INCREASED FGF AND HH SIGNALING IMPAIRS CRANIOFACIAL AND LIMB

MORPHOGENESIS

by

LINNEA GWENDOLYN SCHMIDT

B.A. Gustavus Adolphus College, 2011

A thesis submitted to the

Faculty of the Graduate School of the

University of Colorado in partial fulfillment

of the requirements for the degree of

Doctor of Philosophy

Integrated Physiology Program

2017

This thesis for the Doctor of Philosophy degree by

Linnea Gwendolyn Schmidt

has been approved for the

Integrated Physiology Program

by

Joan Hooper, Chair

David Clouthier

Lee Niswander

Kathleen Connell

Jim McManaman

Trevor Williams, Advisor

Date: 12/15/2017

ii Schmidt, Linnea Gwendolyn (Ph.D., Integrated Physiology)

Increased FGF and HH Signaling Impairs Craniofacial and Limb Morphogenesis

Thesis directed by Professor Trevor Williams

ABSTRACT

Tight regulation of signaling pathways during embryogenesis is required for normal development. In contrast, aberrations in signaling frequently result in fetal and infant mortality and morbidity. While the head and limb have separate evolutionary origins and

dissimilar anatomy, many of the same genes and signaling pathways regulate craniofacial

and limb development. As such, craniofacial and limb defects are often found together in

genetic syndromes. The unique morphology of the head and limb is thus dependent upon the specific location, dosage, and timing of gene expression as well as genetic interactions between genes and signaling pathways. Using the mouse as a model system, the studies described in this thesis investigate the impact of increased FGF8 and HH signaling on

craniofacial and limb development when increased starting at different embryological

timepoints and from different tissues. The specific location, dosage, and timing of increased

FGF8 and HH signaling led to distinct phenotypes and genetic interactions with other genes

and signaling pathways. Together, these results contribute to the fields of embryogenesis,

morphogenesis, and human congenital defects, with implications for the evolution of

craniofacial and limb morphogenesis and therapeutic intervention in birth defects.

The form and content of this abstract are approved. I recommend its publication.

Approved: Trevor Williams

iii

Dedicated with love to my grandmothers, Bernice Olson Schmidt and Julia Andrew Gamon,

who taught me learning is a lifelong pursuit.

iv ACKNOWLEDGEMENTS

Many people contributed to the successful completion of this dissertation. First, I

would like to thank all the people who helped with the work presented here and offered

advice and assistance. I am grateful to the past and present members of the Williams lab,

particularly Eric Van Otterloo, Hong Li, Irene Choi, and Aftab Taiyab for their training and

expertise. I would also like to acknowledge the members of my advisory committee for their suggestions along the way. Finally, I would like to especially thank my advisor, Trevor

Williams, for his patient mentorship and guidance throughout this experience.

I would also like to extend my appreciation to the members and administrators of the

Integrated Physiology graduate program (formally Reproductive Sciences), with an especial thanks to Emily Busta, Ally Roof, and Sydney Coates for their unwavering support.

Additionally, I would like to thank my Gustavus professors for their dedication to teaching and mentoring as well as their encouragement in pursuing graduate school

Finally, all my gratitude to my family and friends that are outside of Colorado, but always inside my heart. I’m lucky to have both parents and grandparents who taught me the value in higher education and the pursuit of knowledge. Many thanks to my Minnesota

Gusties—particularly Rebecca Rasp and Paul Huff –for giving me somewhere to come

“home” to and Elisabeth Graeber for always giving me a great excuse for a European vacation. Finally thank you to my “Iowa Friends” -- Lena Thompson, Lisa Wehr Maves,

Micah Stevens, Rachel Olson, and Ashley Christner -- for being with me since the very beginning and molding me into the person I am today. Our friendship is my greatest accomplishment.

v TABLE OF CONTENTS

CHAPTER

I. THE GENETICS OF HEAD AND LIMB DEVELOPMENT...... 1

Introduction ...... 1

Craniofacial Development ...... 1

Human Craniofacial Defects ...... 2

Mouse Craniofacial Development ...... 8

Limb Development ...... 11

Human Limb Defects...... 12

Mouse Limb Development...... 12

Limb Patterning and Outgrowth ...... 14

Ossification in the Cranium and Limb ...... 18

Critical Signaling Pathways in Craniofacial and Limb Development ...... 21

FGF Signaling ...... 22

Hedgehog (HH) Signaling ...... 28

WNT Signaling ...... 35

BMP Signaling ...... 37

Layout of this Dissertation ...... 39

II. MATERIALS AND METHODS ...... 41

Mice ...... 41

vi Mouse Strains...... 41

Genotyping ...... 44

Skeletal Analysis ...... 45

Bone & Cartilage Staining...... 45

Cartilage Staining ...... 45

Section Histochemistry...... 46

In Situ Hybridization...... 46

Preparation of Probes...... 46

Wholemount In Situ Hybridization (WMISH) ...... 46

Section In Situ Hybridization ...... 48

Skin Barrier Analysis ...... 49

Toluidine Blue ...... 49

Alkaline Phosphatase Staining ...... 49

β-galactosidase Staining ...... 50

Whole Mount ...... 50

Sections ...... 50

RNA Quantification ...... 51

Realtime PCR (qRTPCR) ...... 51

RNAseq ...... 52

III.. CCTSI CLINICAL EXPERIENCE ...... 54

vii Craniosynostosis ...... 54

Interactions with Patients in Clinic ...... 55

Mice as a Model to Study Craniosynostosis ...... 60

Conclusions ...... 61

IV. INCREASED FGF8 SIGNALING SHIFTS CELL FATE FROM OSTEOGENIC

TO CHONDROGENIC IN THE DEVELOPING SKULL ...... 63

Introduction ...... 63

Results ...... 67

Generation of New Alleles for Differential Expression of Fgf8 ...... 67

Analysis of Msx2-Cre Expression Pattern During Embryonic Craniofacial

Development ...... 72

Moderately Increased Levels of Fgf8 Cause Craniosynostosis ...... 74

Histological Analysis of Craniosynostosis in MR26F8 Mice ...... 80

Abnormal Cartilage Replaces Intramembranous Bone at Higher Fgf8 Levels ...... 85

MCAGF8 Mutants Have Impaired Differentiation and Display Dysregulated WNT

Signaling ...... 97

Reducing Axin2 Gene Dosage Partially Rescues the Cranial Skeleton Phenotype.. 110

Summary ...... 118

Discussion ...... 119

Moderate Overexpression of Fgf8 Leads to Craniosynostosis ...... 119

viii Dose-dependent Response to Fgf8 Signaling in the Cranial Vault ...... 121

How Does Fgf8 signaling Provoke the Shift from Bone to Cartilage Formation? .. 122

V. INCREASED FGF8 SIGNALING DRAMATICALLY IMPAIRS OSSIFICATION

OF INTRAMEMBRANOUS, BUT NOT ENDOCHONDRAL BONES ...... 129

Introduction ...... 129

Results ...... 131

Osteocalcin-Cre ...... 131

Col2a1-Cre ...... 138

Creface ...... 143

Discussion ...... 159

VI. INCREASED HEDGEHOG SIGNALING FROM THE ECTODERM AND

MESENCHYME IMPAIRS CRANIOFACIAL AND LIMB DEVELOPMENT ..

...... 165

Introduction ...... 165

Results ...... 168

Mesenchymal Upregulation of HH Signaling Results in Cranial and Limb Defects 168

Ectodermal Upregulation of HH Signaling Results in Cranial and Limb Defects ... 194

Discussion ...... 219

VII. OUTLOOK ...... 232

Craniosynostosis as a Bone Growth Disorder ...... 232

ix FGF Signaling and Endochondral Ossification ...... 234

Spatiotemporal Dynamics ...... 239

Increased Hedgehog Signaling from the Ectoderm and Mesenchyme Impairs Craniofacial

and Limb Development...... 242

Genetic Interactions Between FGF8 and HH ...... 244

REFERENCES ...... 246

APPENDIX A ...... 283

Recipes and Abbreviations ...... 283

x LIST OF FIGURES

FIGURE

1. Overview of human facial development...... 4

2. Anatomy of the human cranial vault...... 6

3. Mouse craniofacial development...... 9

4. Bones and sutures of the mouse cranial vault...... 11

5. Mouse limb bud morphology...... 13

6. Limb signaling centers...... 15

7. Stages of bone and cartilage lineage cell differentiation...... 19

8. FGFR structure, signaling, and dysregulation...... 23

9. The hedgehog signaling pathway...... 29

10. Embryos without SHH do not develop recognizable heads...... 31

11. Craniofacial phenotypes of Wnt1-Cre; SmoGOF embryos...... 32

12. Canonical WNT signaling pathway...... 36

13. Schematic of the BMP signaling pathway...... 38

14. Detailed map of R26F8 and CAGF8 constructs...... 68

15. Conditional alleles modulate Fgf8b expression levels...... 70

16. Embryonic expression of Msx2-Cre during craniofacial development...... 73

17. MR26F8 mice exhibit cranial defects including coronal craniosynostosis...... 75

18. MR26F8 mice have striking cranial and limb defects...... 77

19. Fur defect in MR26F8 mice...... 78

20. MR26F8 mutants have delayed ossification, followed by over-ossification, of the

lambdoid suture...... 81

xi 21. Delayed ossification, followed by craniosynostosis, of the MR26F8 coronal suture...... 83

22. MCAGF8 mice exhibit cranial defects including loss of ossification...... 86

23. MCAGF8 limb Phenotypes...... 88

24. Cartilage replaces bone throughout the MCAGF8 cranial vault...... 90

25. Histological analysis indicates abnormal cartilage replaces intramembranous bone in

MCAGF8 cranial vault...... 94

26. MCAGF8 mice also exhibit epithelial defects...... 97

27. MCAGF8 cranial vault differentiation shifts from osteogenic to chondrogenic...... 98

28. Two-way comparisons of the RNAseq treatment groups...... 104

29. PCA of all three groups of RNAseq samples...... 105

30. Additional pathways altered in the MCAGF8 skull...... 111

31. Reduced Axin2 gene dosage improves cranial vault ossification...... 113

32. Axin2lacZ expression during early embryogenesis...... 116

33. Axin2lacZ expression at birth in the cranial vault...... 118

34. Osteocalcin-Cre expression from E14.5-E16.5...... 132

35. Gross Morphology of E18.5 OCAGF8 embryos...... 133

36. OR26F8 mice have craniofacial shape and skeletal defects...... 134

37. Intramembranous forming bones are more severely impacted by increased Fgf8 expression than endochondral forming bones...... 135

