K. (2003). The ZO-1-associated Y-box factor ZONAB regulates epithelial cell proliferation and cell density. J Cell Biol 160, 423-432. Balda, M. S. and Matter, K. (2000). The tight junction protein ZO-1 and an interacting transcription factor regulate ErbB-2 expression. Embo J 19, 2024-2033. Balda, M. and Matter, K. (2009). Tight junctions and the regulation of gene expression. Biochim Biophys Acta 1788, 761-767. Balda, M. S., Whitney, J. A., Flores, C., Gonzalez, S., Cereijido, M. and Matter, K. (1996). Functional dissociation of paracellular permeability and transepithelial electrical resistance and disruption of the apical-basolateral intramembrane diffusion barrier by expression of a mutant tight junction membrane protein. J Cell Biol 134, 1031-1049. Baltzegar, D. A., Reading, B. J., Brune, E. S. and Borski, R. J. (2013). Phylogenetic revision of the claudin gene family. Mar Genomics 11, 17-26. Bancroft, M. and Bellairs, R. (1975). Differentiation of the neural plate and neural tube in the young chick embryo. A study by scanning and transmission electron microscopy. Anat Embryol (Berl) 147, 309-335. Bardet, C., Ribes, S., Wu, Y., Diallo, M. T., Salmon, B., Breiderhoff, T., Houillier, P., Muller, D. and Chaussain, C. (2017). Claudin Loss-of-Function Disrupts Tight Junctions and Impairs Amelogenesis. Front Physiol 8, 326. Bashir, Z. E., Latief, N., Belyantseva, I. A., Iqbal, F., Riazuddin, S. A., Khan, S. N., Friedman, T. B., Riazuddin, S. and Riazuddin, S. (2013). Phenotypic variability of CLDN14 mutations causing DFNB29 hearing loss in the Pakistani population. J Hum Genet 58, 102-108. Baumholtz, A. I., Gupta, I. R. and Ryan, A. K. (2017a). Claudins in morphogenesis: Forming an epithelial tube. Tissue barriers, e1361899. Baumholtz, A. I., Simard, A., Nikolopoulou, E., Oosenbrug, M., Collins, M. M., Piontek, A., Krause, G., Piontek, J., Greene, N. D. E. and Ryan, A. K. (2017b). Claudins are essential for cell shape changes and convergent extension movements during neural tube closure. Dev Biol. Bazzoni, G., Martinez-Estrada, O. M., Orsenigo, F., Cordenonsi, M., Citi, S. and Dejana, E. (2000). Interaction of junctional adhesion molecule with the tight junction components ZO-1, cingulin, and occludin. J Biol Chem 275, 20520-20526. Behr, M., Riedel, D. and Schuh, R. (2003). The claudin-like megatrachea is essential in septate junctions for the epithelial barrier function in Drosophila. Dev Cell 5, 611-620. Bell, G. W., Yatskievych, T. A. and Antin, P. B. (2004). GEISHA, a whole-mount in situ hybridization gene expression screen in chicken embryos. Dev Dyn 229, 677-687.

180

Ben-Yosef, T., Belyantseva, I. A., Saunders, T. L., Hughes, E. D., Kawamoto, K., Van Itallie, C. M., Beyer, L. A., Halsey, K., Gardner, D. J., Wilcox, E. R., et al. (2003). Claudin 14 knockout mice, a model for autosomal recessive deafness DFNB29, are deaf due to cochlear hair cell degeneration. Hum Mol Genet 12, 2049-2061. Bhavaniprasad, V., Prabhu Dass, J. F. and Jayanthi, S. (2013). Activation mechanism of claudin-4 by ephrin type-A receptor 2: a molecular dynamics approach. Mol Biosyst 9, 2627-2634. Blencowe, H., Cousens, S., Modell, B. and Lawn, J. (2010). Folic acid to reduce neonatal mortality from neural tube disorders. Int J Epidemiol 39 Suppl 1, i110-121. Blom, N., Gammeltoft, S. and Brunak, S. (1999). Sequence and structure-based prediction of eukaryotic protein phosphorylation sites. J Mol Biol 294, 1351-1362. Bosoi, C. M., Capra, V., Allache, R., Trinh, V. Q., De Marco, P., Merello, E., Drapeau, P., Bassuk, A. G. and Kibar, Z. (2011). Identification and characterization of novel rare mutations in the planar cell polarity gene PRICKLE1 in human neural tube defects. Hum Mutat 32, 1371-1375. Boyer, J. L. (2013). Bile formation and secretion. Comprehensive Physiology 3, 1035-1078. Breiderhoff, T., Himmerkus, N., Stuiver, M., Mutig, K., Will, C., Meij, I. C., Bachmann, S., Bleich, M., Willnow, T. E. and Muller, D. (2012). Deletion of claudin-10 (Cldn10) in the thick ascending limb impairs paracellular sodium permeability and leads to hypermagnesemia and nephrocalcinosis. Proc Natl Acad Sci U.S.A 109, 14241-14246. Brizuela, B. J., Wessely, O. and De Robertis, E. M. (2001). Overexpression of the Xenopus tight-junction protein claudin causes randomization of the left-right body axis. Dev Biol 230, 217-229. Brouns, M. R., De Castro, S. C., Terwindt-Rouwenhorst, E. A., Massa, V., Hekking, J. W., Hirst, C. S., Savery, D., Munts, C., Partridge, D., Lamers, W., et al. (2011). Over- expression of Grhl2 causes spina bifida in the Axial defects mutant mouse. Hum Mol Genet 20, 1536-1546. Brun, R. B. and Garson, J. A. (1983). Neurulation in the Mexican salamander (Ambystoma mexicanum): a drug study and cell shape analysis of the epidermis and the neural plate. J Embryol Exp Morphol 74, 275-295. Burnside, B. (1971). Microtubules and microfilaments in newt neuralation. Developmental biology 26, 416-441. Burren K. A., Savery D., Massa V., Kok R. M., Scott J. M., Blom H. J., Copp A. J. and Greene, N. D. (2008). Gene-environment interactions in the causation of neural tube defects: folate deficiency increases susceptibility conferred by loss of Pax3 function. Hum Mol Genet 17, 3675-85. Cadot, B., Gache, V., Vasyutina, E., Falcone, S., Birchmeier, C. and Gomes, E. R. (2012). Nuclear movement during myotube formation is microtubule and dynein dependent and is regulated by Cdc42, Par6 and Par3. EMBO reports 13, 741-749. Campanale, J. P., Sun, T. Y. and Montell, D. J. (2017). Development and dynamics of cell polarity at a glance. Journal of cell science 130, 1201-1207. Cecconi, F., Alvarez-Bolado, G., Meyer, B. I., Roth, K. A. and Gruss, P. (1998). Apaf1 (CED-4 homolog) regulates programmed cell death in mammalian development. Cell 94, 727-737. Chakrabarti, G., Zhou, X. and McClane, B. A. (2003). Death pathways activated in CaCo-2 cells by Clostridium perfringens enterotoxin. Infect Immun 71, 4260-4270.

181

Chapman, S. C., Schubert, F. R., Schoenwolf, G. C. and Lumsden, A. (2002). Analysis of spatial and temporal gene expression patterns in blastula and gastrula stage chick embryos. Dev Biol 245, 187-199. Chapman, S. C., Schubert, F. R., Schoenwolf, G. C. and Lumsden, A. (2003). Anterior identity is established in chick epiblast by hypoblast and anterior definitive endoderm. Development 130, 5091-5101. Chen, B. and Hales, B. F. (1995). Antisense oligonucleotide down-regulation of E-cadherin in the yolk sac and cranial neural tube malformations. Biol Reprod 53, 1229-1238. Chen, C. M., Freedman, J. A., Bettler, D. R., Jr., Manning, S. D., Giep, S. N., Steiner, J. and Ellis, H. M. (1996). Polychaetoid is required to restrict segregation of sensory organ precursors from proneural clusters in Drosophila. Mech Dev 57, 215-227. Chen, F., Ma, L., Parrini, M. C., Mao, X., Lopez, M., Wu, C., Marks, P. W., Davidson, L., Kwiatkowski, D. J., Kirchhausen, T., et al. (2000). Cdc42 is required for PIP(2)- induced actin polymerization and early development but not for cell viability. Curr Biol 10, 758-765. Chen, X., An, Y., Gao, Y., Guo, L., Rui, L., Xie, H., Sun, M., Lam Hung, S., Sheng, X., Zou, J., et al. (2017). Rare Deleterious PARD3 Variants in the aPKC-Binding Region are Implicated in the Pathogenesis of Human Cranial Neural Tube Defects Via Disrupting Apical Tight Junction Formation. Hum Mutat 38, 378-389. Chen, Y., Merzdorf, C., Paul, D. L. and Goodenough, D. A. (1997). COOH terminus of occludin is required for tight junction barrier function in early Xenopus embryos. J Cell Biol 138, 891-899. Cheung, I. D., Bagnat, M., Ma, T. P., Datta, A., Evason, K., Moore, J. C., Lawson, N. D., Mostov, K. E., Moens, C. B. and Stainier, D. Y. (2012). Regulation of intrahepatic biliary duct morphogenesis by Claudin 15-like b. Dev Biol 361, 68-78. Chiang, C., Litingtung, Y., Lee, E., Young, K. E., Corden, J. L., Westphal, H. and Beachy, P. A. (1996). Cyclopia and defective axial patterning in mice lacking Sonic hedgehog gene function. Nature 383, 407-413. Chiba, K., Yamaguchi, K., Ando, M., Miyake, H. and Fujisawa, M. (2012). Expression pattern of testicular claudin-11 in infertile men. Urology 80, 1161.e1113-1167. Choi, W., Jung, K. C., Nelson, K. S., Bhat, M. A., Beitel, G. J., Peifer, M. and Fanning, A. S. (2011). The single Drosophila ZO-1 protein Polychaetoid regulates embryonic morphogenesis in coordination with Canoe/afadin and Enabled. Mol Biol Cell 22, 2010- 2030. Ciruna, B., Jenny, A., Lee, D., Mlodzik, M. and Schier, A. F. (2006). Planar cell polarity signalling couples cell division and morphogenesis during neurulation. Nature 439, 220- 224. Claverie-Martin, F., Garcia-Nieto, V., Loris, C., Ariceta, G., Nadal, I., Espinosa, L., Fernandez-Maseda, A., Anton-Gamero, M., Avila, A., Madrid, A., et al. (2013). Claudin-19 mutations and clinical phenotype in Spanish patients with familial hypomagnesemia with hypercalciuria and nephrocalcinosis. PloS one 8, e53151. Colas, J. F. and Schoenwolf, G. C. (2001). Towards a cellular and molecular understanding of neurulation. Dev Dyn 221, 117-145. Colas, J. F. and Schoenwolf, G. C. (2003a). Assessing the contributions of gene products to the form-shaping events of neurulation: a transgenic approach in chick. Genesis 37, 64-75.

182

Colas, J. F. and Schoenwolf, G. C. (2003b). Differential expression of two cell adhesion molecules, Ephrin-A5 and Integrin alpha6, during cranial neurulation in the chick embryo. Dev Neurosci 25, 357-365. Colegio, O. R., Van Itallie, C., Rahner, C. and Anderson, J. M. (2003). Claudin extracellular domains determine paracellular charge selectivity and resistance but not tight junction fibril architecture. Am J Cell Physiol 284, C1346-1354. Collins, M. M., Baumholtz, A. I. and Ryan, A. K. (2012). Claudin-5 expression in the vasculature of the developing chick embryo. Gene Expr Patterns 12, 123-129. Collins, M. M., Baumholtz, A. I. and Ryan, A. K. (2013). Claudin family members exhibit unique temporal and spatial expression boundaries in the chick embryo. Tissue barriers 1, e24517. Collins, M. M., Baumholtz, A. I., Simard, A., Gregory, M., Cyr, D. G. and Ryan, A. K. (2015). Claudin-10 is required for relay of left-right patterning cues from Hensen's node to the lateral plate mesoderm. Dev Biol 401, 236-248. Collins, M. M. and Ryan, A. K. (2011). Manipulating claudin expression in avian embryos. Methods Mol Biol 762, 195-212. Consortium, I. C. G. S. (2004). Sequence and comparative analysis of the chicken genome provide unique perspectives on vertebrate evolution. Nature 432, 695-716. Cook, T. A., Nagasaki, T. and Gundersen, G. G. (1998). Rho guanosine triphosphatase mediates the selective stabilization of microtubules induced by lysophosphatidic acid. J C Biol 141, 175-185. Copp, A. J. (2005). Neurulation in the cranial region--normal and abnormal. J Anat 207, 623- 635. Cordenonsi, M., D'Atri, F., Hammar, E., Parry, D. A., Kendrick-Jones, J., Shore, D. and Citi, S. (1999). Cingulin contains globular and coiled-coil domains and interacts with ZO-1, ZO-2, ZO-3, and myosin. J Cell Biol 147, 1569-1582. Cunningham, S. A., Arrate, M. P., Rodriguez, J. M., Bjercke, R. J., Vanderslice, P., Morris, A. P. and Brock, T. A. (2000). A novel protein with homology to the junctional adhesion molecule. Characterization of leukocyte interactions. J Biol Chem 275, 34750- 34756. Curtin, J. A., Quint, E., Tsipouri, V., Arkell, R. M., Cattanach, B., Copp, A. J., Henderson, D. J., Spurr, N., Stanier, P., Fisher, E. M., et al. (2003). Mutation of Celsr1 disrupts planar polarity of inner ear hair cells and causes severe neural tube defects in the mouse. Curr Biol : CB 13, 1129-1133. D'Atri, F. and Citi, S. (2001). Cingulin interacts with F-actin in vitro. FEBS letters 507, 21-24. D'Atri, F., Nadalutti, F. and Citi, S. (2002). Evidence for a functional interaction between cingulin and ZO-1 in cultured cells. J Biol Chem 277, 27757-27764. D'Souza, T., Agarwal, R. and Morin, P. J. (2005). Phosphorylation of claudin-3 at threonine 192 by cAMP-dependent protein kinase regulates tight junction barrier function in ovarian cancer cells. J Biol Chem 280, 26233-26240. Darnell, D. K., Kaur, S., Stanislaw, S., Davey, S., Konieczka, J. H., Yatskievych, T. A. and Antin, P. B. (2007). GEISHA: an in situ hybridization gene expression resource for the chicken embryo. Cytogenet Genome Res 117, 30-35. Daugherty, B. L., Ward, C., Smith, T., Ritzenthaler, J. D. and Koval, M. (2007). Regulation of heterotypic claudin compatibility. J Biol Chemi 282, 30005-30013.

183

Davidson, L. A. and Keller, R. E. (1999). Neural tube closure in Xenopus laevis involves medial migration, directed protrusive activity, cell intercalation and convergent extension. Development 126, 4547-4556. De Marco, P., Merello, E., Cama, A., Kibar, Z. and Capra, V. (2011). Human neural tube defects: genetic causes and prevention. BioFactors 37, 261-268. De Marco, P., Merello, E., Consales, A., Piatelli, G., Cama, A., Kibar, Z. and Capra, V. (2013). Genetic analysis of disheveled 2 and disheveled 3 in human neural tube defects. J Mol Neurosci 49, 582-588. De Marco, P., Merello, E., Rossi, A., Piatelli, G., Cama, A., Kibar, Z. and Capra, V. (2012). FZD6 is a novel gene for human neural tube defects. Hum Mutat 33, 384-390. De Wals, P., Tairou, F., Van Allen, M. I., Uh, S. H., Lowry, R. B., Sibbald, B., Evans, J. A., Van den Hof, M. C., Zimmer, P., Crowley, M., et al. (2007). Reduction in neural-tube defects after folic acid fortification in Canada. N Engl J Med 357, 135-142. Denninger, A. R., Breglio, A., Maheras, K. J., LeDuc, G., Cristiglio, V., Deme, B., Gow, A. and Kirschner, D. A. (2015). Claudin-11 Tight Junctions in Myelin Are a Barrier to Diffusion and Lack Strong Adhesive Properties. Biophys J 109, 1387-1397. Devaux, J. and Gow, A. (2008). Tight junctions potentiate the insulative properties of small CNS myelinated axons. J Cell Biol 183, 909-921. Devenport, D. (2016). Tissue morphodynamics: Translating planar polarity cues into polarized cell behaviors. Semin Cell Dev Biol 55, 99-110. Dickinson, M. E., Selleck, M. A., McMahon, A. P. and Bronner-Fraser, M. (1995). Dorsalization of the neural tube by the non-neural ectoderm. Development 121, 2099- 2106. Ding, L., Lu, Z., Foreman, O., Tatum, R., Lu, Q., Renegar, R., Cao, J. and Chen, Y. H. (2012). Inflammation and disruption of the mucosal architecture in claudin-7-deficient mice. Gastroenterology 142, 305-315. Ebihara, C., Kondoh, M., Hasuike, N., Harada, M., Mizuguchi, H., Horiguchi, Y., Fujii, M. and Watanabe, Y. (2006). Preparation of a claudin-targeting molecule using a C- terminal fragment of Clostridium perfringens enterotoxin. J Pharmacol Exp Ther 316, 255-260. Ebnet, K., Aurrand-Lions, M., Kuhn, A., Kiefer, F., Butz, S., Zander, K., Meyer zu Brickwedde, M. K., Suzuki, A., Imhof, B. A. and Vestweber, D. (2003). The junctional adhesion molecule (JAM) family members JAM-2 and JAM-3 associate with the cell polarity protein PAR-3: a possible role for JAMs in endothelial cell polarity. J Cell Sci 116, 3879-3891. Ekinci, Z., Karabas, L. and Konrad, M. (2012). Hypomagnesemia-hypercalciuria- nephrocalcinosis and ocular findings: a new claudin-19 mutation. Turk J Pediatr 54, 168- 170. Elul, T., Koehl, M. A. and Keller, R. (1997). Cellular mechanism underlying neural convergent extension in Xenopus laevis embryos. Dev Biol 191, 243-258. Eom, D. S., Amarnath, S., Fogel, J. L. and Agarwala, S. (2011). Bone morphogenetic proteins regulate neural tube closure by interacting with the apicobasal polarity pathway. Development 138, 3179-3188. Escobedo, N., Contreras, O., Munoz, R., Farias, M., Carrasco, H., Hill, C., Tran, U., Pryor, S. E., Wessely, O., Copp, A. J., et al. (2013). Syndecan 4 interacts genetically with

184

Vangl2 to regulate neural tube closure and planar cell polarity. Development 140, 3008- 3017. Escuin, S., Vernay, B., Savery, D., Gurniak, C. B., Witke, W., Greene, N. D. and Copp, A. J. (2015). Rho-kinase-dependent actin turnover and actomyosin disassembly are necessary for mouse spinal neural tube closure. J Cell Sci 128, 2468-2481. Everett, R. S., Vanhook, M. K., Barozzi, N., Toth, I. and Johnson, L. G. (2006). Specific modulation of airway epithelial tight junctions by apical application of an occludin peptide. Mol Pharmacol 69, 492-500. Farquhar, M. G. and Palade, G. E. (1963). Junctional complexes in various epithelia. The J Cell Biol 17, 375-412. Feldmeyer, L., Huber, M., Fellmann, F., Beckmann, J. S., Frenk, E. and Hohl, D. (2006). Confirmation of the origin of NISCH syndrome. Hum Mutat 27, 408-410. Fernandez, J. G., de Paz, P. and Chamorro, C. A. (1987). Effects of colchicine on the shape of chick neuroepithelial cells during neurulation. Anat Rec 219, 296-303. Findley, M. K. and Koval, M. (2009). Regulation and roles for claudin-family tight junction proteins. IUBMB life 61, 431-437. Flower, N. E. and Filshie, B. K. (1975). Junctional structures in the midgut cells of lepidopteran caterpillars. J Cell Sci 17, 221-239. Fredriksson, K., Van Itallie, C. M., Aponte, A., Gucek, M., Tietgens, A. J. and Anderson, J. M. (2015). Proteomic analysis of proteins surrounding occludin and claudin-4 reveals their proximity to signaling and trafficking networks. PloS one 10, e0117074. Fujibe, M., Chiba, H., Kojima, T., Soma, T., Wada, T., Yamashita, T. and Sawada, N. (2004). Thr203 of claudin-1, a putative phosphorylation site for MAP kinase, is required to promote the barrier function of tight junctions. Exp Cell Res 295, 36-47. Fujita, H., Hamazaki, Y., Noda, Y., Oshima, M. and Minato, N. (2012). Claudin-4 deficiency results in urothelial hyperplasia and lethal hydronephrosis. PloS one 7, e52272. Fujita, K., Katahira, J., Horiguchi, Y., Sonoda, N., Furuse, M. and Tsukita, S. (2000). Clostridium perfringens enterotoxin binds to the second extracellular loop of claudin-3, a tight junction integral membrane protein. FEBS letters 476, 258-261. Furuse, M., Fujita, K., Hiiragi, T., Fujimoto, K. and Tsukita, S. (1998a). Claudin-1 and -2: novel integral membrane proteins localizing at tight junctions with no sequence similarity to occludin. J Cell Biol 141, 1539-1550. Furuse, M., Furuse, K., Sasaki, H. and Tsukita, S. (2001). Conversion of zonulae occludentes from tight to leaky strand type by introducing claudin-2 into Madin-Darby canine kidney I cells. The Journal of cell biology 153, 263-272. Furuse, M., Hata, M., Furuse, K., Yoshida, Y., Haratake, A., Sugitani, Y., Noda, T., Kubo, A. and Tsukita, S. (2002). Claudin-based tight junctions are crucial for the mammalian epidermal barrier: a lesson from claudin-1-deficient mice. J Cell Biol 156, 1099-1111. Furuse, M., Hirase, T., Itoh, M., Nagafuchi, A., Yonemura, S., Tsukita, S. and Tsukita, S. (1993). Occludin: a novel integral membrane protein localizing at tight junctions. J Cell Biol 123, 1777-1788. Furuse, M., Itoh, M., Hirase, T., Nagafuchi, A., Yonemura, S., Tsukita, S. and Tsukita, S. (1994). Direct association of occludin with ZO-1 and its possible involvement in the localization of occludin at tight junctions. J Cell Biol 127, 1617-1626.

185

Furuse, M., Sasaki, H., Fujimoto, K. and Tsukita, S. (1998b). A single gene product, claudin- 1 or -2, reconstitutes tight junction strands and recruits occludin in fibroblasts. J Cell Biol 143, 391-401. Furuse, M., Sasaki, H. and Tsukita, S. (1999). Manner of interaction of heterogeneous claudin species within and between tight junction strands. J Cell Biol 147, 891-903. Geelen, J. A. and Langman, J. (1979). Ultrastructural observations on closure of the neural tube in the mouse. Anat Embryol (Berl) 156, 73-88. Gilula, N. B., Branton, D. and Satir, P. (1970). The septate junction: a structural basis for intercellular coupling. Proc Natl Acad Sci U.S.A. 67, 213-220. Gong, Y., Renigunta, V., Zhou, Y., Sunq, A., Wang, J., Yang, J., Renigunta, A., Baker, L. A. and Hou, J. (2015a). Biochemical and biophysical analyses of tight junction permeability made of claudin-16 and claudin-19 dimerization. Mol Biol Cell 26, 4333- 4346. Gong, Y., Wang, J., Yang, J., Gonzales, E., Perez, R. and Hou, J. (2015b). KLHL3 regulates paracellular chloride transport in the kidney by ubiquitination of claudin-8. Proc Natl Acad Sci U.S.A. 112, 4340-4345. Gonzalez-Mariscal, L., Islas, S., Contreras, R. G., Garcia-Villegas, M. R., Betanzos, A., Vega, J., Diaz-Quinonez, A., Martin-Orozco, N., Ortiz-Navarrete, V., Cereijido, M., et al. (1999). Molecular characterization of the tight junction protein ZO-1 in MDCK cells. Exp Cell Res 248, 97-109. González-Mariscal, L., Quirós, M., Díaz-Coránguez, M. and Bautista, P. (2012). Tight junctions. Curr Front Perspect Cell Biol (ed. S. Najman). Rijeka, Croatia InTech. Goto, T. and Keller, R. (2002). The planar cell polarity gene strabismus regulates convergence and extension and neural fold closure in Xenopus. Dev Biol 247, 165-181. Gow, A., Davies, C., Southwood, C. M., Frolenkov, G., Chrustowski, M., Ng, L., Yamauchi, D., Marcus, D. C. and Kachar, B. (2004). Deafness in Claudin 11-null mice reveals the critical contribution of basal cell tight junctions to stria vascularis function. J Neurosci 24, 7051-7062. Gow, A., Southwood, C. M., Li, J. S., Pariali, M., Riordan, G. P., Brodie, S. E., Danias, J., Bronstein, J. M., Kachar, B. and Lazzarini, R. A. (1999). CNS myelin and sertoli cell tight junction strands are absent in Osp/claudin-11 null mice. Cell 99, 649-659. Graper, L. (1929). Die Primitiventwicklung des Hünchens nach stereokinematographischen Untersuchungen, kontrolliert durch vitale Farbmarkierung und verglichen mit der Entwicklung anderer Wirbeltiere. Arch. EntwMech. Org 116, 382-429. Greene, N. D. and Copp, A. J. (1997). Inositol prevents folate-resistant neural tube defects in the mouse. Nat Med 3, 60-66. Greene, N. D. and Copp, A. J. (2009). Development of the vertebrate central nervous system: formation of the neural tube. Prenatal diagnosis 29, 303-311. Greene, N. D., Gerrelli, D., Van Straaten, H. W. and Copp, A. J. (1998). Abnormalities of floor plate, notochord and somite differentiation in the loop-tail (Lp) mouse: a model of severe neural tube defects. Mech Dev 73, 59-72. Greene, N. D., Leung, K. Y., Gay, V., Burren, K., Mills, K., Chitty, L. S. and Copp, A. J. (2016). Inositol for the prevention of neural tube defects: a pilot randomised controlled trial. Br J Nutr 115, 974-983.