38. OCAGF8 ribs have patterning defects, but ossify...... 138

39. Embryonic expression of Col2a1-Cre...... 139

40. ColCAGF8 embryos have severe cartilage formation defects...... 140

41. E13.5 ColR26F8 and ColCAGF8 embryos have ectopic cranial cartilage formation. .. 142

xii 42. Characterization of AP-2CRE (Creface) transgene activity in mid-embryogenesis using

the R26R reporter strain...... 145

43. Creface transgene activity in skull and limb derivatives...... 147

44. FaceCAGF8 can survive past birth, but are grow slower than their littermate controls.148

45. FaceCAGF8 mice exhibit craniofacial shape and skeletal defects...... 149

46. Post-natal cranial skeletal staining...... 151

47. Cartilage staining in the E14.5 FaceCAGF8 head...... 152

48. Limb defects in FaceCAGF8 mice...... 153

49. FaceCAGF8 E12.5 embryos have decreased WNT Signaling in the FNP...... 156

50. Decreased Osterix expression in the FaceCAGF8 frontal bone...... 157

51. PtcLacZ expression in E12.5 control and FaceCAGF8 embryos...... 158

52. SmoM2 allele in mice...... 168

53. Gross morphology of FaceSMO heads...... 170

54. FaceSMO mice exhibit shortened snout and hypertelorism...... 171

55. FaceSMO neonates develop distended abdomens and lack milk sacs...... 173

56. FaceSMO neonates have cranial and cerebral defects...... 174

57. FaceSMOs exhibit multiple limb patterning defects...... 177

58. Increased hedgehog signaling in E12.5 FaceSMO embryos corresponds with regions of

Creface expression...... 179

59. Hedgehog signaling is increased in FaceSMO neonates...... 181

60. Increased Gli1 expression in the FaceSMO limb and face mesenchyme...... 182

61. Gli3 expression in the FaceSMO face and limb...... 184

62. Domain of Hand2 expression expanded in FaceSMO limbs...... 186

xiii 63. Bmp2 expression is decreased in the FaceSMO face and limb...... 188

64. Increased BMP4 expression in the FaceSMO face and limb...... 190

65. Increased Gremlin expression in the FaceSMO hindlimb...... 192

66. Increased, followed by decreased, expression of Hox genes in FaceSMO limbs...... 193

67. Axin2lacZ expression in FaceSMO embryos...... 195

68. Crect drives recombination in the surface ectoderm...... 196

69. EctSMO embryos exhibit torso defects and exencephaly...... 198

70. EctSMO brains without exencephaly have indistinct boundaries between sections. .... 199

71. EctSMO embryos have skin barrier defects...... 200

72. EctSMO mice exhibit multiple craniofacial defects...... 203

73. Incisors are missing in EctSMO mice...... 204

74. EctSMO mice have limb defects...... 205

75. Hedgehog signaling is increased in the EctSMO ectoderm, but not mesenchyme...... 207

76. Expression pattern of Gli3 and Hand2 is altered in the EctSMO E11.5 hindlimbs...... 208

77. FGF ligand expression in EctSMO E11.5 hindlimbs...... 210

78. Bmp expression is increased in E11.5 EctSMO limbs...... 211

79. Increased expression of Hoxd11 and Hoxd13 in the anterior of EctSMO hindlimbs. .. 213

80. WNT signaling is upregulated in the EctSMO ectoderm...... 214

Figure 81. MSMO neonates have pigment variation and skin barrier defects...... 216

Figure 82. MSMO neonates exhibit no skeletal defects...... 219

Figure 83. Facial dysmorphology of a child with frontonasal dysplasia...... 221

Figure 84. Model of FaceSMO and EctSMO signaling in the E11.5 hindlimb...... 226

Figure 85. Model of A-P patterning in the FaceSMO limb...... 227

xiv Figure 86. Model of FGF8 and WNT signaling in the developing MCAGF8 mesenchyme over time...... 240

xv LIST OF TABLES

TABLE

1. Examples of human syndromes affecting craniofacial development...... 3

2. Common genetic syndromes associated with craniosynostosis...... 8

3. Primers...... 42

4. Gene comparisons from RNAseq...... 103

xvi ABBREVIATIONS

Transgenics

GOF – Gain-of-Function

LOF – Loss-of-Function

LSL – Lox-Stop-Lox

Cre – Cre recombinase

AP-2Cre – see Creface

BGLAP-Cre – see OC-Cre

Col2-Cre – Cre recombinase driven by the Col2a1 promoter

Crect – Cre recombinase driven by the AP2α ectodermal enhancer element

Creface (a.k.a. AP-2Cre) – Cre recombinase driven by the AP-2α FNP and

limb enhancer elements

Msx2-Cre – Cre recombinase driven by the Msx2 promoter

OC-Cre (a.k.a. BGLAP-Cre) – Cre recombinase driven by the BGLAP

promoter/enhancer; expressed in osteoblasts

CAGF8 (CAG Fgf8) – R26LSL CAG Fgf8

R26F8 (R26 Fgf8) – R26LSL Fgf8b

SmoM2 -- R26SmoM2

Ptc-lacZ – Patched LacZ

Axin2-lacZ – a.k.a. Conductin-lacZ

Mouse Breeding Schemes

ColCAGF8 – Col2a1-Cre; CAG Fgf8

ColR26F8 – Col2a1-Cre; R26 Fgf8

xvii EctSMO – Crect; SmoM2

FaceCAGF8 – Creface; CAG Fgf8

FaceSMO – Creface; SmoM2

MR26F8 – Msx2-Cre; R26 Fgf8

MCAGF8 – Msx2-Cre; CAG Fgf8

MSO – Msx2-Cre; SmoM2

OR26F8 – OC-Cre; R26 Fgf8

OCAGF8 – OC-Cre; CAG Fgf8

Tissue, Morphology, and Orientation

AER – apical ectodermal ridge

FNP – frontal nasal prominence

ZPA – zone of polarizing activity

P-D – proximal to distal

A-P – anterior to posterior

D-V – dorsoventral

Genes and Signaling pathways

β- catenin – catenin (cadherin-associated protein), beta

BMP – bone morphogenetic proteins

BGLAP – bone gamma carboxyglutamate protein

COL2A1 – type II collagen alpha 1

FGF – fibroblast growth factor

FGFR – fibroblast growth receptor

GLI – gli family zinc finger

xviii HH – hedgehog

MSX2 – msh homeobox 2

PTC/PTCH – patched

RUNX2 – runt related transcription factor 2

SHH – sonic hedgehog

SOX9 – sex-determining region Y-box 9

SP7/OSX – osterix

TWIST – twist family basic helix loop helix transcription factor

WNT – wingless INT

Experimental Techniques

RT-PCR – reverse transcriptase polymerase chain reaction

ISH – in situ hybridization

WMISH – wholemount in situ hybridization

PCA – principal component analysis

Clinical

CCTSI – Colorado Clinical Translational Science Institute

ICP – intracranial pressure

CT – computed tomography

TPTPS – triphalangeal thumb polydactyly

xix CHAPTER I

THE GENETICS OF HEAD AND LIMB DEVELOPMENT

Introduction

Tight regulation of signaling pathways during embryogenesis is required for normal

development [1]. In contrast, aberrations in signaling frequently result in fetal and infant

mortality and morbidity [2]. While the head and limb have separate evolutionary origins and

dissimilar anatomy, many of the same genes and signaling pathways regulate craniofacial

and limb development [3]. As such, craniofacial and limb defects are often found together in

genetic syndromes [4]. The unique morphology of the head and limb is thus dependent upon

the specific location, dosage, and timing of gene expression as well as genetic interactions

between genes and signaling pathways. The following sections give an overview of

craniofacial and limb development, the process of ossification in the head and limb, and the

genes and signaling pathways involved in craniofacial and limb morphogenesis that are

critical for my studies, particularly Hedgehog (HH) and Fibroblast Growth Factor (FGF) signaling.

Craniofacial Development

The evidence that human craniofacial diversity is genetically determined surrounds us–

population and sex differences, family resemblances, and identical twins [5, 6]. The

development of the head and face is tightly controlled both temporally and spatially,

requiring the coordinated expression of hundreds of genes. While certain variations in the

genetic control of craniofacial development contribute to facial diversity, others increase the

likelihood of craniofacial malformations.

1 Around 3% of infants are born with a congenital defect [7]. Of these, craniofacial

malformations are the most common defects, with 75% of all birth defects involving

structures of the head and face [8]. Craniofacial defects are both economically and

psychologically expensive. As the face is the most recognizable feature in humans, those

with facial abnormalities are at an increased risk of bullying and societal prejudice.

Additionally, depending on the craniofacial defect, affected individuals may have difficulties

with vision, breathing, hearing, olfaction, speech and/or eating. Thus, they often face

multiple corrective surgeries with lifetime treatment costs accumulating to over $100,000 per

patient [9, 10]

Human Craniofacial Defects

Human craniofacial defects manifest as a wide variety of phenotypes that can impact

multiple areas in the head/face, including the ear, nose, eye, mouth, jaw, skull, and brain.

The four main categories of defects are shown in Table 1 and include orofacial clefting,

craniosynostosis, branchial arch disorders, and syndromes affecting the bone/cartilage.

While some of these defects occur in isolation (non-syndromic defects), many craniofacial

defects are due to a genetic syndrome, in which defects are also observed in other areas of the

body, such as the limbs. The search for causative genes has revealed complex inheritance,

with both genetic and environmental factors involved. Genetic factors contributing to

craniofacial defects are discussed in greater detail throughout this chapter. Environmental

factors include exposure to alcohol, diabetes, smoking, maternal thyroid disease, and certain

medications such as the fertility drug clomiphene citrate. These preventable exposures increase the risk of facial dysmorphology (alcohol), cleft lip (diabetes/smoking), and craniosynostosis (maternal thyroid disease/clomiphene citrate) [11–15].

2

Table 1. Examples of human syndromes affecting craniofacial development. Orofacial • Cleft lip and palate Clefting • Pierre-Robin syndrome • Crouzon syndrome • Apert syndrome • Crouzondermoskeletal syndrome Craniosynostosis • Pfeiffer syndrome • Carpenter syndrome • Jackson-Weiss syndrome • Saethre-Chotzen syndrome • Hemifacial microsomia • Treacher Collins syndrome Branchial arch • Goldenhar syndrome disorders • DiGeorge syndrome • Nager syndrome • Miller syndrome • Oro-facial-digital syndrome Bone/cartilage • Achondroplasia disorders • Cleido-cranial dysplasia • Binder’s syndromes (maxilla-nasal dysplasia) Others • Fetal alcohol syndrome

Orofacial clefts, including cleft lip with or without cleft palate, are the most common craniofacial birth defect, occurring in approximately 1/700 live births [16–19]. The frequency varies widely across racial and ethnic groups with Asian and Amerindian populations having the highest reported prevalence, often as high as 1/500, and African- derived populations having the lowest prevalence, at about 1/2500 [18]. Gender also alters the frequency of cleft lip/palate; males are twice as likely to experience clefts involving the lip whereas females are more likely to have cleft palate only. In unilateral cleft lip cases, the left side is twice as likely to be cleft than the right [16]. Approximately 70% of all cases of orofacial cleft occur as a single deformity, known as non-syndromic cleft lip and palate [20–

3 22]. The remaining cases are associated with one of 500 developmental syndromes that can include cleft lip/palate [16].