186

Grego-Bessa, J., Hildebrand, J. and Anderson, K. V. (2015). Morphogenesis of the mouse neural plate depends on distinct roles of cofilin 1 in apical and basal epithelial domains. Development 142, 1305-1314. Guillemot, L. and Citi, S. (2006). Cingulin regulates claudin-2 expression and cell proliferation through the small GTPase RhoA. Mol Biol Cell 17, 3569-3577. Guillemot, L., Hammar, E., Kaister, C., Ritz, J., Caille, D., Jond, L., Bauer, C., Meda, P. and Citi, S. (2004). Disruption of the cingulin gene does not prevent tight junction formation but alters gene expression. J Cell Sci 117, 5245-5256. Guillemot, L., Paschoud, S., Jond, L., Foglia, A. and Citi, S. (2008a). Paracingulin regulates the activity of Rac1 and RhoA GTPases by recruiting Tiam1 and GEF-H1 to epithelial junctions. Mol Biol Cell 19, 4442-4453. Guillemot, L., Paschoud, S., Pulimeno, P., Foglia, A. and Citi, S. (2008b). The cytoplasmic plaque of tight junctions: a scaffolding and signalling center. Biochim Biophys Acta 1778, 601-613. Gumbiner, B., Lowenkopf, T. and Apatira, D. (1991). Identification of a 160-kDa polypeptide that binds to the tight junction protein ZO-1. Proc Natl Acad Sci U.S.A. 88, 3460-3464. Guney, O., Canbilen, A., Konak, A. and Acar, O. (2003). The effects of folic acid in the prevention of neural tube development defects caused by phenytoin in early chick embryos. Spine 28, 442-445. Gunzel, D. and Yu, A. S. (2013). Claudins and the modulation of tight junction permeability. Physiol Rev 93, 525-569. Guo, N., Hawkins, C. and Nathans, J. (2004). Frizzled6 controls hair patterning in mice. Proc Natl Acad Sci U.S.A. 101, 9277-9281. Guo, P., Weinstein, A. M. and Weinbaum, S. (2003). A dual-pathway ultrastructural model for the tight junction of rat proximal tubule epithelium. Am J Physiol Renal Physiol 285, F241-257. Hackett, D. A., Smith, J. L. and Schoenwolf, G. C. (1997). Epidermal ectoderm is required for full elevation and for convergence during bending of the avian neural plate. Dev Dyn 210, 397-406. Hadj-Rabia, S., Baala, L., Vabres, P., Hamel-Teillac, D., Jacquemin, E., Fabre, M., Lyonnet, S., De Prost, Y., Munnich, A., Hadchouel, M., et al. (2004). Claudin-1 gene mutations in neonatal sclerosing cholangitis associated with ichthyosis: a tight junction disease. Gastroenterology 127, 1386-1390. Hadj-Rabia, S., Brideau, G., Al-Sarraj, Y., Maroun, R. C., Figueres, M. L., Leclerc- Mercier, S., Olinger, E., Baron, S., Chaussain, C., Nochy, D., et al. (2017). Multiplex epithelium dysfunction due to CLDN10 mutation: the HELIX syndrome. Genet Med. Hamazaki, Y., Itoh, M., Sasaki, H., Furuse, M. and Tsukita, S. (2002). Multi-PDZ domain protein 1 (MUPP1) is concentrated at tight junctions through its possible interaction with claudin-1 and junctional adhesion molecule. J Biol Chem 277, 455-461. Hamblet, N. S., Lijam, N., Ruiz-Lozano, P., Wang, J., Yang, Y., Luo, Z., Mei, L., Chien, K. R., Sussman, D. J. and Wynshaw-Boris, A. (2002). Dishevelled 2 is essential for cardiac outflow tract development, somite segmentation and neural tube closure. Development 129, 5827-5838. Hamburger, V. and Hamilton, H. L. (1951). A series of normal stages in the development of the chick embryo. J Morphol 88, 49-92.

187

Hanna, P. C. and McClane, B. A. (1991). A recombinant C-terminal toxin fragment provides evidence that membrane insertion is important for Clostridium perfringens enterotoxin cytotoxicity. Mol Microbiol 5, 225-230. Hanna, P. C., Mietzner, T. A., Schoolnik, G. K. and McClane, B. A. (1991). Localization of the receptor-binding region of Clostridium perfringens enterotoxin utilizing cloned toxin fragments and synthetic peptides. The 30 C-terminal amino acids define a functional binding region. J Biol Chem 266, 11037-11043. Hanna, P. C., Wnek, A. P. and McClane, B. A. (1989). Molecular cloning of the 3' half of the Clostridium perfringens enterotoxin gene and demonstration that this region encodes receptor-binding activity. J Bacteriol 171, 6815-6820. Haorah, J., Heilman, D., Knipe, B., Chrastil, J., Leibhart, J., Ghorpade, A., Miller, D. W. and Persidsky, Y. (2005). Ethanol-induced activation of myosin light chain kinase leads to dysfunction of tight junctions and blood-brain barrier compromise. Alcohol Clin Exp Res 29, 999-1009. Harada, M., Kondoh, M., Ebihara, C., Takahashi, A., Komiya, E., Fujii, M., Mizuguchi, H., Tsunoda, S., Horiguchi, Y., Yagi, K., et al. (2007). Role of tyrosine residues in modulation of claudin-4 by the C-terminal fragment of Clostridium perfringens enterotoxin. Biochem Pharmacol 73, 206-214. Harris, M. J. and Juriloff, D. M. (2010). Birth Defects Res A Clin Mol Teratol. Birth defects research. Part A, Clinical and molecular teratology 88, 653-669. Haskins, J., Gu, L., Wittchen, E. S., Hibbard, J. and Stevenson, B. R. (1998). ZO-3, a novel member of the MAGUK protein family found at the tight junction, interacts with ZO-1 and occludin. J Cell Biol 141, 199-208. Hawley, S. H., Wunnenberg-Stapleton, K., Hashimoto, C., Laurent, M. N., Watabe, T., Blumberg, B. W. and Cho, K. W. (1995). Disruption of BMP signals in embryonic Xenopus ectoderm leads to direct neural induction. Genes Dev 9, 2923-2935. Hayashi, D., Tamura, A., Tanaka, H., Yamazaki, Y., Watanabe, S., Suzuki, K., Suzuki, K., Sentani, K., Yasui, W., Rakugi, H., et al. (2012). Deficiency of claudin-18 causes paracellular H+ leakage, up-regulation of interleukin-1beta, and atrophic gastritis in mice. Gastroenterology 142, 292-304. Heid, M. K., Bills, N. D., Hinrichs, S. H. and Clifford, A. J. (1992) J Nutr 122, 888-894. Hemmati-Brivanlou, A. and Melton, D. (1997). Vertebrate embryonic cells will become nerve cells unless told otherwise. Cell 88, 13-17. Hol, F. A., Geurds, M. P., Chatkupt, S., Shugart, Y. Y., Balling, R., Schrander-Stumpel, C. T., Johnson, W. G., Hamel, B. C. and Mariman, E. C. (1996). PAX genes and human neural tube defects: an amino acid substitution in PAX1 in a patient with spina bifida. J Med Genet 33, 655-660. Holmberg, J., Clarke, D. L. and Frisen, J. (2000). Regulation of repulsion versus adhesion by different splice forms of an Eph receptor. Nature 408, 203-206. Horiguchi, Y., Akai, T. and Sakaguchi, G. (1987). Isolation and function of a Clostridium perfringens enterotoxin fragment. Infect Immun 55, 2912-2915. Hou, J., Rajagopal, M. and Yu, A. S. (2013). Claudins and the kidney. Annual review of physiology 75, 479-501. Hou, J., Renigunta, A., Gomes, A. S., Hou, M., Paul, D. L., Waldegger, S. and Goodenough, D. A. (2009). Claudin-16 and claudin-19 interaction is required for their assembly into

188

tight junctions and for renal reabsorption of magnesium. Proc Natl Acad Sci U.S.A. 106, 15350-15355. Hou, J., Renigunta, A., Konrad, M., Gomes, A. S., Schneeberger, E. E., Paul, D. L., Waldegger, S. and Goodenough, D. A. (2008). Claudin-16 and claudin-19 interact and form a cation-selective tight junction complex. J Clin Invest 118, 619-628. Hou, J., Renigunta, A., Yang, J. and Waldegger, S. (2010). Claudin-4 forms paracellular chloride channel in the kidney and requires claudin-8 for tight junction localization. Proc Natl Acad Sci U.S.A. 107, 18010-18015. Hou, J., Shan, Q., Wang, T., Gomes, A. S., Yan, Q., Paul, D. L., Bleich, M. and Goodenough, D. A. (2007). Transgenic RNAi depletion of claudin-16 and the renal handling of magnesium. J Biol Chem 282, 17114-17122. Huerta, M., Munoz, R., Tapia, R., Soto-Reyes, E., Ramirez, L., Recillas-Targa, F., Gonzalez-Mariscal, L. and Lopez-Bayghen, E. (2007). Cyclin D1 is transcriptionally down-regulated by ZO-2 via an E box and the transcription factor c-Myc. Mol Biol Cell 18, 4826-4836. Hughes, S. H. (2004). The RCAS vector system. Folia Biol (Praha) 50, 107-119. Hurd, T. W., Gao, L., Roh, M. H., Macara, I. G. and Margolis, B. (2003). Direct interaction of two polarity complexes implicated in epithelial tight junction assembly. Nat Cell Biol 5, 137-142. Ikari, A., Matsumoto, S., Harada, H., Takagi, K., Hayashi, H., Suzuki, Y., Degawa, M. and Miwa, M. (2006). Phosphorylation of paracellin-1 at Ser217 by protein kinase A is essential for localization in tight junctions. J Cell Sci 119, 1781-1789. Ikenouchi, J., Furuse, M., Furuse, K., Sasaki, H., Tsukita, S. and Tsukita, S. (2005). Tricellulin constitutes a novel barrier at tricellular contacts of epithelial cells. J Cell Biol 171, 939-945. Ikenouchi, J., Matsuda, M., Furuse, M. and Tsukita, S. (2003). Regulation of tight junctions during the epithelium-mesenchyme transition: direct repression of the gene expression of claudins/occludin by Snail. J Cell Sci 116, 1959-1967. Ikenouchi, J., Sasaki, H., Tsukita, S., Furuse, M. and Tsukita, S. (2008). Loss of occludin affects tricellular localization of tricellulin. Mol Biol Cell 19, 4687-4693. Inagaki, T., Wang, K. C., Higbee, R. G., McLone, D. G. and Knepper, P. A. (1996). Concanavalin-A-induced open neural tube defects in chick embryos. Neurologia medico- chirurgica 36, 691-696; discussion 696-697. Inai, T., Kobayashi, J. and Shibata, Y. (1999). Claudin-1 contributes to the epithelial barrier function in MDCK cells. Eur J Cell Biol 78, 849-855. Itoh, M., Furuse, M., Morita, K., Kubota, K., Saitou, M. and Tsukita, S. (1999a). Direct binding of three tight junction-associated MAGUKs, ZO-1, ZO-2, and ZO-3, with the COOH termini of claudins. J Cell Biol 147, 1351-1363. Itoh, M., Morita, K. and Tsukita, S. (1999b). Characterization of ZO-2 as a MAGUK family member associated with tight as well as adherens junctions with a binding affinity to occludin and alpha catenin. J Biol Chem 274, 5981-5986. Itoh, M., Tsukita, S., Yamazaki, Y. and Sugimoto, H. (2012). Rho GTP exchange factor ARHGEF11 regulates the integrity of epithelial junctions by connecting ZO-1 and RhoA- myosin II signaling. Proc Natl Acad Sci U.S.A. 109, 9905-9910. Jacobson, A. G. and Moury, J. D. (1995). Tissue boundaries and cell behavior during neurulation. Dev Biol 171, 98-110.