Cleft lip and palate result from abnormal growth and proliferation of the prominences during early development. Facial development begins during week 4 of human embryonic development when 5 mesenchymal swellings (prominences) emerge. Development of the face proceeds via growth and fusion of these prominences as shown in Figure 1. By the end of the sixth week, the medial nasal processes have merged with the maxillary processes to form the upper lip and primary palate. Cleft lip arises from the incomplete fusion of the

Figure 1. Overview of human facial development. Schematic diagrams of the development of the lip and palate in humans, illustrating the primary and secondary palate. Modified from Dixon 2011[16].

4 maxillary and nasal processes on one or both sides [16, 23]. Similarly, the secondary palate develops as bilateral outgrowths from the maxillary processes that eventually fuse to separate the oral and nasal cavities. If this fusion does not occur, cleft palate results [16, 24, 25].

While twin studies and familial clustering studies have provided compelling evidence for a genetic component to non-syndromic cleft lip/palate, the genetic analysis of non- syndromic cases is complicated by the lack of clear-cut Mendelian inheritance and susceptibility to environmental risk factors [11–15, 26–29]. Therefore, most cases are thought to arise via a multifactorial model of inheritance [16, 30]. Nevertheless, variants in over 20 genes have been implicated as risk factors, including Msx1, Tfap2a, Irf6, Bmp4, and

FGF family members, Fgfr2 and Fgf8 [16, 30–35]. The genetic causes of syndromic cleft lip/palate are better characterized, with the causative gene mutation identified in over 50 syndromes. These include mutations in TGFβ family members as well as members of the

WNT, HH, and FGF signaling pathways. Notably, cleft palate is a feature in several syndromes, including Apert, Crouzon, and Saethre-Chotzen, that are characterized by the second most common craniofacial anomaly, craniosynostosis [16, 31].

Craniosynostosis occurs in 1 in 2100 to 2500 births and is a complex condition that always involves premature fusion of one or more cranial vault sutures [36, 37]. This premature fusion causes a distortion in skull shape that frequently necessitates surgical correction, both for cosmetic reasons and to prevent potential cognitive impairment [38]. As with orofacial clefts, craniosynostosis can be both isolated (non-syndromic) and syndromic, with 85% being classified as non-syndromic. The other 15% of cases are diagnosed with one of nearly 200 known syndromes, about half of which follow a Mendelian inheritance pattern

[39, 40].

5 As shown in Figure 2, the human calvarium is a dynamic structure, formed primarily

from five bones – the paired frontal and parietal bones and the occipital, which is a fusion of

interparietal and supraoccipital bones. The bones are connected at the edges by sutures, which are critical as they allow for deformation of the skull at birth and the free expansion of the cranium as the brain grows during development. The sutures of the sagittal plane – the metopic suture and the sagittal suture – separate the paired frontal bones and paired parietal bones, respectively. The coronal suture separates the frontal bones from the parietal bones, while the lambdoid suture separates the parietal bones from the occipital bone [41].

Figure 2. Anatomy of the human cranial vault. Bones (blue), sutures (black), and fontanelles (green) of an infant cranial vault.

6 Simple craniosynostosis involves the premature fusion of a single suture, whereas

compound craniosynostosis includes premature closure of two or more sutures [42]. While

simple craniosynostosis can involve the fusion of any suture, the sagittal suture is most often

fused, accounting for 40-60% of human cases. The next most commonly prematurely fused

suture is the coronal suture, accounting for 20-30% of human cases. Premature fusion of the metopic and lambdoid sutures is relatively rare, accounting for 10% and 1% of craniosynostosis cases, respectively [43]. Craniosynostosis of multiple sutures accounts for

5% of craniosynostosis cases and is clinically separated into two groups: two-suture disease

(including bicoronal synostosis) and complex craniosynostosis, with the fusion of more than two sutures [44]. Complex craniosynostosis is associated with a higher rate of reoperation and increased intracranial pressure (ICP), which can cause developmental delay [45]. As with orofacial clefts, gender alters the frequency of craniosynostosis presentation—males are more likely to present with sagittal or metopic craniosynostosis whereas coronal craniosynostosis is more prevalent in females [39]. The association with known familial inheritance also varies between the types of craniosynostosis. While only 2% of sagittal craniosynostosis cases are thought to be familial, 8-10% of coronal craniosynostosis patients have a positive family history [44].

Several genetic mutations have been implicated in craniosynostosis and its related syndromes (Table 2). Activating mutations in the FGF signaling pathway, mostly gain-of function mutations in FGFRs, account for the majority of the known causes of craniosynostosis [46]. Mutations in the TWIST genes as well as MSX2 and EFNB1 also

account for a large number of craniosynostosis cases. Gain-of-function (GOF) mutations in

FGFR1, FGFR2, and FGFR3 are associated with Pfeiffer, Apert, Crouzon, Beare-Stevenson,

7 Table 2. Common genetic syndromes associated with craniosynostosis. Common syndromes associated with craniosynostosis with the corresponding genotypes and phenotypes. Used with permission from Ko, 2016 [47].

Jackson-Weiss, and Muenke syndomes, all of which are characterized by bicoronal craniosynostosis, distinctive facial features, and variable hand and foot findings [44].

The role of FGF mutations in craniosynostosis will be further elaborated upon in the FGF signaling portion of this chapter. The severe morphological abnormalities and cognitive deficits resulting from craniosynostosis and the potential morbidity of surgical correction espouse the need for a deeper understanding of the complex etiology of the condition. Work in animal models, particularly mice, has been pivotal in advancing our understanding of normal suture biology and elucidating pathological disease mechanisms and will be discussed in greater detail in the following section.

Mouse Craniofacial Development

Similarities in craniofacial development and molecular pathways between the human and mouse make the mouse an excellent model for the study of facial morphogenesis and suture biology. As in human, the facial development in the mouse begins with the formation

8 of facial prominences populated by mesenchyme derived from cranial neural crest cells and

mesoderm, surrounded by an overlying epithelium [48]. During normal mouse development, the facial prominences form by E9.5 and by 10.5 are undergoing rapid outgrowth (Figure 3A,

B). At E11.5, the medial nasal processes meet at the midline and by E12.5 the prominences

have fused to form a continuous band of tissue at the front of the face (Figure 3C, D) [24]. In

contrast to humans, primary cleft palates in mice are relatively unusual and are typically

Figure 3. Mouse craniofacial development. Growth and migration of the prominences over time. Red represents the nasal prominences, green the maxillary prominence, and blue the mandibular prominence. Abbreviations: fnp, frontonasal prominence; lnp, lateral nasal prominence; mnd, mandibular prominence; mnp, medial nasal prominence; mx, maxillary prominence. Adapted by Eric Van Otterloo from Thomason, H. A. and Dixon, M. J. 2009. [24]

9 centrally located, due to a failure of fusion at the midline [49–51]. The secondary palate, which will become the roof of the mouth, forms later in mice, at about E15, from the oral part of the maxillary process. The palatal shelves rise around the tongue and then grow towards the midline, where they fuse. Secondary cleft palate results from a failure in fusion or growth [24].

Tissue-tissue interactions between the ectoderm and mesenchyme of the facial prominences are required for normal facial growth and patterning to occur. A region of facial ectoderm, called the Frontonasal Ectodermal Zone (FEZ), regulates outgrowth and dorsoventral patterning in the face, particularly the frontonasal prominence (FNP). The FEZ is defined by the juxtaposition of a Sonic hedgehog (Shh) expression domain and a Fibroblast growth factor 8 (Fgf8) expression domain. While the FEZ was first discovered in chick, a similar region has since been identified in mice [52]. However, in contrast to the chick FEZ, the mouse FEZ is not a single signaling center; instead, Shh is expressed bilaterally. In both mice and humans, Shh is expressed in domains on the right and left side of the face, creating a left and right FEZ [52, 53]. Similar epithelial-mesenchymal interactions are required in the limb bud and will be discussed in further detail in a later section.

As shown in Figure 4, humans and mice also share similar cranial vault bones and sutures, though the shape of the mouse skull results in mice having more prominent nasal bones with additional sutures regulating their growth. In the mouse, the sagittal suture between the paired parietal bones, coronal suture between the frontal and parietal bones, and lambdoid suture between the parietal and interparietal bone are all named similarly to humans. In contrast, the mouse interfrontal suture, between the paired frontal bones, is analogous to the metopic suture in humans. Additionally, the mouse has an internasal suture

10

Figure 4. Bones and sutures of the mouse cranial vault. A) Skull bones of the cranial vault shown on a P0 mouse skull. B) Sutures of the cranial vault shown on an adult (P36) mouse skull.

between the paired nasal bones and a frontonasal suture between the nasal and frontal bones

[54, 55].

Limb Development

For over 60 years, vertebrate limb development has been of significant research

interest [56, 57]. The fundamental processes underlying limb morphogenesis are consistently

used in the development and patterning of many structures, including the face, kidneys,

lungs, and teeth. As the vertebrate limb is not critical for survival and can be easily

manipulated, it is an ideal organ to study the signaling pathways and mechanisms governing

patterning and growth during embryogenesis. The coordinated efforts of several highly

conserved signaling pathways, including FGF, HH, WNT, and BMP, are necessary for

proper limb development [58–60]. The regulation of these pathways is of significant interest

as a better understanding of the signals controlling limb development may provide insight into potential avenues to regenerate amputated limbs or correct limb defects in humans.

11 Human Limb Defects

Congenital limb defects occur in 1 in 500 to 3 in 1000 live births [61–63]. These malformations include both major defects in limb formation as well as more subtle defects in digit number and patterning. Limb formation defects range from the loss of entire limbs, as in amelia, to limb reductions, seen in micromelia, hemimelia, phocomelia, or peromelia.

Supernumerary limbs are rarer, but also occur occasionally. Alterations in digit number includes both loss of digits (oligodactyly) and/or phalanges as well as gain of digits

(polydactyly). Digits may also be fused (syndactyly) or have defects in length [63–65].

Like craniofacial defects, limb defects are often found as part of a syndrome, in which

the limb abnormalities may be found in conjunction with defects in the head, skin, genitals,

or other organs [63, 66–69]. Currently, there are 221 syndromes associated with polydactyly

and 120 syndromes with oligodactyly. Of these, only 84 causative genes have been

identified, 15 of which are associated with syndromes that include polydactyly [70]. While

some limb abnormalities result from genetic inheritance, they also are one of the most common phenotypic effects of several human teratogens. The effects on limb morphogenesis varies by the specific teratogen. One of the more well-known limb teratogens is thalidomide which causes severe limb malformations, including shortening of the limbs and preaxial polydactyly or deficiency (loss of the thumb) [71, 72].

Mouse Limb Development

The process of mouse limb development is shown in Figure 5. In the mouse, limb development begins at E9.5 with the initiation of the forelimb bud from the lateral plate mesoderm. Hindlimb bud initiation begins half a day later and its development remains half to full day delayed compared to the forelimb throughout the critical timepoints of limb

12

Figure 5. Mouse limb bud morphology. The morphology of the mouse forelimb (top) and hindlimb (bottom) during the critical stages of limb development – E9.5 to E13.5. Light blue indicates mesenchymal condensations and dark blue indicates cartilage. Hindlimbs are morphologically delayed by 0.5 to 1 day. Imaged modified from Taher, et al, open access [73].

development, although the developmental difference between them diminishes as pregnancy proceeds. The following timepoints listed in this paragraph are for the forelimb. By E10.5, the apical ectodermal ridge (AER) is visible; the AER is an epithelium that runs anterior to posterior along the limb bud, separates dorsal and ventral sides of the limb bud, and is critical for limb growth and differentiation. Limb outgrowth results in the limb being divided into more proximal and distal elements by E11.5; this outgrowth continues and is responsible for the division of the limb into three zones: the stylopod (future humerus/femur), zeugopod

(future radius/ulna/tibia/fibula), and autopod (future hand/foot/digits). At E12.5, digits are first visible as digital rays; the digits are more distinct by E13.5 due to outgrowth and programmed cell death in the interdigital space [73–76].

13 Due to the growth and differentiation of the embryo, the position of the forelimbs and

hindlimbs relative to other structures changes as development proceeds. While the mouse forelimb bud emerges across from the heart at the mid-abdominal level opposite somites 8-

12, this region eventually forms the future lower cervical/upper thoracic region. Similarly, while the hindlimb bud emerges in the mid or distal region of the tail opposite somites 23-28, this region eventually is located opposite the pelvic region, near the origin of the tail. This illusionary migration or ascent of the limbs is due to the differential elongation of the vertebrae and neck rather than a change in the location of the limb buds, which remain at the same level as the somites they emerged across from [76].

Limb Patterning and Outgrowth

Vertebrate limbs are patterned on three axes: the proximal to distal (shoulder to digit

tips), anterior to posterior (thumb to little finger), and dorsal to ventral (back of hand to palm)

axes. Proximal-distal (P-D) growth is regulated by the apical ectodermal ridge (AER), which

secretes FGF ligands that signal to the underlying limb mesenchyme (Figure 6) [74, 77].

Anterior-posterior (A-P) patterning is controlled by the zone of polarizing activity (ZPA) in the posterior limb mesenchyme which secretes SHH, the main controller of this axis (Figure

6) [78–80]. Dorsoventral patterning (D-V) requires the restriction of WNT7A to the dorsal limb ectoderm by Engrailed1, which is localized in the ventral ectoderm [81, 82].

Additionally, expression differences in Hox, T-box family, and other genes along the embryo’s rostral-caudal axis specify forelimb and hindlimb identity [60]. Several major signaling pathways are integral to limb patterning, specifically HH, FGF, WNT, and BMP.

The role of several members of these pathways in limb outgrowth and patterning is described

14

Figure 6. Limb signaling centers. A sagittal section through a limb bud illustrating the A-P and P-D axes as well as the primary signaling centers of the developing limb, the ZPA and AER. Image modified from Ibrahim, 2015 [83].

in the following section. A more in-depth explanation of these signaling pathways and

further discussion of their roles in limb development will be described later in this chapter.

Proximal-Distal Limb Bud Development

The P-D limb bud axis is defined by the direction of outgrowth that results in the formation of the stylopod and zeugopod. As previously discussed, limb buds are derived from the mesoderm and their outgrowth is initiated at defined positions along the embryonic axis. Limb buds consists of proliferating mesenchyme surrounded by ectoderm. Epithelial- mesenchymal interactions between these two layers are critical for limb bud outgrowth and patterning. The AER is a specialized epithelium that runs anterior to posterior along the distal limb bud tip and is the signaling center responsible for limb bud outgrowth. One of the earliest markers of markers of the presumptive AER is Fgf8. Bmps as well as at least three

15 other Fgfs are expressed in the AER (Fgf4, Fgf9, Fgf17) and there is functional redundancy

between several of these Fgfs [59, 84–86].

Interest in AER function began in 1948 when it was found that limb skeletal

abnormalities could be caused by removal of the AER from chick limb buds at successive

developmental stages. When the AER was removed early, both the zeugopod and autopod

were absent, but when it was removed later, only the autopod was missing [86]. These

observations led to the “Progress Zone Model” of P-D limb development which proposes that

the progressive specification of the skeletal elements depends on the amount of time the

progenitors have spent in the undifferentiated region immediately adjacent to the AER,

where they are exposed to AER-FGFs that stimulate their proliferation and act as distalising

factors (Figure 6) [87, 88]. However, more recent research has called this model into question as the Progress Zone Model fails to explain limb phenotypes resulting from conditional removal of FGFs from the AER. Thus, the newest model, the “two-signal model” proposes that two opposing signals, such as retinoic acid secretion from the flank

(proximal) and FGF secretion from the AER (distal), are responsible for patterning the skeletal elements along the P-D axis [84]. This model postulates that AER-FGFs have instructive roles in specifying the P-D fates, and thus regulating skeletal identities, during limb bud development. The two-signal model is supported by genetic analysis of the role of individual FGFs expressed in the AER during limb bud development; however, neither model is fully able to integrate all of the experimental data, and thus another new model is needed [56, 84, 89].

16 Anterior-Posterior Patterning of the Limb Bud

The A-P limb bud axis is defined by the growth that results in the sequence of digits

1-5 (thumb to little finger). Identification of the ZPA as the signaling center responsible for

A-P patterning laid the foundation for understanding the patterning processes during limb

bud development. A gradient of the morphogen Sonic Hedgehog (SHH) emanating from the

posterior mesenchyme is responsible for the ZPA’s polarizing activity. Loss of SHH in mice

leads to a severe limb phenotype characterized by the loss of posterior digit development (2-

5); in contrast, grafting posterior (SHH containing) grafts onto the anterior limb leads to extra

digits displayed as mirror image duplications [90–93]. During the establishment of A-P

patterning, the mutual antagonism between the transcription factors GLI3 and HAND2

restricts SHH to the posterior limb bud. Expansion of GLI3R into the posterior mesenchyme

leads to loss of Shh expression whereas expansion of HAND2 into the anterior mesenchyme

induces a second Shh expression region in the anterior mesenchyme which leads to mirror

image duplications of digits on the anterior side [93–95]. SHH enhancers, such as the ZRS

(zone of polarizing activity regulatory sequence) are also responsible for restricting Shh

expression to the posterior limb bud. Point mutations in the ZRS of both humans and mice

result in ectopic expression of Shh and consequently, pre-axial polydactyly. In contrast,

deletion of the ZRS from the genome leads to loss of Shh expression and loss of digits 2-5, a

similar phenotype to that observed in Shh LOF mice [85, 96, 97].

The ZPA also plays an important role in AER maintenance and regression. Likewise, maintenance of the ZPA requires signaling from the AER, which again demonstrates the importance of epithelial-mesenchymal interactions during limb development. AER regression occurs following skeletal progenitor specification at E12.5, and is important for

17 limiting limb outgrowth [98]. Ectodermal expression of Fgfs is maintained by Shh from the

ZPA through the induction of Gremlin, a BMP inhibitor that prevents BMPs in the mesenchyme from inhibiting the AER-FGFs [78, 99, 100]. In turn, Fgf expression in the ectoderm activates the expression of Shh in the ZPA [101]. Therefore, the ZPA and AER support each other through a positive loop of HH and FGF signaling between the limb ectoderm and mesenchyme.

Dorsoventral Patterning of the Limb Bud

As compared to the proximal-distal and anterior-posterior axes, little is known about the dorsoventral axis, responsible for patterning the front (palm, ventral) and back (dorsal) of the limbs. Early experiments in the chick limb bud suggest that dorsoventral identity is established prior to limb bud outgrowth in both the presumptive limb mesoderm and ectoderm [102]. During dorsal patterning, WNT7A from the dorsal ectoderm induces LMX1 in the dorsal mesenchyme [103]. Engrailed-1 is required for ventral limb patterning [82].

Together, WNT7A and Engrailed-1 are the necessary for normal dorsoventral patterning;

WNT7A KO mice lose dorsal identity whereas Engrailed-1 KO mice lose ventral identity

[81, 82].

Ossification in the Cranium and Limb

Bone forms via two processes: intramembranous or endochondral ossification. The majority of the skeleton, including the long bones, vertebrae, and basicranium, forms via endochondral ossification during which condensed mesenchyme cells first differentiate into chondrocytes that form cartilage tissue. This intermediate cartilaginous template is then replaced by bone, formed through osteogenesis. In contrast, most of the skull, including the cranial vault, jaw, and most of the facial bones, is generated via intramembranous

18 ossification, in which condensed mesenchyme cells pass through a preosteoblast stage and

differentiate directly into osteoblasts that form bone without any cartilaginous precursor

[104, 105].

Intramembranous ossification of the cranial vault bones is initiated by condensation of the mesenchymal cells between the epithelium and the dura mater. Similarly, in the facial bones, intramembranous ossification is initiated by mesenchymal condensations within the

facial, maxillary, and mandibular prominences [39, 106]. As these cells differentiate

(preosteoblasts) into osteoblasts, they are marked by a sequence of transcription factors – first Sox9, followed by Runx2, and finally Osterix (Osx) (Figure 7) [105]. Sox9+ cells have

Figure 7. Stages of bone and cartilage lineage cell differentiation. Differentiation of osteoblast lineage cells (top row) during cranial vault bone development (intramembranous ossification), though similar paths are taken by these cells during endochondral ossification. The bottom row shows the chondrocyte lineage, which is involved in endochondral ossification. Dashed lines show differentiation relationships which have yet to be confirmed in vivo. Image modified with permission from Flaherty et al, 2016 [39].

19 the potential to differentiate into osteoblasts or chondrocytes. In cranial neural crest cells, the

deletion of Sox9 causes a shift from chondrocyte to osteoblast differentiation [107].

Conversely, the deletion of Osx causes Runx2+ cells to undergo chondrocyte rather than osteoblast differentiation [108]. In addition, deletion of β−catenin in the WNT signaling pathway can shift cell fate from osteogenic to chondrogenic in both Runx2+ and

Runx2+Osx+ cells [109–111] . Osteoblasts secrete collagen-rich, non-mineralized bone

matrix known as osteoid. Once the osteoid matrix is established, the osteoblasts deposit

hydroxyapatite as a scaffold which serves to mineralize the matrix while continuing to

deposit osteoid in the periphery of the forming bone. These osteoblasts then become trapped

within the bone matrix as the osteoid expands and mineralizes, causing them to differentiate into mature bone cells, known as osteocytes, or undergo apoptosis (Figure 7) [39, 112, 113].

The initial stage of endochondral ossification is formation of a cartilage model. The formation of this model is initiated when aggregated mesenchymal cells differentiate into chondroblasts. These chondroblasts condense and subsequently differentiate into chondrocytes that proliferate to form a template, called the cartilage anlagen, in the shape of the eventual bone. Upon differentiation, the chondrocytes eventually stop dividing and instead increase in volume, becoming hypertrophic chondrocytes. As some of these hypertrophic chondrocytes undergo apoptosis, surrounding cells differentiate into osteoblasts that produce osteoid matrix and eventually replace the cartilage with bone [39, 114, 115].