189

Jeansonne, B., Lu, Q., Goodenough, D. A. and Chen, Y. H. (2003). Claudin-8 interacts with multi-PDZ domain protein 1 (MUPP1) and reduces paracellular conductance in epithelial cells. Cell Mol Biol 49, 13-21. Jesuthasan, S. and Stahle, U. (1997). Dynamic microtubules and specification of the zebrafish embryonic axis. Curr Biol 7, 31-42. Jonusaite, S., Donini, A. and Kelly, S. P. (2016). Occluding junctions of invertebrate epithelia. J Comp Physiol B 186, 17-43. Jovov, B., Van Itallie, C. M., Shaheen, N. J., Carson, J. L., Gambling, T. M., Anderson, J. M. and Orlando, R. C. (2007). Claudin-18: a dominant tight junction protein in Barrett's esophagus and likely contributor to its acid resistance. Am J Physiol Gastrointest Liver Physiol 293, G1106-1113. Jung, A. C., Ribeiro, C., Michaut, L., Certa, U. and Affolter, M. (2006). Polychaetoid/ZO-1 is required for cell specification and rearrangement during Drosophila tracheal morphogenesis. Curr Biol 16, 1224-1231. Juriloff, D. M. and Harris, M. J. (2012). A consideration of the evidence that genetic defects in planar cell polarity contribute to the etiology of human neural tube defects. Birth Defects Res A Clin Mol Teratol 94, 824-840. Kaarteenaho, R., Merikallio, H., Lehtonen, S., Harju, T. and Soini, Y. (2010). Divergent expression of claudin -1, -3, -4, -5 and -7 in developing human lung. Respiratory research 11, 59. Kage, H., Flodby, P., Gao, D., Kim, Y. H., Marconett, C. N., DeMaio, L., Kim, K. J., Crandall, E. D. and Borok, Z. (2014). Claudin 4 knockout mice: normal physiological phenotype with increased susceptibility to lung injury. Am J Physiol Lung Cell Mol Physiol 307, L524-536. Kajiwara, K., Berson, E. L. and Dryja, T. P. (1994). Digenic retinitis pigmentosa due to mutations at the unlinked peripherin/RDS and ROM1 loci. Science 264, 1604-1608. Kamberov, E., Makarova, O., Roh, M., Liu, A., Karnak, D., Straight, S. and Margolis, B. (2000). Molecular cloning and characterization of Pals, proteins associated with mLin-7. J Biol Chem 275, 11425-11431. Karfunkel, P. (1971). The role of microtubules and microfilaments in neurulation in Xenopus. Dev Biol 25, 30-56. Karfunkel, P.(1972). The activity of microtubules and microfilaments in neurulation in the chick. The Journal of experimental zoology 181, 289-301. Katahira, J., Inoue, N., Horiguchi, Y., Matsuda, M. and Sugimoto, N. (1997a). Molecular cloning and functional characterization of the receptor for Clostridium perfringens enterotoxin. J Cell Biol 136, 1239-1247. Katahira, J., Sugiyama, H., Inoue, N., Horiguchi, Y., Matsuda, M. and Sugimoto, N. (1997b). Clostridium perfringens enterotoxin utilizes two structurally related membrane proteins as functional receptors in vivo. J Biol Chem 272, 26652-26658. Katsanis, N., Ansley, S. J., Badano, J. L., Eichers, E. R., Lewis, R. A., Hoskins, B. E., Scambler, P. J., Davidson, W. S., Beales, P. L. and Lupski, J. R. (2001). Triallelic inheritance in Bardet-Biedl syndrome, a Mendelian recessive disorder. Science 293, 2256-2259. Katsuno, T., Umeda, K., Matsui, T., Hata, M., Tamura, A., Itoh, M., Takeuchi, K., Fujimori, T., Nabeshima, Y., Noda, T., et al. (2008). Deficiency of zonula occludens-1

190

causes embryonic lethal phenotype associated with defected yolk sac angiogenesis and apoptosis of embryonic cells. Mol Biol Cell 19, 2465-2475. Keller, R., Shih, J. and Sater, A. (1992). The cellular basis of the convergence and extension of the Xenopus neural plate. Dev Dyn 193, 199-217. Kemphues, K. J., Priess, J. R., Morton, D. G. and Cheng, N. S. (1988). Identification of genes required for cytoplasmic localization in early C. elegans embryos. Cell 52, 311-320. Kharfallah, F., Guyot, M. C., El Hassan, A. R., Allache, R., Merello, E., De Marco, P., Di Cristo, G., Capra, V. and Kibar, Z. (2017). Scribble1 plays an important role in the pathogenesis of neural tube defects through its mediating effect of Par-3 and Vangl1/2 localization. Hum Mol Genet 26, 2307-2320. Khokha, M. K., Yeh, J., Grammer, T. C. and Harland, R. M. (2005). Depletion of three BMP antagonists from Spemann's organizer leads to a catastrophic loss of dorsal structures. Dev Cell 8, 401-411. Khudyakov, J. and Bronner-Fraser, M. (2009). Comprehensive spatiotemporal analysis of early chick neural crest network genes. Dev Dyn 238, 716-723. Kibar, Z., Bosoi, C. M., Kooistra, M., Salem, S., Finnell, R. H., De Marco, P., Merello, E., Bassuk, A. G., Capra, V. and Gros, P. (2009). Novel mutations in VANGL1 in neural tube defects. Hum Mutat 30, E706-715. Kibar, Z., Capra, V. and Gros, P. (2007). Toward understanding the genetic basis of neural tube defects. Clinical genetics 71, 295-310. Kibar, Z., Salem, S., Bosoi, C. M., Pauwels, E., De Marco, P., Merello, E., Bassuk, A. G., Capra, V. and Gros, P. (2011). Contribution of VANGL2 mutations to isolated neural tube defects. Clin Genet 80, 76-82. Kiener, T. K., Selptsova-Friedrich, I. and Hunziker, W. (2008). Tjp3/zo-3 is critical for epidermal barrier function in zebrafish embryos. Dev Biol 316, 36-49. Kieserman, E. K. and Wallingford, J. B. (2009). In vivo imaging reveals a role for Cdc42 in spindle positioning and planar orientation of cell divisions during vertebrate neural tube closure. J Cell Sci 122, 2481-2490. Kim, E., Harris, A., Hyland, V. and Murrell, D. (2016). Digenic Inheritance in Epidermolysis Bullosa Simplex involving two novel mutations in KRT5 and KRT14. Br J Dermatol. Kimura-Yoshida, C., Mochida, K., Ellwanger, K., Niehrs, C. and Matsuo, I. (2015). Fate Specification of Neural Plate Border by Canonical Wnt Signaling and Grhl3 is Crucial for Neural Tube Closure. EBioMedicine 2, 513-527. Kimura, J., Abe, H., Kamitani, S., Toshima, H., Fukui, A., Miyake, M., Kamata, Y., Sugita- Konishi, Y., Yamamoto, S. and Horiguchi, Y. (2010). Clostridium perfringens enterotoxin interacts with claudins via electrostatic attraction. J Biol Chem 285, 401-408. Kinoshita, N., Sasai, N., Misaki, K. and Yonemura, S. (2008). Apical accumulation of Rho in the neural plate is important for neural plate cell shape change and neural tube formation. Mol Biol Cell 19, 2289-2299. Kirchmeier, P., Sayar, E., Hotz, A., Hausser, I., Islek, A., Yilmaz, A., Artan, R. and Fischer, J. (2014). Novel mutation in the CLDN1 gene in a Turkish family with neonatal ichthyosis sclerosing cholangitis (NISCH) syndrome. Br J Dermatol 170, 976-978. Kitajiri, S., Miyamoto, T., Mineharu, A., Sonoda, N., Furuse, K., Hata, M., Sasaki, H., Mori, Y., Kubota, T., Ito, J., et al. (2004). Compartmentalization established by claudin-11-based tight junctions in stria vascularis is required for hearing through generation of endocochlear potential. J Cell Sci 117, 5087-5096.

191

Klar, J., Piontek, J., Milatz, S., Tariq, M., Jameel, M., Breiderhoff, T., Schuster, J., Fatima, A., Asif, M., Sher, M., et al. (2017). Altered paracellular cation permeability due to a rare CLDN10B variant causes anhidrosis and kidney damage. PLoS Genet 13, e1006897. Kokai-Kun, J. F. and McClane, B. A. (1997). Deletion analysis of the Clostridium perfringens enterotoxin. Infect Immun 65, 1014-1022. Kooij, G., Kopplin, K., Blasig, R., Stuiver, M., Koning, N., Goverse, G., van der Pol, S. M., van Het Hof, B., Gollasch, M., Drexhage, J. A., et al. (2014). Disturbed function of the blood-cerebrospinal fluid barrier aggravates neuro-inflammation. Acta neuropathologica 128, 267-277. Krause, G., Protze, J. and Piontek, J. (2015). Assembly and function of claudins: Structure- function relationships based on homology models and crystal structures. Semin Cell Dev Biol 42, 3-12. Krause, G., Winkler, L., Mueller, S. L., Haseloff, R. F., Piontek, J. and Blasig, I. E. (2008). Structure and function of claudins. Biochim Biophys Acta 1778, 631-645. Kubota, K., Furuse, M., Sasaki, H., Sonoda, N., Fujita, K., Nagafuchi, A. and Tsukita, S. (1999). Ca(2+)-independent cell-adhesion activity of claudins, a family of integral membrane proteins localized at tight junctions. Curr Biol 9, 1035-1038. Lal-Nag, M. and Morin, P. J. (2009). The claudins. Genome biology 10, 235. Lameris, A. L., Huybers, S., Kaukinen, K., Makela, T. H., Bindels, R. J., Hoenderop, J. G. and Nevalainen, P. I. (2013). Expression profiling of claudins in the human gastrointestinal tract in health and during inflammatory bowel disease. Scand J Gastroenterol 48, 58-69. Laukoetter, M. G., Nava, P., Lee, W. Y., Severson, E. A., Capaldo, C. T., Babbin, B. A., Williams, I. R., Koval, M., Peatman, E., Campbell, J. A., et al. (2007). JAM-A regulates permeability and inflammation in the intestine in vivo. J Exp Med 204, 3067- 3076. Lawson, A., Anderson, H. and Schoenwolf, G. C. (2001). Cellular mechanisms of neural fold formation and morphogenesis in the chick embryo. The Anatomical record 262, 153-168. Lee, H., Sheffield, J. B., Nagele, R. G., Jr. and Kalmus, G. W. (1976). The role of extracellular material in chick neurulation. I. Effects of concanavalin A. J Exp Zool 198, 261-266. Lee, H. Y., Kosciuk, M. C., Nagele, R. G. and Roisen, F. J. (1983). Studies on the mechanisms of neurulation in the chick: possible involvement of myosin in elevation of neural folds. J Exp Zool 225, 449-457. Lee, H. Y. and Nagele, R. G. (1985). Studies on the mechanisms of neurulation in the chick: interrelationship of contractile proteins, microfilaments, and the shape of neuroepithelial cells. J Exp Zool 235, 205-215. Lee, H. Y., Sheffield, J. B. and Nagele, R. G., Jr. (1978). The role of extracellular material in chick neurulation. II. Surface morphology of neuroepithelial cells during neural fold fusion. J Exp Zool 204, 137-153. Lee, J. Y. and Harland, R. M. (2007). Actomyosin contractility and microtubules drive apical constriction in Xenopus bottle cells. Dev Biol 311, 40-52. Lee, J. Y. and Harland, R. M. (2010). Endocytosis is required for efficient apical constriction during Xenopus gastrulation. Curr Biol 20, 253-258.