Additionally, recent studies have shown that chondrocytes can also survive and directly differentiate into osteoblasts (Figure 7) [116, 117].

Bone development in the cranial vault is a complex process. The anterior cranial bones of the cranial vault (nasal, frontal) are derived from cranial neural crest cells, [118,

20 119] whereas the posterior portion (parietal) is derived from the paraxial mesoderm [119,

120]. Together, these cranial neural crest and paraxial mesoderm derived cells form the

intramembranous bones and sutures of the skull [119, 121–123]. Intramembranous

ossification of the skull vault involves direct bone matrix deposition to form calvarial plates,

which expand during development but do not fuse with other cranial bones during

embryogenesis and infancy [124]. Instead, sutures connect the individual intramembranous

bones and serve as growth centers that regulate the expansive growth of the skull [125]. Like

the cranial bones, the cellular origins of the sutures are varied, with the interfrontal and

sagittal sutures being neural crest derived and the coronal suture being mesoderm derived in

the mouse [126, 127]. Because sutures are the major sites of bone growth during cranial

vault development, signaling at the sutures is essential for the regulation of intramembranous

ossification [128, 129]. Additionally, cranial suture fusion and suture patency is regulated by

secretion from the dura mater of a variety of growth factors, which signal to the sutures in a

paracrine fashion [130–132]. Several signaling pathways are implicated in proper skull

ossification and growth including FGF, HH, WNT, and BMP signaling [133]. The

contributions of each signaling pathway to ossification will be discussed in the following

sections.

Critical Signaling Pathways in Craniofacial and Limb Development

Signaling can occur through two different mechanisms: autocrine and paracrine signaling. During autocrine signaling, a cell secretes a substance that binds to autocrine receptors on that same cell, leading to changes in that cells signaling. In contrast, during paracrine signaling, a cell produces a signal that binds to receptors on nearby cells, leading to changes in the cell signaling of neighboring cells. While autocrine signaling has important

21 roles in immunology and cancer, paracrine signaling predominates in most developmental

contexts [134]. As such, the FGF, HH, WNT, and BMP signaling pathways regulate

craniofacial and limb development through paracrine signaling. In the following section, I

discuss FGF and HH signaling in detail, as these pathways are the focus of this dissertation,

followed by a brief overview of WNT and BMP signaling.

FGF Signaling

Dynamics of the FGF Signaling Pathway

In both humans and mice, the FGF (Fibroblast Growth Factor) signaling

pathway is composed of four FGFRs (Fibroblast Growth Factor Receptors) and 22

FGF ligands that function to regulate cell proliferation, differentiation, and

migration in a variety of systems [101]. Phylogenetic analysis has grouped the 22

FGF ligands into 7 subfamilies, each containing 2-4 members. The focus of this

dissertation, FGF8, is most closely related to FGF17 and FGF18 [135]. Eighteen of these FGFs function by binding to four receptor tyrosine kinases (FGFRs); the remaining FGFs (FGF11-14) are intracellular proteins and do not interact with

FGFRs [136, 137]. As shown in Figure 8A, FGFRs consist of a transmembrane

domain, a split intracellular tyrosine kinase domain, and an extracellular ligand

binding domain containing three immunoglobulin-like loops (IgI, IgII, IgIII) [138].

Mutations in any of these regions can cause craniosynostosis [44]. The second half

of the IgIII domain of FGFR 1-3 is encoded by the IIIb or IIIc exon, as the gene is

alternatively spliced in a tissue specific manner; IIIb is expressed in the epithelia

whereas IIIc is expressed in the mesenchyme [138]. Therefore, alternative splicing of

22 Figure 8. FGFR structure, signaling, and dysregulation. (A): Basic structure of an FGFR and downstream signaling. (B): FGF signaling dysregulation can be ligand dependent or independent. Ligand depend activation of FGFRs can be dysregulated when a cell overproduces FGF ligand (1) or when a cell produces splice-variant FGFRs (2) that have altered specificity to endogenous FGF ligands. Ligand- independent dysregulation of FGFRs can occur when an FGFR becomes mutated (3), leading constitutive activation of the kinase, or when a gene translocation occurs (4), whereby the FGFR fuses with a transcription factor or promoter region resulting in overexpression or activation of the FGFR. A third mechanism is when a gene amplification for the receptor occurs (5), resulting in grossly exaggerated expression of the receptor. Image used with permission from Brooks et al. 2012 [139].

the receptors allows ligands to activate receptors in the adjacent mesenchymal or epithelial tissue without activating autocrine signaling. For example, Fgf8 is

23 normally expressed in the epithelium and signals to FGFRc isoforms in the

neighboring mesenchyme whereas Fgf10 is normally expressed in the mesenchyme

and signals to FGFRb isoforms in the neighboring epithelium [140–143]. Ligands

can also undergo alternative splicing. For example, there are 8 isoforms of FGF8 in

mice and 4 in humans, at least some of which have distinct bioactivities and

functions [144, 145].

FGFRs are normally activated by the binding of FGFs, and a coreceptor,

heparan sulfate proteoglycan. This induces dimerization of FGFRs which leads to

phosphorylation of the intracellular tyrosine kinase domains [146]. Subsequent downstream signaling occurs through two main pathways, Ras-Erk1/2 dependent and Ras-Erk1/2 independent signaling. During Ras-Erk1/2 dependent signaling, phosphorylation of the intracellular tyrosine kinase domain on the FGFR promotes the binding of adaptor molecules that function as a scaffold to engage additional signaling proteins that are involved in the RAS/RAF/MAPK signaling cascade.

Through Ras-Erk1/2 independent signaling, FGFRs are able to activate several additional signaling pathways, including PI3K/Akt, Pkc, Src, Stat1, p38, and Jnk

[139, 147]. FGF signaling is critical to a number of developmental processes including angiogenesis, wound healing, mesoderm induction/patterning, neuronal differentiation, malignant transformation, facial development, limb development, and skeletogenesis [44]; the last three of these will be the primary focus of the following sections, with particular emphasis on the role of Fgf8 in these processes.

24 FGF Signaling in Craniofacial Patterning and Skeletogenesis

FGF signaling is critical from the earliest stages of craniofacial development as it induces neural crest formation [148–150]. As development progresses, FGF signaling is present in both the epithelia and mesenchyme and mediates the epithelial-mesenchymal interactions that are required for normal development of almost all of the facial structures

[125]. Both FGFR1 and FGFR2 are expressed in the facial primordia [151, 152]. FGF ligands are also expressed the facial primordia in redundant and restricted domains. For example, Fgf8, Fgf9, and Fgf10 are expressed in the nasal pits whereas Fgf3, Fgf15, and

Fgf17 are also expressed in the nasal pits, but restricted to the medial side [152].

Fgf8 is particularly important in early craniofacial patterning and growth. Fgf8 expressed from the ectoderm is critical for formation of several structures derived from the facial prominences. While deletion of Fgf8 leads to embryonic death at gastrulation [153,

154] conditional loss of FGF8 in the ectoderm of the first pharyngeal arch leads to near complete loss of the teeth, maxillary prominence, and mandibular prominence, as well as clefting of the nasal prominences [155–159]. Additionally, while removal of the facial ectoderm in chick results in craniofacial defects, addition of Fgf8 via beads rescues certain aspects of the phenotype [160]. Finally, FGF8, together with SHH, is sufficient to promote chondrogenesis and outgrowth in the face [161, 162]. In Chapter 4 and 5, I further elucidate the role of Fgf8, expressed from both from the cranial ectoderm and mesenchyme, on facial patterning by examining the phenotypic effects of increased Fgf8 in the face.

In addition to FGFs critical role in chondrogenesis, FGF signaling is an important regulator of skeletogenesis. Detailed studies using genetically modified mice with dominant or LOF FGFR mutations have confirmed the central role of this receptor family in bone

25 formation, skeletal development, and craniosynostosis [163]. Moreover, activating mutations

in FGFRs can lead to both ossification of the sutures as well as thinner calvaria [164].

Additionally, FGF2 stimulates proliferation of osteoprogenitors in the dura mater and

overlying suture mesenchyme [165, 166]. Finally, mouse mutations that increase the

diffusion properties and effective range of FGF9 also lead to craniosynostosis [167]. This

latter result may reflect the mechanism underlying human multiple synostoses syndrome,

which is caused by rare autosomal mutations in FGF9 [168].

As mentioned previously, activating mutations in the FGF signaling pathway, mostly gain-of function mutations in FGFRs, account for the majority of the known causes of craniosynostosis [46]. The etiology of craniosynostosis is complicated, however, by the fact that activating mutations in the same gene can cause a variety of different syndromes and/or types of craniosynostosis. For example, mutations in FGFR2 can cause seven of the eight

FGFR-related craniosynostosis disorders [169]. Interestingly, identical FGFR2 mutations have been found in patients with Crouzon, Pfeiffer, and Jackson-Weiss craniosynostosis syndromes, suggesting genetic modifiers play a role in clinical presentation [170, 171]. On the other hand, the same clinical phenotype can result from mutations in different genes, suggesting functional redundancy among different FGFR molecules; for example, Pfeiffer

syndrome is associated with mutations in both FGFR1 and FGFR2 [170].

While some FGFR activating mutations cause ligand independent dimerization and

activation of the receptors [172–175], others do not lead to ligand independent receptor

activation, but instead cause increased affinity and altered specificity for different FGF

ligands (Figure 8B). These ligand dependent activating mutations are found in FGFR1-3 and

include mutations associated with Apert, Pfeiffer, and Muenke syndromes [176–180]. Thus,

26 craniosynostosis patients with these ligand dependent activating mutations are likely more

sensitive to FGF ligand dosage than those with ligand independent mutations. However,

most current mouse models have ligand independent FGFR mutations. In chapter 4, I

examine the effects of FGF ligand dosage on craniosynostosis presentation and cranial vault

development by overexpressing the FGF8 ligand at different doses using two novel FGF8

alleles.

FGF Signaling in Limb Development

FGF signaling in both the epithelium and mesenchyme is also critical for limb

development. Fgf10 expressed in the limb bud mesenchyme activates FGFR2b which in turn induces Fgf8 in the AER [140, 141, 181]. The regulation of ectodermal Fgf8 by the mesodermal Fgf10 is accomplished through a pathway involving Wnt3a in the ectoderm

[182, 183]. Together with Fgf8, the AER-FGFs Fgf4, Fgf9, and Fgf17, promote expression

of Fgf10 in the mesenchyme [84, 184] Loss of Fgf10 in the mesenchyme results in complete

loss of the limbs [141]. While conditional loss of Fgf8 alone only causes a mild skeletal

phenotype [89, 185], inactivation of Fgf8 in conjunction with Fgf4 in the ectoderm also leads

to almost complete loss of the limb [186]. Similarly, inactivation of Fgf8 in conjunction with

Fgf9 and Fgf17 leads to severe skeletal phenotypes [84]. However, in mice, limbs lacking

Fgf4, Fgf9, and Fgf17 have a normal skeletal pattern, suggesting Fgf8 expression in the AER

is sufficient for normal limb formation [84]. Furthermore, the AER has been shown to

depend on FGFR signaling [187]. Interestingly, in addition to its well-established role in P-D

patterning, FGF signaling may play a role in A-P patterning. In chick, aberrant FGF

signaling has been shown to result in pre-axial polydactyly through ectopic expression of

27 FGFs in the AER [188]. In summary, reciprocal FGF signaling between the mesenchyme

and ectoderm in the limb regulates induction, P-D patterning, and potentially, A-P patterning.

FGF signaling also plays a role in endochondral ossification and limb skeletogenesis.

FGFRs are known to be expressed in the developing skeleton. Mutations in FGFR3 cause

achondroplasia, the most common form of skeletal dwarfism in humans, as well as other

forms of dwarfism [189]. FGFR1-3, as well as several FGF ligands, are expressed in

developing bone. Fgf2 is expressed in chondrocytes and osteoblasts whereas Fgf7, Fgf8,

Fgf9, Fgf17, and Fgf18 are all expressed in the perichondrium [104, 189]. Mice lacking

FGF18 have defects in both chondrogenesis and osteogenesis [190, 191] whereas mice

lacking FGF7 and FGF17 have apparently normal chondrogenesis [192, 193]. As mentioned

previously, mice lacking FGF8 die prior to skeletogenesis [154]. FGF9 null mice are slightly smaller than controls and have disproportionally short proximal skeletal elements [189].

However, dominant mutations in FGF9 in both mice and humans affect chondrogenesis and formation of the joints [104]. Thus, while FGFs have a clear role in limb ossification, several issues still need to be addressed, including functional redundancy with other FGFs, mechanism of action, and delineating the roles in chondrogenesis from osteogenesis.

Hedgehog (HH) Signaling

Dynamics of HH Signaling

Hedgehog signaling was first identified in Drosophila, and its expression was later found in all vertebrates [194]. In vertebrates, the Hedgehog (HH) family consists of three protein ligands: Indian Hedgehog (IHH), Desert Hedgehog (DHH), and Sonic Hedgehog

(SHH). While Dhh expression is primarily limited to the male reproductive tract [195, 196],

28 the other two ligands, Shh and Ihh, are expressed during craniofacial and limb development and will be discussed in greater detail below.

All of the HH ligands signal through the same highly conserved HH signaling pathway through a three-step process. The HH pathway signals through paracrine signaling as shown in Figure 9; thus, the first-step is to prepare the ligand for diffusion. To do this, the

HH ligand precursor undergoes a series of modifications that convert it into a multimeric, active form (shown in yellow in Figure 9) that is diffusible and has an increased affinity for

the cell membrane [198]. Next, Dispatched, a large transmembrane protein, releases HH

from the signaling cell. Once HH has been secreted, it can bind to the Patched receptor, an

Figure 9. The hedgehog signaling pathway. Depiction of the Hh signaling pathway. Described in greater detail in text. In brief, in the absence of hedgehog, patched inhibits smoothened. When hedgehog is present, it binds to patched, which releases the inhibition of smoothened and allows for the transcription of hedgehog target genes. Image acquired from Pan et al. 2013, open access [197].

29 inhibitor of signal transduction. In the absence of HH, Patched inhibits Smoothened (SMO), another transmembrane protein. However, once HH has bound to Patched, SMO is freed from Patched repression and shuttles through the cilia, which ultimately leads to the transcription of HH target genes. This occurs through SMO interaction with the glioblastoma gene products (GLI) family of transcription factors (GLI1, GLI2, GLI3). GLI1 and GLI2 act primary as transcriptional activators, whereas GLI3 exists in two forms: a full-length activator form (GLI3A) and a truncated repressor form (GLI3R). The cleavage that results in the truncated form of GLI3 occurs in the absence of HH signaling and is inhibited by HH signaling. Thus, in the presence of HH signaling, GLI3 is a transcriptional activator of HH target genes, which include Gli1 and Patched [197, 199].

Hedgehog Signaling in Craniofacial Development

Mutations in the HH signaling pathway, including mutations in the HH effectors and co-receptors as well as those affecting ciliary proteins, cause a wide range of craniofacial abnormalities, including holoprosencephaly, hypotelorism, cleft lip/palate, frontonasal dysplasia, and other skeletal and craniofacial deformities [197, 200, 201]. Hedgehog signaling is essential for normal head development; as SHH knockout mice do not develop any recognizable craniofacial features (Figure 10). Thus, mouse studies typically examine

HH signaling in the head though conditional knockouts by using Cre-recombinases. The role of IHH and SHH in cartilage and bone patterning has been well studied [48, 202]. However, studies examining the role of HH signaling in calvarial ossification and cranial suture morphogenesis are newer and thus the role of HH in these processes is less well understood.

In Chapter 6, I will examine the effects on increased HH signaling on both craniofacial patterning and ossification.

30

Figure 10. Embryos without SHH do not develop recognizable heads. Lateral view of E15.5 control (left) and SHH knockout (right) embryos. Image modified with permission from Chiang et al, 1996 [92]

During facial development, Shh is expressed in the ectoderm of the frontonasal and maxillary processes. In chick, transient loss of SHH signaling during the later stages of craniofacial morphogenesis causes hypotelorism, where the space between the eyes is reduced, whereas excess SHH causes hypertelorism, or widening between the eyes.

Additionally, disrupting SHH signaling in the frontonasal or maxillary processes leads to loss of outgrowth and clefting [203]. In mice, the role of increased hedgehog signaling during craniofacial development has been examined using a SmoGOF construct, in which smoothened in constitutively active, in conjunction with Wnt1-Cre. In these mice, formation of the face and head, including many of the cranial bones, is severely impaired (Figure 11).

31

Figure 11. Craniofacial phenotypes of Wnt1-Cre; SmoGOF embryos. (A-D) Heads of control (A, C) and Wnt1-Cre; SmoGOF (B,D) embryos at the stages indicated. (E, F) Skeletal preparations of heads from control (E) and Wnt1-Cre; SmoGOF (F) embryos at E18.5. Adapted from Jeong et al 2004 [204].

32 Teratogens can also impair SHH mediated craniofacial patterning. For example, exposure to

cyclopamine causes holoprosencephaly in mammalian embryos [205]. Additionally, the

administration of ethanol to chick embryos leads to a dramatic loss of SHH, but not other

signaling molecules. However, both the alcohol-induced craniofacial growth defects as well

as the cranial neural crest cell death can be rescued by application of SHH [206]. This is of

vital importance as the pattern of facial abnormalities found in Fetal Alcohol Syndrome are

similar to the facial defects observed with decreased HH signaling in the developing embryo.

In both cases, there is a reduction in head size as well as apoptosis of the cranial neural crest

cells that results in morphological abnormalities along the midline of the face [207].

Compared to FGF, TGF-β, and BMP signaling, the role of HH signaling in cranial suture and bone morphogenesis is a relatively new area of study, yet one of clear importance.

Both SHH and IHH are critical regulators of osteogenesis and are therefore of importance to cranial suture biology. However, they have different expression patterns that indicate probable differences in function. The domains of Shh expression in the skull are still debated

but occur in later embryonic development. While some investigators argue that Shh expression in the mouse skull does not begin until E18.0 in the midline suture mesenchyme

[208], others have found that Shh is expressed in the parietal bones at E16.5, but not in the midline suture mesenchyme [209]. In contrast, Ihh is expressed in mice during calvarial osteoblast development and during osteoblast proliferation in the osteogenic fronts [210].

Additionally, Ihh is expressed at lower levels in the sagittal suture [209].

The differing regions of Shh and Ihh expression suggest they may have different functions. Given the presence of Shh in the suture mesenchyme, it has been postulated that

Shh has a role in maintaining suture patency [208]. Unlike other cytokines, such as Fgf2 and

33 TFG-β, that are concentrated in the dura mater underlying the sutures, Shh/Ihh expression does not seem to come from the dura. Ihh, on the other hand, is expressed by calvarial

osteoblasts and promotes ossification and fusion of the cranial and palatine bones [209, 211].

Disruption of IHH signaling impairs osteoblastogenesis and leads to a reduction in cranial

bone size, likely via downregulation of Bmp expression [209], In rabbits, Ihh expression may lead to craniosynostosis [212]. Additionally, duplications of Ihh in mice and humans result in craniosynostosis [213] and mutations in Rab23 and Ptch1, which lead to increased

HH signaling, also cause craniosynostosis [214, 215]. The preceding evidence suggests that

SHH may prevent suture fusion whereas IHH promotes calvarial ossification and suture closure. However, IHH has also been shown to repress osteogenic lineage differentiation, with loss of IHH resulting in premature osteoprogenitor cell differentiation [216].

Additionally, a large number of in vitro studies have demonstrated the pro-osteogenic effects of SHH in diverse cells types, including pre-osteoblastic cells and primary mesenchymal stem cells [217–221]. Thus, the role of hedgehog signaling in cranial ossification is not yet well understood.

Hedgehog Signaling in Limb Development

Defects in hedgehog signaling cause several inherited limb disorders. Specifically, these malformations include both pre-axial and post-axial polydactyly as well as syndactyly

[222]. However, these defects are not actually due to alterations in the coding sequence of

Shh, as mutations within Shh have not been associated with malformations. This is presumably because mutations in Shh are not compatible with life. However, point mutations in the ZRS enhancer, which are predicted to lead to ectopic Shh expression specifically in the limb bud, are found in patients with pre-axial polydactyly and

34 triphalangeal thumb polydactyly (TPTPS). TPTPS patients can have either pre-axial or post- axial polydactyly [223]. This suggests that point mutations in the ZRS perturb normal regulation of SHH in the posterior limb bud as well. Additionally, defects in the response to

SHH signaling are found in syndromes that include polydactyly. For example, mutations in

Gli3 cause Pallister-Hall and Grieg Cephalopolysyndactyly syndromes which present with

pre-axial and post- axial polydactyly [224, 225]. In these cases, the polydactyly is likely due

to the loss of restriction of SHH to the posterior of the limb. Finally, mutations in Hoxd13, a putative target of SHH signaling, are associated with polydactyly, syndactyly, and brachydactyly (short digits) [226, 227].

In addition to SHH’s role in A-P patterning, discussed previously, hedgehog signaling acts as a master regulator of endochondral ossification, and thus limb ossification. However, in limb ossification, IHH, rather than SHH, is the major active HH ligand. Ihh is expressed in the chondrocytes of developing endochondral bones, where it stimulates chondrocyte proliferation and regulates chondrocyte differentiation. Later in development, IHH is critical for osteoblast differentiation and stimulates endochondral bone formation [228, 229]. In the absence of IHH signaling, osteoblast differentiation is abolished and endochondral bone ossification is therefore lost [229, 230]. While IHH’s main function in chondrocyte differentiation is to suppress GLI3R function, during osteoblast differentiation IHH is required to activate GLI2 activator [231–233]. Thus, hedgehog signaling is critical for several different aspects of limb development.