192

Lei, Y., Zhu, H., Duhon, C., Yang, W., Ross, M. E., Shaw, G. M. and Finnell, R. H. (2013). Mutations in planar cell polarity gene SCRIB are associated with spina bifida. PloS one 8, e69262. Lei, Y. P., Zhang, T., Li, H., Wu, B. L., Jin, L. and Wang, H. Y. (2010). VANGL2 mutations in human cranial neural-tube defects. N Engl J Med 362, 2232-2235. Lemay, P., Guyot, M. C., Tremblay, E., Dionne-Laporte, A., Spiegelman, D., Henrion, E., Diallo, O., De Marco, P., Merello, E., Massicotte, C., et al. (2015). Loss-of-function de novo mutations play an important role in severe human neural tube defects. J Med Genet 52, 493-497. Lemmers, C., Michel, D., Lane-Guermonprez, L., Delgrossi, M. H., Medina, E., Arsanto, J. P. and Le Bivic, A. (2004). CRB3 binds directly to Par6 and regulates the morphogenesis of the tight junctions in mammalian epithelial cells. Mol Biol Cell 15, 1324-1333. Leonard, J. R., Klocke, B. J., D'Sa, C., Flavell, R. A. and Roth, K. A. (2002). Strain- dependent neurodevelopmental abnormalities in caspase-3-deficient mice. J Neuropathol Exp Neurol 61, 673-677. Li, G., Flodby, P., Luo, J., Kage, H., Sipos, A., Gao, D., Ji, Y., Beard, L. L., Marconett, C. N., DeMaio, L., et al. (2014). Knockout mice reveal key roles for claudin 18 in alveolar barrier properties and fluid homeostasis. Am J Respir Cell Mol Biol 51, 210-222. Li, Z., Ren, A., Zhang, L., Ye, R., Li, S., Zheng, J., Hong, S., Wang, T. and Li, Z. (2006). Extremely high prevalence of neural tube defects in a 4-county area in Shanxi Province, China. Birth Defects Res A Clin Mol Teratol 76, 237-240. Loh, Y. H., Christoffels, A., Brenner, S., Hunziker, W. and Venkatesh, B. (2004). Extensive expansion of the claudin gene family in the teleost fish, Fugu rubripes. Genome Res 14, 1248-1257. Lohrberg, D., Krause, E., Schumann, M., Piontek, J., Winkler, L., Blasig, I. E. and Haseloff, R. F. (2009). A strategy for enrichment of claudins based on their affinity to Clostridium perfringens enterotoxin. BMC molecular biology 10, 61. Mak, L. L. (1978). Ultrastructural studies of amphibian neural fold fusion. Dev Biol 65, 435- 446. Marean, A., Graf, A., Zhang, Y. and Niswander, L. (2011). Folic acid supplementation can adversely affect murine neural tube closure and embryonic survival. Hum Mol Genet 20, 3678-3683. Martin-Padura, I., Lostaglio, S., Schneemann, M., Williams, L., Romano, M., Fruscella, P., Panzeri, C., Stoppacciaro, A., Ruco, L., Villa, A., et al. (1998). Junctional adhesion molecule, a novel member of the immunoglobulin superfamily that distributes at intercellular junctions and modulates monocyte transmigration. J Cell Biol 142, 117-127. Matsumoto, K., Imasato, M., Yamazaki, Y., Tanaka, H., Watanabe, M., Eguchi, H., Nagano, H., Hikita, H., Tatsumi, T., Takehara, T., et al. (2014). Claudin 2 deficiency reduces bile flow and increases susceptibility to cholesterol gallstone disease in mice. Gastroenterology 147, 1134-1145 e1110. Mazaud-Guittot, S., Meugnier, E., Pesenti, S., Wu, X., Vidal, H., Gow, A. and Le Magueresse-Battistoni, B. (2010). Claudin 11 deficiency in mice results in loss of the Sertoli cell epithelial phenotype in the testis. Biol Reprod 82, 202-213.

193

McLaughlin, M. E., Kruger, G. M., Slocum, K. L., Crowley, D., Michaud, N. A., Huang, J., Magendantz, M. and Jacks, T. (2007). The Nf2 tumor suppressor regulates cell-cell adhesion during tissue fusion. Proc Natl Acad Sci U.S.A. 104, 3261-3266. Merte, J., Jensen, D., Wright, K., Sarsfield, S., Wang, Y., Schekman, R. and Ginty, D. D. (2010). Sec24b selectively sorts Vangl2 to regulate planar cell polarity during neural tube closure. Nat Cell Biol. 12, 41-46. McShane, S. G., Mole, M. A., Savery, D., Greene, N. D., Tam, P. P. and Copp, A. J. (2015). Cellular basis of neuroepithelial bending during mouse spinal neural tube closure. Dev Biol 404, 113-124. Milatz, S., Piontek, J., Hempel, C., Meoli, L., Grohe, C., Fromm, A., Lee, I. M., El-Athman, R. and Gunzel, D. (2017). Tight junction strand formation by claudin-10 isoforms and claudin-10a/-10b chimeras. Ann N Y Acad Sci. Milatz, S., Piontek, J., Schulzke, J. D., Blasig, I. E., Fromm, M. and Gunzel, D. (2015). Probing the cis-arrangement of prototype tight junction proteins claudin-1 and claudin-3. Biochem J 468, 449-458. Miyamoto, T., Morita, K., Takemoto, D., Takeuchi, K., Kitano, Y., Miyakawa, T., Nakayama, K., Okamura, Y., Sasaki, H., Miyachi, Y., et al. (2005). Tight junctions in Schwann cells of peripheral myelinated axons: a lesson from claudin-19-deficient mice. J Cell Biol 169, 527-538. Monteiro, A. C., Sumagin, R., Rankin, C. R., Leoni, G., Mina, M. J., Reiter, D. M., Stehle, T., Dermody, T. S., Schaefer, S. A., Hall, R. A., et al. (2013). JAM-A associates with ZO-2, afadin, and PDZ-GEF1 to activate Rap2c and regulate epithelial barrier function. Mol Biol Cell 24, 2849-2860. Morita, H., Kajiura-Kobayashi, H., Takagi, C., Yamamoto, T. S., Nonaka, S. and Ueno, N. (2012). Cell movements of the deep layer of non-neural ectoderm underlie complete neural tube closure in Xenopus. Development 139, 1417-1426. Morita, K., Sasaki, H., Furuse, M. and Tsukita, S. (1999). Endothelial claudin: claudin- 5/TMVCF constitutes tight junction strands in endothelial cells. J Cell Biol 147, 185-194. Moriwaki, K., Tsukita, S. and Furuse, M. (2007). Tight junctions containing claudin 4 and 6 are essential for blastocyst formation in preimplantation mouse embryos. Dev Biol 312, 509-522. Moury, J. D. and Schoenwolf, G. C. (1995). Cooperative model of epithelial shaping and bending during avian neurulation: autonomous movements of the neural plate, autonomous movements of the epidermis, and interactions in the neural plate/epidermis transition zone. Dev Dyn 204, 323-337. Mukendi, C., Dean, N., Lala, R., Smith, J., Bronner, M. E. and Nikitina, N. V. (2016). Evolution of the vertebrate claudin gene family: insights from a basal vertebrate, the sea lamprey. Int J Dev Bioll 60, 39-51. Murdoch, J. N., Damrau, C., Paudyal, A., Bogani, D., Wells, S., Greene, N. D., Stanier, P. and Copp, A. J. (2014). Genetic interactions between planar cell polarity genes cause diverse neural tube defects in mice. Dis Model Mech 7, 1153-1163. Murdoch, J. N., Rachel, R. A., Shah, S., Beermann, F., Stanier, P., Mason, C. A. and Copp, A. J. (2001). Circletail, a new mouse mutant with severe neural tube defects: chromosomal localization and interaction with the loop-tail mutation. Genomics 78, 55- 63.

194

Muto, S., Hata, M., Taniguchi, J., Tsuruoka, S., Moriwaki, K., Saitou, M., Furuse, K., Sasaki, H., Fujimura, A., Imai, M., et al. (2010). Claudin-2-deficient mice are defective in the leaky and cation-selective paracellular permeability properties of renal proximal tubules. Proc Natl Acad Sci U.S.A. 107, 8011-8016. Naeem, M., Hussain, S. and Akhtar, N. (2011). Mutation in the tight-junction gene claudin 19 (CLDN19) and familial hypomagnesemia, hypercalciuria, nephrocalcinosis (FHHNC) and severe ocular disease. Am J Nephrol 34, 241-248. Nagai, H., Lin, M. C. and Sheng, G. (2011). A modified cornish pasty method for ex ovo culture of the chick embryo. Genesis 49, 46-52. Nagai, T., Aruga, J., Minowa, O., Sugimoto, T., Ohno, Y., Noda, T. and Mikoshiba, K. (2000). Zic2 regulates the kinetics of neurulation. Proc Natl Acad Sci U.S.A. 97, 1618- 1623. Nakano, Y., Kim, S. H., Kim, H. M., Sanneman, J. D., Zhang, Y., Smith, R. J., Marcus, D. C., Wangemann, P., Nessler, R. A. and Banfi, B. (2009). A claudin-9-based ion permeability barrier is essential for hearing. PLoS Genet 5, e1000610. Nakatsu, T., Uwabe, C. and Shiota, K. (2000). Neural tube closure in humans initiates at multiple sites: evidence from human embryos and implications for the pathogenesis of neural tube defects. Anat Embryol (Berl) 201, 455-466. Nanni, L., Ming, J. E., Bocian, M., Steinhaus, K., Bianchi, D. W., Die-Smulders, C., Giannotti, A., Imaizumi, K., Jones, K. L., Campo, M. D., et al. (1999). The mutational spectrum of the sonic hedgehog gene in holoprosencephaly: SHH mutations cause a significant proportion of autosomal dominant holoprosencephaly. Hum Mol Genet 8, 2479-2488. National Center on Birth Defects and Developmental Disabilities, C. f. D. C. a. P. (2016). Folic Acid: Birth Defects COUNT. Nave, K. A. and Werner, H. B. (2014). Myelination of the nervous system: mechanisms and functions. Annu Rev Cell Dev Biol 30, 503-533. Nelson, K. S., Furuse, M. and Beitel, G. J. (2010). The Drosophila Claudin Kune-kune is required for septate junction organization and tracheal tube size control. Genetics 185, 831-839. Nikolopoulou, E., Galea, G. L., Rolo, A., Greene, N. D. and Copp, A. J. (2017). Neural tube closure: cellular, molecular and biomechanical mechanisms. Development 144, 552-566. Nishimura, T., Honda, H. and Takeichi, M. (2012). Planar cell polarity links axes of spatial dynamics in neural-tube closure. Cell 149, 1084-1097. Nitta, T., Hata, M., Gotoh, S., Seo, Y., Sasaki, H., Hashimoto, N., Furuse, M. and Tsukita, S. (2003). Size-selective loosening of the blood-brain barrier in claudin-5-deficient mice. J Cell Biol 161, 653-660. Nunbhakdi-Craig, V., Machleidt, T., Ogris, E., Bellotto, D., White, C. L., 3rd and Sontag, E. (2002). Protein phosphatase 2A associates with and regulates atypical PKC and the epithelial tight junction complex. J Cell Biol 158, 967-978. O'Rahilly, R. and Muller, F. (1994). Neurulation in the normal human embryo. Ciba Foundation symposium 181, 70-82; discussion 82-79. O'Rahilly, R. and Muller, F. (2002). The two sites of fusion of the neural folds and the two neuropores in the human embryo. Teratology 65, 162-170. Ooshio, T., Kobayashi, R., Ikeda, W., Miyata, M., Fukumoto, Y., Matsuzawa, N., Ogita, H. and Takai, Y. (2010). Involvement of the interaction of afadin with ZO-1 in the

195

formation of tight junctions in Madin-Darby canine kidney cells. J Biol Chem 285, 5003- 5012. Ossipova, O., Kim, K., Lake, B. B., Itoh, K., Ioannou, A. and Sokol, S. Y. (2014). Role of Rab11 in planar cell polarity and apical constriction during vertebrate neural tube closure. Nat Commun 5, 3734. Ossipova, O., Kim, K. and Sokol, S. Y. (2015). Planar polarization of Vangl2 in the vertebrate neural plate is controlled by Wnt and Myosin II signaling. Biol Open 4, 722-730. Osterhues, A., Ali, N. S. and Michels, K. B. (2013). The role of folic acid fortification in neural tube defects: a review. Crit Rev Food Sci Nutr 53, 1180-1190. Pai, Y. J., Abdullah, N. L., Mohd-Zin, S. W., Mohammed, R. S., Rolo, A., Greene, N. D., Abdul-Aziz, N. M. and Copp, A. J. (2012). Epithelial fusion during neural tube morphogenesis. Birth Defects Res A Clin Mol Teratol 94, 817-823. Palazzo, A. F., Cook, T. A., Alberts, A. S. and Gundersen, G. G. (2001a). mDia mediates Rho-regulated formation and orientation of stable microtubules. Nat Cell Biol 3, 723-729. Palazzo, A. F., Joseph, H. L., Chen, Y. J., Dujardin, D. L., Alberts, A. S., Pfister, K. K., Vallee, R. B. and Gundersen, G. G. (2001b). Cdc42, dynein, and dynactin regulate MTOC reorientation independent of Rho-regulated microtubule stabilization. Curr Biol 11, 1536-1541. Pangilinan, F., Molloy, A. M., Mills, J. L., Troendle, J. F., Parle-McDermott, A., Signore, C., O'Leary, V. B., Chines, P., Seay, J. M., Geiler-Samerotte, K., et al. (2012). Evaluation of common genetic variants in 82 candidate genes as risk factors for neural tube defects. BMC Med Genet 13, 62. Patten, I. and Placzek, M. (2002). Opponent activities of Shh and BMP signaling during floor plate induction in vivo. Curr Biol 12, 47-52. Paudyal, A., Damrau, C., Patterson, V. L., Ermakov, A., Formstone, C., Lalanne, Z., Wells, S., Lu, X., Norris, D. P., Dean, C. H., et al. (2010). The novel mouse mutant, chuzhoi, has disruption of Ptk7 protein and exhibits defects in neural tube, heart and lung development and abnormal planar cell polarity in the ear. BMC Dev Biol 10, 87. Persidsky, Y., Heilman, D., Haorah, J., Zelivyanskaya, M., Persidsky, R., Weber, G. A., Shimokawa, H., Kaibuchi, K. and Ikezu, T. (2006). Rho-mediated regulation of tight junctions during monocyte migration across the blood-brain barrier in HIV-1 encephalitis (HIVE). Blood 107, 4770-4780. Piontek, A., Rossa, J., Protze, J., Wolburg, H., Hempel, C., Gunzel, D., Krause, G. and Piontek, J. (2017a). Polar and charged extracellular residues conserved among barrier- forming claudins contribute to tight junction strand formation. Ann N Y Acad Sci 1397, 143-156. Piontek, A., Rossa, J., Protze, J., Wolburg, H., Hempel, C., Gunzel, D., Krause, G. and Piontek, J. (2017b). Polar and charged extracellular residues conserved among barrier- forming claudins contribute to tight junction strand formation. Ann N Y Acad Sci. Piontek, J., Winkler, L., Wolburg, H., Muller, S. L., Zuleger, N., Piehl, C., Wiesner, B., Krause, G. and Blasig, I. E. (2008). Formation of tight junction: determinants of homophilic interaction between classic claudins. FASEB J 22, 146-158. Pisam, M. and Ripoche, P. (1976). Redistribution of surface macromolecules in dissociated epithelial cells. J Cell Biol 71, 907-920. Poulain, F. E. and Yost, H. J. (2015). Heparan sulfate proteoglycans: a sugar code for vertebrate development? Development 142, 3456-3467.