WNT Signaling

The WNT signaling pathway is a highly conserved signal transduction pathway used

extensively during animal development, controlling multiple aspects of development

35 including proliferation, fate specification, polarity, and migration of cells [234–237]. There

are 19 secreted WNT ligands that signal though the canonical, β-catenin dependent pathway,

or at least two different noncanonical pathways [238–240] . Here, I will focus on the canonical pathway, shown in Figure 12, in which regulation of β-catenin activity, including

both protein abundance and nuclear localization, is paramount. In the absence of WNT

ligand, the destruction complex (CKIα, GSK3β, APC, Axin) phosphorylates β-catenin, which

allows the ubiquitin ligase complex to bind to it. This targets β-catenin for proteasome-

mediated degradation. In contrast, binding of the WNT ligand to coreceptors Frizzled and

LFP5/6 leads to the activation of Disheveled, which suppresses the formation of the

destruction complex. As a result, β-catenin is not phosphorylated and thus not degraded.

Figure 12. Canonical WNT signaling pathway. WNT signaling pathway in the absence (A) and presence (B) of WNT ligand. Image used with permission from Eisenmann, 2005, Creative Commons [237].

36 Therefore, β-catenin accumulates in the cytoplasm and is translocated to the nucleus. In the nucleus, it interacts with TCF/LEF proteins to activate transcription of WNT target genes, including Axin2 [237, 241, 242]. Axin2 then contributes to the regulation of the pathway by serving as a negative regulator of WNT signaling [243].

WNT signaling is absolutely imperative to normal morphogenesis; β-catenin knockouts die during early embryogenesis, around the time of gastrulation [244]. Thus, Cre recombinase transgenes with different tissue specificities have been employed to better investigate WNT signaling effects on specific aspects of development. During craniofacial development, conditional loss of β-catenin using Wnt1-Cre leads to defects in brain formation and cranial neural crest survival, resulting in the loss of all craniofacial elements

[245]. In the developing limb, WNT signaling has been shown to have critical roles in limb bud initiation, limb outgrowth, early limb patterning, and later limb morphogenesis events

[242]. Regulation of WNT signaling is also critical for normal osteoblast and chondrocyte differentiation in the developing head and limb as WNT signaling suppresses chondrocyte differentiation which allows for osteoblast differentiation [246–249].

BMP Signaling

The BMP gene family belongs to the TGF-β superfamily and contains more than 20 secreted ligands [250]. These ligands are divided into subfamilies based on phylogenetic analysis and sequence similarity. The dpp and 60A subfamilies have been extensively investigated during embryonic development; BMP2 and BMP4 are part of the dpp subfamily due to their similarity to the ancestral dpp gene in Drosophila and BMP5-8 are part of the

60A subfamily [251]. As shown in Figure 13, BMPs bind to receptor serine/threonine protein kinases known as type I (BMPR1A, BMPR1B, and ALK2) and type II receptors

37 Figure 13. Schematic of the BMP signaling pathway. (A): BMP ligand unbound state and (B) BMP ligand bound state. Used with permission, Bandyopadhyay et al. 2013 [252].

(BMPR2, ActRIIA, ActRIIB) [250]. Binding of BMP ligands to the Type II receptor leads to

phosphorylation of the Type I receptor which in turn results in phosphorylation of

downstream Smad proteins and thus triggers the intracellular signal cascade. The activated

Smads (Smad1, 5, or 8) can form heterodimers with Smad4 and translocate into the nucleus

where it activates transcription of a host of BMP downstream target genes [253, 252].

BMP signaling is critical for normal craniofacial development. Specifically, BMP

signaling is involved in the early patterning of the head, the development of cranial neural crest cells, and facial patterning as well as the development of mineralized structures, such as

cranial bones, maxilla, mandible, palate, and teeth [254]. BMPs are expressed in developing

sutures which results in the expression of the BMP antagonist Noggin. Noggin is expressed

postnatally in the suture mesenchyme of patent, but not fusing cranial sutures and is thought

to be an important contributor to maintaining open sutures [255]. Mutations within the BMP

pathway have been associated with various human craniofacial malformations including

neural tube defects, cleft palate, tooth defects, and bone and cartilage defects [254].

38 Additionally, knockout mouse models have been made for most of the BMP family

members; these knockouts frequently result in malformations in the head and limb [251].

BMP2, BMP4, and BMP7 are of particular interest in craniofacial development as they are

prominently expressed during early craniofacial morphogenesis [256–258].

As in the face, BMP2, 4, and 7 are expressed in the limb, with BMP4 having the most

central role in digit patterning [259–261]. Dosage and timing of BMP signaling is critical in limb development as decreased BMP signaling during early development (prior to E10.5) results in digit loss, whereas BMP signaling inhibition after the limb bud has been initiated results in polydactylous limbs [259–264]. Additionally, loss of BMP2 and 4 results in syndactyly of the digits, as BMPs induce apoptosis of the interdigital mesenchyme [259,

265]. As BMPs promote compaction of mesodermal cells into cartilage elements [266],

alterations in BMP signaling can also impact limb bone shape. Overexpression of BMP2 or 4

in the chick led to a dramatic increase in the volume of cartilage elements, altered their

shapes and led to joint fusions [267]. Given the importance of the BMP signaling, it is not

surprising that BMP activity in the limb bud is tightly regulated. SHH expressed in the

posterior limb bud causes the upregulation of Gremlin, a BMP inhibitor [268, 269].

Layout of this Dissertation

Above, I have introduced the main themes of this dissertationF. In the following

chapter, I describe the methods used throughout this dissertation. Chapter 3 details my clinical experience as part of the Colorado Clinical Translational Science Institute (CCTSI).

An abridged version of Chapter 4 combined with the first section of Chapter 5 (OC-Cre) is under revision in Disease Models and Mechanisms as “Increased FGF8 Signaling Shifts Cell

Fate from Osteogenic to Chondrogenic in the Developing Skull.” The later part of Chapter 5

39 continues to expand upon FGF8’s role in intramembranous vs. endochondral ossification, but

is more preliminary in nature. A version of Chapter 6, “Increased Hedgehog Signaling from

the Ectoderm and Mesenchyme Impairs Craniofacial and Limb Development” is being

prepared for publication. In Chapter 7, I will discuss some of the conclusions from my work

and how they fit into the field, as well as explore some potential future directions of this work and the direction the field may be heading.

40 CHAPTER II

MATERIALS AND METHODS

Mice

Mouse Strains

All mouse experiments were done with the approval of the Institutional Animal Care and Use Committee (IACUC) of the University of Colorado Denver; IACUC number was

B59016(02)1C. Embryonic day 0.5 was considered to be noon on the day a copulatory plug was found.

The Williams lab generated two new Rosa26-based Lox-Stop-Lox alleles that can be used to regulate the levels of FGF8 expression under the control of Cre recombinase transgene activity—R26F8 and CAGF8. R26F8 GOF mice have an allele with the endogenous Rosa-26 promoter separated from FGF8 by a LoxP STOP LoxP cassette. When the Stop cassette is removed by a Cre recombinase, mice express moderate levels of FGF8

expression in a tissue specific manner. Similarly, CAGF8 GOF mice have an allele with the strong CAG promoter separated from FGF8 by a LoxP STOP LoxP cassette. Given the stronger promoter activity of CAG [270], Cre recombinase mediated removal of the stop cassette leads to mice expressing high levels of FGF8.

The R26F8 and CAGF8 alleles are shown in detail in Figure 14 and Figure 15. Below

I describe their derivation by Drs. Melvin and Williams as this information is also described

in my FGF8 paper that is under revision. For R26LSL Fgf8b (R26F8), an ~0.8kb mouse Fgf8b cDNA was PCR amplified using Elongase (Thermo Fisher Scientific Waltham, MA) from a plasmid vector provided by Dr. Mark Lewandoski (NCI) using the forward primer Fgf8b

FWD and the reverse primer Fgf8b REV (all primer sequences are provided in Table 3).

41 Table 3. Primers. Primer name, purpose, and sequences for all primers utilized.

Primer Name Purpose Sequence

Fgf8b FWD Cloning GGATC CCTCG AGCGC GCCAT GGGCA GCCCC CGCTC C Fgf8b REV Cloning CGAGC TGAAG CTTCG CCTAT CGGGG CTCCG GGGCC CAAG BTR ES Screening/ CTAGA GCGGC CTCGA CTCTA CGATA Genotyping CCGTC GATCC CC R5F ES screening, GGCTG TGCTT TGGGG CTCCG GCTCC outside 5' homology TCAG GFP F ES screening CCAAC GAGAA GCGCG ATCAC ATGGT CCTGC TGGAG TTCGT G R3R ES screening, CCTCA GAGAA ATGGA GTAGT TACTC outside 3' homology CACTT TCAAG TTCC CMV R1 ES screening/ CGTTG GGCGG TCAGC CAGGC GGGCC Genotyping ATTTA CCG Rosa F Genotyping GGGAG TTCTC TGCTG CCTCC TGGCT TCTGA GG Rosa R Genotyping CCTGC AGGAC AACGC CCACA CACCA GG Cre1 Cre Genotyping GCTGG TTAGC ACCGC AGGTG TAGAG

Cre3 Cre Genotyping CGCCA TCTTC CAGCA GGCGC ACC

oIMR0039 LacZ Genotyping ATCCT CTGCA TGGTC AGGTC

oIMR0040 LacZ Genotyping CGTGG CCTGA TTCAT TCC

Fgf8-F qRT-PCR CCGGA CCTAC CAGCT CTACA

Fgf8-R qRT-PCR GGCAA TTAGC TTCCC CTTCT

Bactin-qF qRT-PCR GCGAG CACAG CTTCT TTG

Bactin-qR qRT-PCR CCATG TTCAA TGGGG TACTT C

This procedure introduced an XhoI site just prior to an ATG start codon and a HindIII site downstream of the stop codon. Following subcloning into TA vector (Thermo Fisher

42 Scientific Waltham, MA) and sequence confirmation, the insert was digested with HindIII and this site blunted with Klenow fragment of DNA polymerase I in the presence of all 4 dNTPs (all enzymes for cloning were obtained from New England Biolabs, Ipswich, MA).

Subsequently, the insert was released using XhoI digestion and cloned into the XhoI and

SmaI digested pBTG vector (pBigT-IRES-GFP; Addgene, Cambridge, MA). Next, this insert was released with PacI and AscI and subcloned into the vector pRosa26PAm1

(Addgene) using the same enzymes to generate the R26F8 GOF targeting vector.

For R26LSL CAG Fgf8b (CAGF8), an ~1.7kb SalI – EcoRI fragment containing the CAG promoter sequence was isolated from the plasmid CAG-GFP (Addgene) and cloned into a vector containing a PacI site upstream of the SalI site as well as NotI and NheI sites downstream of the EcoRI site. Subsequently, the NotI – NheI restriction fragment from

PGKneotpAlox2 (Addgene) which contains the floxed selection cassette was cloned downstream of the CAG promoter. Finally, the new PacI – NheI insert fragment was inserted into the R26F8 targeting vector to generate the CAGF8 GOF targeting vector. Both targeting vectors were then linearized with MluI, gel purified, and then electroporated into

129S1/Sv W9.5 ES cells.