196

Protze, J., Eichner, M., Piontek, A., Dinter, S., Rossa, J., Blecharz, K. G., Vajkoczy, P., Piontek, J. and Krause, G. (2015). Directed structural modification of Clostridium perfringens enterotoxin to enhance binding to claudin-5. Cell Mol Life Sci 72, 1417-1432. Pryor, S. E., Massa, V., Savery, D., Greene, N. D. and Copp, A. J. (2012). Convergent extension analysis in mouse whole embryo culture. Methods Mol Biol 839, 133-146. Pulimeno, P., Paschoud, S. and Citi, S. (2011). A role for ZO-1 and PLEKHA7 in recruiting paracingulin to tight and adherens junctions of epithelial cells. J Biol Chem 286, 16743- 16750. Pyrgaki, C., Liu, A. and Niswander, L. (2011). Grainyhead-like 2 regulates neural tube closure and adhesion molecule expression during neural fold fusion. Dev Biol 353, 38-49. Qiu, R. G., Abo, A. and Steven Martin, G. (2000). A human homolog of the C. elegans polarity determinant Par-6 links Rac and Cdc42 to PKCzeta signaling and cell transformation. Curr Biol 10, 697-707. Quiros, M. and Nusrat, A. (2014). RhoGTPases, actomyosin signaling and regulation of the epithelial Apical Junctional Complex. Semin Cell Dev Biol 36, 194-203. Raleigh, D. R., Marchiando, A. M., Zhang, Y., Shen, L., Sasaki, H., Wang, Y., Long, M. and Turner, J. R. (2010). Tight junction-associated MARVEL proteins marveld3, tricellulin, and occludin have distinct but overlapping functions. Mol Biol Cell 21, 1200- 1213. Ray, H. J. and Niswander, L. (2012). Mechanisms of tissue fusion during development. Development 139, 1701-1711. Ray, H. J. and Niswander, L. A. (2016). Grainyhead-like 2 downstream targets act to suppress epithelial-to-mesenchymal transition during neural tube closure. Development 143, 1192- 1204. Rehder, D., Iden, S., Nasdala, I., Wegener, J., Brickwedde, M. K., Vestweber, D. and Ebnet, K. (2006). Junctional adhesion molecule-a participates in the formation of apicobasal polarity through different domains. Exp Cell Res 312, 3389-3403. Riazuddin, S., Ahmed, Z. M., Fanning, A. S., Lagziel, A., Kitajiri, S., Ramzan, K., Khan, S. N., Chattaraj, P., Friedman, P. L., Anderson, J. M., et al. (2006). Tricellulin is a tight- junction protein necessary for hearing. Am J Hum Genet 79, 1040-1051. Rifat, Y., Parekh, V., Wilanowski, T., Hislop, N. R., Auden, A., Ting, S. B., Cunningham, J. M. and Jane, S. M. (2010). Regional neural tube closure defined by the Grainy head-like transcription factors. Dev Biol 345, 237-245. Robertson, S. L., Smedley, J. G., 3rd and McClane, B. A. (2010). Identification of a claudin-4 residue important for mediating the host cell binding and action of Clostridium perfringens enterotoxin. Infect Immun 78, 505-517. Robertson, S. L., Smedley, J. G., 3rd, Singh, U., Chakrabarti, G., Van Itallie, C. M., Anderson, J. M. and McClane, B. A. (2007). Compositional and stoichiometric analysis of Clostridium perfringens enterotoxin complexes in Caco-2 cells and claudin 4 fibroblast transfectants. Cell Microbiol 9, 2734-2755. Robinson, A., Escuin, S., Doudney, K., Vekemans, M., Stevenson, R. E., Greene, N. D., Copp, A. J. and Stanier, P. (2012). Mutations in the planar cell polarity genes CELSR1 and SCRIB are associated with the severe neural tube defect craniorachischisis. Hum Mutat 33, 440-447. Rodriguez-Boulan, E., Kreitzer, G. and Musch, A. (2005). Organization of vesicular trafficking in epithelia. Nature reviews. Molecular cell biology 6, 233-247.

197

Rodriguez-Gallardo, L., Climent, V., Garcia-Martinez, V., Schoenwolf, G. C. and Alvarez, I. S. (1997). Targeted over-expression of FGF in chick embryos induces formation of ectopic neural cells. Int J Dev Bioll 41, 715-723. Rogers, C. D., Ferzli, G. S. and Casey, E. S. (2011). The response of early neural genes to FGF signaling or inhibition of BMP indicate the absence of a conserved neural induction module. BMC Dev Biol 11, 74. Roh, M. H., Fan, S., Liu, C. J. and Margolis, B. (2003). The Crumbs3-Pals1 complex participates in the establishment of polarity in mammalian epithelial cells. J Cell Sci 116, 2895-2906. Roh, M. H., Liu, C. J., Laurinec, S. and Margolis, B. (2002a). The carboxyl terminus of zona occludens-3 binds and recruits a mammalian homologue of discs lost to tight junctions. J Biol Chem 277, 27501-27509. Roh, M. H., Makarova, O., Liu, C. J., Shin, K., Lee, S., Laurinec, S., Goyal, M., Wiggins, R. and Margolis, B. (2002b). The Maguk protein, Pals1, functions as an adapter, linking mammalian homologues of Crumbs and Discs Lost. J Cell Biol 157, 161-172. Rolo, A., Savery, D., Escuin, S., de Castro, S. C., Armer, H. E., Munro, P. M., Mole, M. A., Greene, N. D. and Copp, A. J. (2016). Regulation of cell protrusions by small GTPases during fusion of the neural folds. Elife 5, e13273. Rolo, A., Skoglund, P. and Keller, R. (2009). Morphogenetic movements driving neural tube closure in Xenopus require myosin IIB. Dev Biol 327, 327-338. Rosenthal, R., Gunzel, D., Theune, D., Czichos, C., Schulzke, J. D. and Fromm, M. (2017). Water channels and barriers formed by claudins. Ann N Y Acad Sci 1397, 100-109. Ross, A. J., May-Simera, H., Eichers, E. R., Kai, M., Jagger, D. J., Leitch, C. C., Chapple, J. P., Munro, P. M., Tan, P. L., Phillips, H. M., Leroux, M. R., Henderson, D. J., Murdoch, J. N., Copp, A. J., Eliot, M. M., Lupski, J. K., Kemp, D. T., Dollfus, H., Tada, M., Katsanis, N., Forge, A. and Beales, P. L. (2005) Disruption of Bardet-Biedle syndrome ciliary proteins perturbs planar cell polarity in vertebrates. Nat Genet 10, 1135- 1140. Rossa, J., Ploeger, C., Vorreiter, F., Saleh, T., Protze, J., Gunzel, D., Wolburg, H., Krause, G. and Piontek, J. (2014). Claudin-3 and claudin-5 protein folding and assembly into the tight junction are controlled by non-conserved residues in the transmembrane 3 (TM3) and extracellular loop 2 (ECL2) segments. J Biol Chem 289, 7641-7653. Saitou, M., Fujimoto, K., Doi, Y., Itoh, M., Fujimoto, T., Furuse, M., Takano, H., Noda, T. and Tsukita, S. (1998). Occludin-deficient embryonic stem cells can differentiate into polarized epithelial cells bearing tight junctions. J Cell Biol 141, 397-408. Saitou, M., Furuse, M., Sasaki, H., Schulzke, J. D., Fromm, M., Takano, H., Noda, T. and Tsukita, S. (2000). Complex phenotype of mice lacking occludin, a component of tight junction strands. Mol Biol Cell 11, 4131-4142. Sanchez-Pulido, L., Martin-Belmonte, F., Valencia, A. and Alonso, M. A. (2002). MARVEL: a conserved domain involved in membrane apposition events. Trends Biochem Sci 27, 599-601. Sausedo, R. A., Smith, J. L. and Schoenwolf, G. C. (1997). Role of nonrandomly oriented cell division in shaping and bending of the neural plate. J Comp Neurol 381, 473-488. Schoenwolf, G. C. (1979). Observations on closure of the neuropores in the chick embryo. Am J Anat 155, 445-465.

198

Schoenwolf, G. C. (1988). Microsurgical analyses of avian neurulation: separation of medial and lateral tissues. J Comp Neurol 276, 498-507. Schoenwolf, G. C. (1991a). Cell movements driving neurulation in avian embryos. Dev Suppl Suppl 2, 157-168. Schoenwolf, G. C. (1991b). Cell movements driving neurulation in avian embryos. Development Suppl 2, 157-168. Schoenwolf, G. C. and Alvarez, I. S. (1989). Roles of neuroepithelial cell rearrangement and division in shaping of the avian neural plate. Development 106, 427-439. Schoenwolf, G. C. and Alvarez, I. S. (1991). Specification of neurepithelium and surface epithelium in avian transplantation chimeras. Development 112, 713-722. Schoenwolf, G. C., Folsom, D. and Moe, A. (1988). A reexamination of the role of microfilaments in neurulation in the chick embryo. Anat Rec 220, 87-102. Schoenwolf, G. C. and Franks, M. V. (1984). Quantitative analyses of changes in cell shapes during bending of the avian neural plate. Dev Biol 105, 257-272. Schoenwolf, G. C. and Powers, M. L. (1987). Shaping of the chick neuroepithelium during primary and secondary neurulation: role of cell elongation. Anat Rec 218, 182-195. Selleck, M. A. and Bronner-Fraser, M. (1995). Origins of the avian neural crest: the role of neural plate-epidermal interactions. Development 121, 525-538. Seller, M. J. (1995). Further evidence for an intermittent pattern of neural tube closure in humans. J Med Genet 32, 205-207. Seller, M. J. and Nevin, N. C. (1984). Periconceptional vitamin supplementation and the prevention of neural tube defects in south-east England and Northern Ireland. J Med Genet 21, 325-330. Senga, K., Mostov, K. E., Mitaka, T., Miyajima, A. and Tanimizu, N. (2012). Grainyhead- like 2 regulates epithelial morphogenesis by establishing functional tight junctions through the organization of a molecular network among claudin3, claudin4, and Rab25. Mol Biol Cell 23, 2845-2855. Severson, E. A. and Parkos, C. A. (2009). Structural determinants of Junctional Adhesion Molecule A (JAM-A) function and mechanisms of intracellular signaling. Curr Opin Cell Biol 21, 701-707. Shafer, M. E. R., Nguyen, A. H. T., Tremblay, M., Viala, S., Beland, M., Bertos, N. R., Park, M. and Bouchard, M. (2017). Lineage Specification from Prostate Progenitor Cells Requires Gata3-Dependent Mitotic Spindle Orientation. Stem Cell Reports 8, 1018- 1031. Shang, X., Lin, X. and Howell, S. B. (2014). Claudin-4 controls the receptor tyrosine kinase EphA2 pro-oncogenic switch through beta-catenin. Cell Commun Signal 12, 59. Sharma, S., Place, E., Lord, K., Leroy, B. P., Falk, M. J. and Pradhan, M. (2016). Claudin 19-based familial hypomagnesemia with hypercalciuria and nephrocalcinosis in a sibling pair. Clinical nephrology 85, 346-352. Shrestha, A. and McClane, B. A. (2013). Human claudin-8 and -14 are receptors capable of conveying the cytotoxic effects of Clostridium perfringens enterotoxin. mBio 4. Siddiqui, M., Sheikh, H., Tran, C. and Bruce, A. E. (2010). The tight junction component Claudin E is required for zebrafish epiboly. Dev Dyn 239, 715-722. Simard, A., Di Pietro, E. and Ryan, A. K. (2005). Gene expression pattern of Claudin-1 during chick embryogenesis. Gene Expr Patterns 5, 553-560.

199

Simske, J. S. (2013). Claudins reign: The claudin/EMP/PMP22/gamma channel protein family in C. elegans. Tissue barriers 1, e25502. Simske, J. S. and Hardin, J. (2011). Claudin family proteins in Caenorhabditis elegans. Methods Mol Biol 762, 147-169. Simske, J. S., Koppen, M., Sims, P., Hodgkin, J., Yonkof, A. and Hardin, J. (2003). The cell junction protein VAB-9 regulates adhesion and epidermal morphology in C. elegans. Nat Cell Biol 5, 619-625. Singh, U., Van Itallie, C. M., Mitic, L. L., Anderson, J. M. and McClane, B. A. (2000). CaCo-2 cells treated with Clostridium perfringens enterotoxin form multiple large complex species, one of which contains the tight junction protein occludin. J Biol Chem 275, 18407-18417. Smedley, J. G., 3rd and McClane, B. A. (2004). Fine mapping of the N-terminal cytotoxicity region of Clostridium perfringens enterotoxin by site-directed mutagenesis. Infect Immun 72, 6914-6923. Smith, J. L. and Schoenwolf, G. C. (1987). Cell cycle and neuroepithelial cell shape during bending of the chick neural plate. Anat Rec 218, 196-206. Smith, J. L. and Schoenwolf, G. C. (1988). Role of cell-cycle in regulating neuroepithelial cell shape during bending of the chick neural plate. Cell Tissue Res 252, 491-500. Smith, J. L. and Schoenwolf, G. C. (1991). Further evidence of extrinsic forces in bending of the neural plate. J Comp Neurol 307, 225-236. Smith, J. L., Schoenwolf, G. C. and Quan, J. (1994). Quantitative analyses of neuroepithelial cell shapes during bending of the mouse neural plate. J Comp Neurol 342, 144-151. Smithells, R. W., Nevin, N. C., Seller, M. J., Sheppard, S., Harris, R., Read, A. P., Fielding, D. W., Walker, S., Schorah, C. J. and Wild, J. (1983). Further experience of vitamin supplementation for prevention of neural tube defect recurrences. Lancet 1, 1027-1031. Smithells, R. W., Sheppard, S., Schorah, C. J., Seller, M. J., Nevin, N. C., Harris, R., Read, A. P. and Fielding, D. W. (2011). Apparent prevention of neural tube defects by periconceptional vitamin supplementation. 1981. Int J Epidemiol 40, 1146-1154. Soma, T., Chiba, H., Kato-Mori, Y., Wada, T., Yamashita, T., Kojima, T. and Sawada, N. (2004). Thr(207) of claudin-5 is involved in size-selective loosening of the endothelial barrier by cyclic AMP. Exp Cell Res 300, 202-212. Sonoda, N., Furuse, M., Sasaki, H., Yonemura, S., Katahira, J., Horiguchi, Y. and Tsukita, S. (1999). Clostridium perfringens enterotoxin fragment removes specific claudins from tight junction strands: Evidence for direct involvement of claudins in tight junction barrier. J Cell Biol 147, 195-204. Sourisseau, T., Georgiadis, A., Tsapara, A., Ali, R. R., Pestell, R., Matter, K. and Balda, M. S. (2006). Regulation of PCNA and cyclin D1 expression and epithelial morphogenesis by the ZO-1-regulated transcription factor ZONAB/DbpA. Mol Cell Biol 26, 2387-2398. Staehelin, L. A., Mukherjee, T. M. and Williams, A. W. (1969). Fine structure of frozen- etched tight junctions. Naturwissenschaften 56, 142. Stanton, P. G. (2016). Regulation of the blood-testis barrier. Semin Cell Dev Biol 59, 166-173. Steed, E., Rodrigues, N. T., Balda, M. S. and Matter, K. (2009). Identification of MarvelD3 as a tight junction-associated transmembrane protein of the occludin family. BMC Cell Biol 10, 95. Stern, C. D. (2005). Neural induction: old problem, new findings, yet more questions. Development 132, 2007-2021.

200

Stevenson, B. R., Anderson, J. M., Goodenough, D. A. and Mooseker, M. S. (1988). Tight junction structure and ZO-1 content are identical in two strains of Madin-Darby canine kidney cells which differ in transepithelial resistance. J Cell Biol 107, 2401-2408. Stevenson, B. R., Siliciano, J. D., Mooseker, M. S. and Goodenough, D. A. (1986). Identification of ZO-1: a high molecular weight polypeptide associated with the tight junction (zonula occludens) in a variety of epithelia. J Cell Biol 103, 755-766. Stiefel, D., Shibata, T., Meuli, M., Duffy, P. G. and Copp, A. J. (2003) Tethering of the spinal cord in mouse fetuses and neonates with spinal bifida. J Neurosurg 99, 206-213. Streit, A., Berliner, A. J., Papanayotou, C., Sirulnik, A. and Stern, C. D. (2000). Initiation of neural induction by FGF signalling before gastrulation. Nature 406, 74-78. Streit, A., Lee, K. J., Woo, I., Roberts, C., Jessell, T. M. and Stern, C. D. (1998). Chordin regulates primitive streak development and the stability of induced neural cells, but is not sufficient for neural induction in the chick embryo. Development 125, 507-519. Sugihara, K., Nakatsuji, N., Nakamura, K., Nakao, K., Hashimoto, R., Otani, H., Sakagami, H., Kondo, H., Nozawa, S., Aiba, A., et al. (1998). Rac1 is required for the formation of three germ layers during gastrulation. Oncogene 17, 3427-3433. Sun, J., Wang, X., Li, C. and Mao, B. (2015). Xenopus Claudin-6 is required for embryonic pronephros morphogenesis and terminal differentiation. Biochem Biophys Res Commun 462, 178-183. Suzuki, H., Nishizawa, T., Tani, K., Yamazaki, Y., Tamura, A., Ishitani, R., Dohmae, N., Tsukita, S., Nureki, O. and Fujiyoshi, Y. (2014). Crystal structure of a claudin provides insight into the architecture of tight junctions. Science 344, 304-307. Suzuki, H., Tani, K., Tamura, A., Tsukita, S. and Fujiyoshi, Y. (2015). Model for the architecture of claudin-based paracellular ion channels through tight junctions. J Mol Biol 427, 291-297. Suzuki, M., Morita, H. and Ueno, N. (2012). Molecular mechanisms of cell shape changes that contribute to vertebrate neural tube closure. Dev Growth Differ 54, 266-276. Swager, S. A., Delfin, D. A., Rastogi, N., Wang, H., Canan, B. D., Fedorov, V. V., Mohler, P. J., Kilic, A., Higgins, R. S. D., Ziolo, M. T., et al. (2015). Claudin-5 levels are reduced from multiple cell types in human failing hearts and are associated with mislocalization of ephrin-B1. Cardiovasc Pathol 24, 160-167. Tabuse, Y., Izumi, Y., Piano, F., Kemphues, K. J., Miwa, J. and Ohno, S. (1998). Atypical protein kinase C cooperates with PAR-3 to establish embryonic polarity in Caenorhabditis elegans. Development 125, 3607-3614. Tamura, A., Hayashi, H., Imasato, M., Yamazaki, Y., Hagiwara, A., Wada, M., Noda, T., Watanabe, M., Suzuki, Y. and Tsukita, S. (2011). Loss of claudin-15, but not claudin- 2, causes Na+ deficiency and glucose malabsorption in mouse small intestine. Gastroenterology 140, 913-923. Tamura, A., Kitano, Y., Hata, M., Katsuno, T., Moriwaki, K., Sasaki, H., Hayashi, H., Suzuki, Y., Noda, T., Furuse, M., et al. (2008). Megaintestine in claudin-15-deficient mice. Gastroenterology 134, 523-534. Tanaka, H., Takechi, M., Kiyonari, H., Shioi, G., Tamura, A. and Tsukita, S. (2015). Intestinal deletion of Claudin-7 enhances paracellular organic solute flux and initiates colonic inflammation in mice. Gut 64, 1529-1538. Tanaka, M., Kamata, R. and Sakai, R. (2005a). EphA2 phosphorylates the cytoplasmic tail of Claudin-4 and mediates paracellular permeability. J Biol Chem 280, 42375-42382.

201

Tanaka, M., Kamata, R. and Sakai, R. (2005b). Phosphorylation of ephrin-B1 via the interaction with claudin following cell-cell contact formation. Embo J 24, 3700-3711. Tatum, R., Zhang, Y., Salleng, K., Lu, Z., Lin, J. J., Lu, Q., Jeansonne, B. G., Ding, L. and Chen, Y. H. (2010). Renal salt wasting and chronic dehydration in claudin-7-deficient mice. Am J Physiol Renal Physiol 298, F24-34. Tawk, M., Araya, C., Lyons, D. A., Reugels, A. M., Girdler, G. C., Bayley, P. R., Hyde, D. R., Tada, M. and Clarke, J. D. (2007). A mirror-symmetric cell division that orchestrates neuroepithelial morphogenesis. Nature 446, 797-800. Terry, S. J., Zihni, C., Elbediwy, A., Vitiello, E., Leefa Chong San, I. V., Balda, M. S. and Matter, K. (2011). Spatially restricted activation of RhoA signalling at epithelial junctions by p114RhoGEF drives junction formation and morphogenesis. Nat Cell Biol 13, 159-166. Torban, E., Patenaude, A. M., Leclerc, S., Rakowiecki, S., Gauthier, S., Andelfinger, G., Epstein, D. J. and Gros, P. (2008). Genetic interaction between members of the Vangl family causes neural tube defects in mice. Proc Natl Acad Sci USA 105, 3449-3454. Tsukita, S., Furuse, M. and Itoh, M. (2001). Multifunctional strands in tight junctions. Nature reviews. Mol Cell Biol 2, 285-293. Turksen, K. and Troy, T. C. (2002). Permeability barrier dysfunction in transgenic mice overexpressing claudin 6. Development 129, 1775-1784. Ullmer, C., Schmuck, K., Figge, A. and Lubbert, H. (1998). Cloning and characterization of MUPP1, a novel PDZ domain protein. FEBS letters 424, 63-68. Umeda, K., Ikenouchi, J., Katahira-Tayama, S., Furuse, K., Sasaki, H., Nakayama, M., Matsui, T., Tsukita, S., Furuse, M. and Tsukita, S. (2006). ZO-1 and ZO-2 independently determine where claudins are polymerized in tight-junction strand formation. Cell 126, 741-754. Van Allen, M. I., Kalousek, D. K., Chernoff, G. F., Juriloff, D., Harris, M., McGillivray, B. C., Yong, S. L., Langlois, S., MacLeod, P. M., Chitayat, D., et al. (1993). Evidence for multi-site closure of the neural tube in humans. Am J Med Genet 47, 723-743. van der Linden, I. J., Afman, L. A., Heil, S. G. and Blom, H. J. (2006). Genetic variation in genes of folate metabolism and neural-tube defect risk. Proc Nutr Soc 65, 204-215. Van Itallie, C., Rahner, C. and Anderson, J. M. (2001). Regulated expression of claudin-4 decreases paracellular conductance through a selective decrease in sodium permeability. J Clin Invest 107, 1319-1327. Van Itallie, C. M. and Anderson, J. M. (1997). Occludin confers adhesiveness when expressed in fibroblasts. J Cell Sci 110 ( Pt 9), 1113-1121. Van Itallie, C. M. and Anderson, J. M. (2014). Architecture of tight junctions and principles of molecular composition. Semin Cell Dev Biol 36, 157-165. Van Itallie, C. M., Aponte, A., Tietgens, A. J., Gucek, M., Fredriksson, K. and Anderson, J. M. (2013). The N and C termini of ZO-1 are surrounded by distinct proteins and functional protein networks. J Biol Chem 288, 13775-13788. Van Itallie, C. M., Betts, L., Smedley, J. G., 3rd, McClane, B. A. and Anderson, J. M. (2008a). Structure of the claudin-binding domain of Clostridium perfringens enterotoxin. J Biol Chem 283, 268-274. Van Itallie, C. M., Colegio, O. R. and Anderson, J. M. (2004). The cytoplasmic tails of claudins can influence tight junction barrier properties through effects on protein stability. J Membr Biol 199, 29-38.

202

Van Itallie, C. M., Gambling, T. M., Carson, J. L. and Anderson, J. M. (2005). Palmitoylation of claudins is required for efficient tight-junction localization. J Cell Sci 118, 1427-1436. Van Itallie, C. M., Holmes, J., Bridges, A., Gookin, J. L., Coccaro, M. R., Proctor, W., Colegio, O. R. and Anderson, J. M. (2008b). The density of small tight junction pores varies among cell types and is increased by expression of claudin-2. J Cell Sci 121, 298- 305. Van Itallie, C. M., Rogan, S., Yu, A., Vidal, L. S., Holmes, J. and Anderson, J. M. (2006). Two splice variants of claudin-10 in the kidney create paracellular pores with different ion selectivities. Am J Physiol Renal Physiol 291, F1288-1299. Van Straaten, H. W., Janssen, H. C., Peeters, M. C., Copp, A. J. and Hekking, J. W. (1996). Neural tube closure in the chick embryo is multiphasic. Dev Dyn 207, 309-318. van Straaten, H. W., Sieben, I. and Hekking, J. W. (2002). Multistep role for actin in initial closure of the mesencephalic neural groove in the chick embryo. Dev Dyn 224, 103-108. Veshnyakova, A., Piontek, J., Protze, J., Waziri, N., Heise, I. and Krause, G. (2012). Mechanism of Clostridium perfringens enterotoxin interaction with claudin-3/-4 protein suggests structural modifications of the toxin to target specific claudins. J Biol Chem 287, 1698-1708. Viswanathan, M., Treiman, K. A., Kish-Doto, J., Middleton, J. C., Coker-Schwimmer, E. J. and Nicholson, W. K. (2017). Folic Acid Supplementation for the Prevention of Neural Tube Defects: An Updated Evidence Report and Systematic Review for the US Preventive Services Task Force. Jama 317, 190-203. Voiculescu, O., Papanayotou, C. and Stern, C. D. (2008). Spatially and temporally controlled electroporation of early chick embryos. Nat Protoc 3, 419-426. Voiculescu, O. and Stern, C. D. (2017). Manipulating Gene Expression in the Chick Embryo. Methods Mol Biol 1565, 105-114. Volcik, K. A., Blanton, S. H., Kruzel, M. C., Townsend, I. T., Tyerman, G. H., Mier, R. J. and Northrup, H. (2002). Testing for genetic associations with the PAX gene family in a spina bifida population. Am J Med Genet 110, 195-202. Wada, M., Tamura, A., Takahashi, N. and Tsukita, S. (2013). Loss of claudins 2 and 15 from mice causes defects in paracellular Na+ flow and nutrient transport in gut and leads to death from malnutrition. Gastroenterology 144, 369-380. Waddington, C. H. (1932). Experiments on the development of chick and duck embryos cultivated in vitro. Phil. Trans. Roy. Soc. Lond. B. 221, 179-230. Wallingford, J. B. and Harland, R. M. (2002). Neural tube closure requires Dishevelled- dependent convergent extension of the midline. Development 129, 5815-5825. Wang, J., Hamblet, N. S., Mark, S., Dickinson, M. E., Brinkman, B. C., Segil, N., Fraser, S. E., Chen, P., Wallingford, J. B. and Wynshaw-Boris, A. (2006a). Dishevelled genes mediate a conserved mammalian PCP pathway to regulate convergent extension during neurulation. Development 133, 1767-1778. Wang, Y., Guo, N. and Nathans, J. (2006b). The role of Frizzled3 and Frizzled6 in neural tube closure and in the planar polarity of inner-ear sensory hair cells. J Neurosci 26, 2147- 2156. Wang, Y., Thekdi, N., Smallwood, P. M., Macke, J. P. and Nathans, J. (2002). Frizzled-3 is required for the development of major fiber tracts in the rostral CNS. J Neurosci 22, 8563-8573.

203

Waterman, R. E. (1976). Topographical changes along the neural fold associated with neurulation in the hamster and mouse. Am J Anat 146, 151-171. Wattenhofer, M., Reymond, A., Falciola, V., Charollais, A., Caille, D., Borel, C., Lyle, R., Estivill, X., Petersen, M. B., Meda, P., et al. (2005). Different mechanisms preclude mutant CLDN14 proteins from forming tight junctions in vitro. Hum Mutat 25, 543-549. Weber, S., Schneider, L., Peters, M., Misselwitz, J., Ronnefarth, G., Boswald, M., Bonzel, K. E., Seeman, T., Sulakova, T., Kuwertz-Broking, E., et al. (2001). Novel paracellin- 1 mutations in 25 families with familial hypomagnesemia with hypercalciuria and nephrocalcinosis. J Am Soc Nephrol 12, 1872-1881. Weil, M., Abeles, R., Nachmany, A., Gold, V. and Michael, E. (2004). Folic acid rescues nitric oxide-induced neural tube closure defects. Cell Death Differ 11, 361-363. Weil, M., Jacobson, M. D. and Raff, M. C. (1997). Is programmed cell death required for neural tube closure? Curr Biol 7, 281-284. Wells, C. D., Fawcett, J. P., Traweger, A., Yamanaka, Y., Goudreault, M., Elder, K., Kulkarni, S., Gish, G., Virag, C., Lim, C., et al. (2006). A Rich1/Amot complex regulates the Cdc42 GTPase and apical-polarity proteins in epithelial cells. Cell 125, 535- 548. Wen, H., Watry, D. D., Marcondes, M. C. and Fox, H. S. (2004). Selective decrease in paracellular conductance of tight junctions: role of the first extracellular domain of claudin-5. Mol Cell Biol 24, 8408-8417. Werth, M., Walentin, K., Aue, A., Schonheit, J., Wuebken, A., Pode-Shakked, N., Vilianovitch, L., Erdmann, B., Dekel, B., Bader, M., et al. (2010). The transcription factor grainyhead-like 2 regulates the molecular composition of the epithelial apical junctional complex. Development 137, 3835-3845. Wessely, O. and Tran, U. (2011). Xenopus pronephros development--past, present, and future. Pediatr Nephrol 26, 1545-1551. Westphal, J. K., Dorfel, M. J., Krug, S. M., Cording, J. D., Piontek, J., Blasig, I. E., Tauber, R., Fromm, M. and Huber, O. (2010). Tricellulin forms homomeric and heteromeric tight junctional complexes. Cell Mol Life Sci 67, 2057-2068. Wieckowski, E. U., Wnek, A. P. and McClane, B. A. (1994). Evidence that an approximately 50-kDa mammalian plasma membrane protein with receptor-like properties mediates the amphiphilicity of specifically bound Clostridium perfringens enterotoxin. J Biol Chem 269, 10838-10848. Wilcox, E. R., Burton, Q. L., Naz, S., Riazuddin, S., Smith, T. N., Ploplis, B., Belyantseva, I., Ben-Yosef, T., Liburd, N. A., Morell, R. J., et al. (2001). Mutations in the gene encoding tight junction claudin-14 cause autosomal recessive deafness DFNB29. Cell 104, 165-172. Will, C., Breiderhoff, T., Thumfart, J., Stuiver, M., Kopplin, K., Sommer, K., Gunzel, D., Querfeld, U., Meij, I. C., Shan, Q., et al. (2010). Targeted deletion of murine Cldn16 identifies extra- and intrarenal compensatory mechanisms of Ca2+ and Mg2+ wasting. Am J Physiol Renal Physiol 298, F1152-1161. Winkler, L., Gehring, C., Wenzel, A., Muller, S. L., Piehl, C., Krause, G., Blasig, I. E. and Piontek, J. (2009). Molecular determinants of the interaction between Clostridium perfringens enterotoxin fragments and claudin-3. J Biol Chem 284, 18863-18872.

204

Wittchen, E. S., Haskins, J. and Stevenson, B. R. (1999). Protein interactions at the tight junction. Actin has multiple binding partners, and ZO-1 forms independent complexes with ZO-2 and ZO-3. J Biol Chem 274, 35179-35185. Wong, V. and Gumbiner, B. M. (1997). A synthetic peptide corresponding to the extracellular domain of occludin perturbs the tight junction permeability barrier. J Cell Biol 136, 399- 409. Wray, C., Mao, Y., Pan, J., Chandrasena, A., Piasta, F. and Frank, J. A. (2009). Claudin-4 augments alveolar epithelial barrier function and is induced in acute lung injury. Am J Physiol Lung Cell Mol Physiol 297, L219-227. Wu, V. M., Schulte, J., Hirschi, A., Tepass, U. and Beitel, G. J. (2004). Sinuous is a Drosophila claudin required for septate junction organization and epithelial tube size control. J Cell Biol 164, 313-323. Xu, J., Kausalya, P. J., Phua, D. C., Ali, S. M., Hossain, Z. and Hunziker, W. (2008). Early embryonic lethality of mice lacking ZO-2, but Not ZO-3, reveals critical and nonredundant roles for individual zonula occludens proteins in mammalian development. Mol Cell Biol 28, 1669-1678. Yamaguti, P. M., dos Santos, P. A., Leal, B. S., Santana, V. B., Mazzeu, J. F., Acevedo, A. C. and Neves Fde, A. (2015). Identification of the first large deletion in the CLDN16 gene in a patient with FHHNC and late-onset of chronic kidney disease: case report. BMC Nephrol 16, 92. Yamamoto, M., Ramirez, S. H., Sato, S., Kiyota, T., Cerny, R. L., Kaibuchi, K., Persidsky, Y. and Ikezu, T. (2008). Phosphorylation of claudin-5 and occludin by rho kinase in brain endothelial cells. Am J Pathol 172, 521-533. Yao, F., Kausalya, J. P., Sia, Y. Y., Teo, A. S., Lee, W. H., Ong, A. G., Zhang, Z., Tan, J. H., Li, G., Bertrand, D., et al. (2015). Recurrent Fusion Genes in Gastric Cancer: CLDN18- ARHGAP26 Induces Loss of Epithelial Integrity. Cell Rep 12, 272-285. Yates, L. L., Schnatwinkel, C., Hazelwood, L., Chessum, L., Paudyal, A., Hilton, H., Romero, M. R., Wilde, J., Bogani, D., Sanderson, J., et al. (2013). Scribble is required for normal epithelial cell-cell contacts and lumen morphogenesis in the mammalian lung. Dev Biol 373, 267-280. Ybot-Gonzalez, P., Cogram, P., Gerrelli, D. and Copp, A. J. (2002). Sonic hedgehog and the molecular regulation of mouse neural tube closure. Development 129, 2507-2517. Ybot-Gonzalez, P. and Copp, A. J. (1999). Bending of the neural plate during mouse spinal neurulation is independent of actin microfilaments. Dev Dyn 215, 273-283. Ybot-Gonzalez, P., Gaston-Massuet, C., Girdler, G., Klingensmith, J., Arkell, R., Greene, N. D. and Copp, A. J. (2007a). Neural plate morphogenesis during mouse neurulation is regulated by antagonism of Bmp signalling. Development 134, 3203-3211. Ybot-Gonzalez, P., Savery, D., Gerrelli, D., Signore, M., Mitchell, C. E., Faux, C. H., Greene, N. D. and Copp, A. J. (2007b). Convergent extension, planar-cell-polarity signalling and initiation of mouse neural tube closure. Development 134, 789-799. Yip, G. W., Ferretti, P. and Copp, A. J. (2002). Heparan sulphate proteoglycans and spinal neurulation in the mouse embryo. Development 129, 2109-2119. Yu, A. S., Enck, A. H., Lencer, W. I. and Schneeberger, E. E. (2003). Claudin-8 expression in Madin-Darby canine kidney cells augments the paracellular barrier to cation permeation. J Biol Chem 278, 17350-17359.

205

Yuan, T., Pang, Q., Xing, X., Wang, X., Li, Y., Li, J., Wu, X., Li, M., Wang, O., Jiang, Y., et al. (2015). First report of a novel missense CLDN19 mutations causing familial hypomagnesemia with hypercalciuria and nephrocalcinosis in a Chinese family. Calcif Tissue Int 96, 265-273. Zaganjor, I., Sekkarie, A., Tsang, B. L., Williams, J., Razzaghi, H., Mulinare, J., Sniezek, J. E., Cannon, M. J. and Rosenthal, J. (2016). Describing the Prevalence of Neural Tube Defects Worldwide: A Systematic Literature Review. PloS one 11, e0151586. Zeissig, S., Burgel, N., Gunzel, D., Richter, J., Mankertz, J., Wahnschaffe, U., Kroesen, A. J., Zeitz, M., Fromm, M. and Schulzke, J. D. (2007). Changes in expression and distribution of claudin 2, 5 and 8 lead to discontinuous tight junctions and barrier dysfunction in active Crohn's disease. Gut 56, 61-72. Zhang, J., Hagopian-Donaldson, S., Serbedzija, G., Elsemore, J., Plehn-Dujowich, D., McMahon, A. P., Flavell, R. A. and Williams, T. (1996). Neural tube, skeletal and body wall defects in mice lacking transcription factor AP-2. Nature 381, 238-241. Zihni, C., Mills, C., Matter, K. and Balda, M. S. (2016). Tight junctions: from simple barriers to multifunctional molecular gates. Nature Rev. Mol Cell Biol 17, 564-580. Ziomek, C. A., Schulman, S. and Edidin, M. (1980). Redistribution of membrane proteins in isolated mouse intestinal epithelial cells. J Cell Biol 86, 849-857.

206