R26F8 ES cell clones were screened using the primer pairs (RF5 + BTR) and (GFP F

+ R3R) to detect appropriate homologous recombination with the endogenous R26R locus at the 5' and 3' end respectively. PCR reactions employed KOD Hot Start DNA Polymerase as recommended by the manufacturer (EMD Millipore, Billerica, MA). Only correctly targeted clones produced a 1.35kb band (5' end) and ~4.8kb band (3' end) and these were karyotyped prior to injection into blastocysts. CAGF8 ES clones were screened in analogous fashion using the primer pairs (RF5 + CMV R1) and (GFP F + R3R) and produced similar sized

43 bands as the R26LSL Fgf8b clones at both the 5' and 3' ends following homologous recombination.

Following germline transmission, both the R26F8 and CAGF8 mice were maintained

on an outbred Black Swiss background and eventually bred to homozygozity and then

maintained as homozygous colonies. Creface and Crect mice were also generated by the

Williams lab and have been described previously [54, 271, 272]. Msx2-Cre mice were

kindly supplied and originally described for their use in limb studies by Rob Maxson [273].

Additional mouse stains were obtained from Jackson Laboratory (Bar Harbor, ME) including

Rosa26 Cre reporter mice (B6.129S4-Gt(ROSA)26Sortm1Sor/J), Axin2lacZ (B6.129P2-

Axin2tm1Wbm/J), BAT-GAL (B6.Cg-Tg(BAT-lacZ)3Picc/J), OC-Cre (B6N.FVB-

Tg(BGLAP-cre)1Clem/J), Col2-Cre (B6;SJL-Tg(Col2a1-cre)1Bhr/J), and SmoM2 mice

(Gt(ROSA)26Sortm1(Smo/EYFP)Amc/J). Note that: (1) as mice with Cre alleles are being bred to

homozygous GOF mice, all offspring will inherit a GOF allele and mutants with an activated

GOF allele can be identified by the presence of a Cre allele; and (2) the combination of the

Cre alleles and the GOF alleles leads to sirenomelia/severe limb defects or perinatal lethality;

therefore it is not possible to generate mice containing these Cre recombinase transgenes

with homozygous recombined GOF alleles.

Genotyping

PCR based genotyping was performed using DNA extracted from yolk sacs or tail

clips using DirectPCR Lysis Reagent (Viagen Biotech. Inc, Los Angeles, CA) plus 10 ug/ml

Proteinase K (Roche, Basel, Switzerland) followed by heat inactivation at 85°C for 45 min.

Mutants were identified via PCR using the Qiagen DNA polymerase kit, including the

optional Q Buffer solution (Qiagen, Valencia, CA). For a typical PCR reaction, refer to the

44 appendix. Mice carrying Crect, Creface, Col2a1-Cre, Msx2-Cre or OC-Cre transgenes were

identified using the primer pair Cre1 and Cre3 at an annealing temperature of 70°C, yielding

a band at ~450bp. R26R mice and the Axin2lacZ allele were identified using the primer pairs

oIMR0039 and oIMR0040. The R26F8 allele was identified using the primer pair Rosa F

and BTR, whereas the primer pair Rosa F and CMV R1 was used to identify the CAGF8

allele. In both cases, Rosa F + Rosa R was used to identify for the presence of the wildtype

allele (Table 3).

Skeletal Analysis

Bone & Cartilage Staining

Embryos, pups, and adult mice were collected at appropriate time points and

processed as previously described [274]. In brief, following euthanasia and the removal of

the skin and organs, the mice were dehydrated in ethanol for a minimum of three days before

being incubated in acetone (at least two days). Subsequently, they were incubated in staining

solution composed of ethanol (70%), acetic acid (5%), alcian blue (0.3%), and alizarin red

(0.1%) at 37°C for a minimum of 5 days before being cleared in 2% KOH. In skeletal

preparations where toluidine blue was utilized, 0.1% toluidine blue was added to the above

staining solution for one hour before clearing.

Cartilage Staining

Embryos (E13.5-E16.5) were collected and processed as previously described [275].

In brief, embryos were fixed in Bouin’s solution for two hours followed by a series of washes in a solution of 70% ethanol plus 0.1% NH4OH until there was no remaining yellow

(Bouin’s) color. Next, the embryos were equilibrated in 5% acetic acid (2x 1 hr washes) and incubated overnight in a solution of alcian blue (0.05%) and acetic acid (5%). The embryos

45 were then washed twice with 5% acetic acid (~1 hr washes) and twice with methanol

(minimum 1 hr washes). Finally, the embryos were cleared in BABB (1:2 benzyl

alcohol/benzylbenzoate).

Section Histochemistry

For sections from E18.5 embryos, as well as P0 and P12 pups, the embryos/pups were

fixed in 70% ethanol (skin removed from P12 pups first). They were then sent to the Yale

Orthopedic Histology and Histomorphometry Laboratory for plastic embedding and

sectioning, as well as staining. Four stains were utilized: Goldner’s Trichome [276], von

Kossa [277], Alcian Blue + PAS (Periodic Acid-Schiff) [278], and Toluidine Blue [279].

Suture widths were calculated as the average of 6 sections—2 sections each from 3 biological replicates.

In Situ Hybridization

Preparation of Probes

Probes were generated I by cloning a unique fragment into a TOPO vector (Life

Technologies, Grand Island, NY), using cDNA synthesized from mouse embryonic mRNA, as a template. cDNA was generated using the Superscript® III First-Strand Synthesis System

(Life Technologies, Grand Island, NY), as per manufacturer’s instructions. Sequence verified plasmids were linearized and antisense probes synthesized using an appropriate DNA- dependent RNA polymerase (T7/T3/SP6) and DIG RNA labeling mix (Roche, Basel,

Switzerland).

Wholemount In Situ Hybridization (WMISH)

I Probes were generated by past and present members of the Williams lab. 46 Recipes and solution abbreviations are found in the appendix. Whole embryos

(E10.5-E12.5) were dissected and fixed overnight at 4°C in 4% paraformaldehyde (PFA), washed in PBT, and serially dehydrated to 100% methanol and stored at -20°C until use. On the day of in situ hybridization, embryos were rehydrated and treated with 10 µg/ml

Proteinase K in PBT (time varied depending on the stage of the embryo, for E10.5 it was a 20 minute incubation; E11.5 and E12.5 embryos had longer incubations of 25 and 30 minutes, respectively). Embryos were then washed in PBT and re-fixed in 4% paraformaldehyde plus

0.1% glutaraldehyde for 20 minutes at room temperature, followed by another wash in PBT.

For hybridization, embryos were rinsed in hybridization mixture and then incubated in hybridization mixture for three hours at 65°C before a DIG-labelled probed was added to the hybridization mix to 1 µg/ml and incubated at 65°C overnight.

Embryos were rinsed twice in pre-warmed (65°C) hybridization mix and then washed twice with pre-warmed hybridization mix for 30 minutes each at 65°C, followed by a wash in a 1:1 mixture of hybridization mix and MABT (pre-warmed) for 20 minutes at 65°C. After three brief rinses in MABT, the embryos were washed twice with MABT for 30 minutes at room temperature and pre-blocked with 20% sheep serum and 2% Boehringer Blocking

Reagent in MABT for three hours at room temperature. The anti-DIG antibody (Roche) was then added to the pre-block solution to a final concentration of 1: 2000 and incubated at 4°C overnight. The following day, the embryos were rinsed three times in MABT and then washed three times in MABT for one hour each with a final wash overnight. Following the

MABT washes, the embryos were washed twice with Alkaline Phosphatase buffer for 1 hour each and developed in BMP purple reagent (Roche). After the color had developed to the

47 desired extent, embryos were washed in PBT to stop the reaction and fixed in 4% PFA. In all

experiments, control and mutant embryos were hybridized and developed together.

Section In Situ Hybridization

Embryos were fixed in 4% paraformaldehyde overnight at 4°C and then incubated in

30% sucrose in PBS at 4°C until the embryos sank to the bottom of the well (overnight to

several days). The embryos were then incubated in a 1:1 solution of 30% sucrose: OCT

(Sakura Finetek, Torrence, CA), rocking at 4°C, overnight for several days. Finally, the embryos were transferred to 100% OCT and rocked at 4°C for at least half an hour before being embedded in OCT on dry ice and stored at -80°C until sectioning. Sections were cut at

10μm on a Leica CM cryostat (Leica Biosystems Inc., Buffalo Grove, IL) and mounted on

APES (aminopropyltriethoxysilane) coated slides. APES coated slides led to better retention of sections during processing than traditional charged glass slides and were made as previously described (Matt Lewis, www.methodbook.net/probes/insitu.html).

Subsequently, sections were pre-hybridized in slide mailers as follows: 1) Fixed 10 minutes in 4% PFA/PBS at room temperature, followed by 3 washes in PBS for 3 minutes

each; 2) Digested in Proteinase K (1 µg/ml in 50mM Tris pH 7.5, 5mM EDTA) for 4

(E12.5), 6 (E14.5), or 8 (E16.5) minutes; 3) Re-fixed in 4%PFA/PBS for 5 minutes at room

temperature, followed by 3 washes in PBS, 3 minutes each; 4) Acetylated (1.36% triethanolamine, 0.178% HCL, 0.2544% acetic anhydride in water) for 10 minutes at room temperature, followed by 3 washes in PBS at room temperature for 5 minutes each; 5)

Incubated in hybridization buffer at 55°C for 1-2 hours. Next, sections were hybridized as

follows: 1) Incubated in hybridization buffer plus 1 ng/µl probe overnight at 70°C; 2)

Submerged in prewarmed (70°C) 5x SSC, pH7 and incubated on rocker for 30 minutes at

48 room temperature; 3) Incubated in prewarmed (70°C) 0.2x SSC (pH7) for 3 hours at 70°C,

followed by incubation in fresh 0.2x SSC for 5 minutes at room temperature; 4) Incubated in

1x MAB (Maleic Acid Buffer, pH 7.5) for five minutes at room temperature; 5) Incubated in blocking solution (2% Blocking reagent (Roche Ref:11096176001), 10% heat inactivated sheep serum, 0.1% Tween-20, all in 1x MAB) for 1 hour room temperature, followed by incubation in blocking solution plus anti-DIG antibody (1:5000, FAB fragments, Roche) overnight at 4°C. Finally, sections were washed and stained as follows, all at room temperature: 1) Washed 3x in 1X MAB with 0.1% Tween-20 for 15-30 minutes; 2) Washed in DEPC- H2O with 0.1% Tween-20 for 20 minutes; 3) Slides were removed from mailers and 200 µl of BM Purple (Roche) with 0.1% Tween-20 was added to each slide; 4) Slides were kept in dark until desired signal observed; 5) Slides were counterstained with Nuclear

Fast Red (Vector Laboratories, Inc., Burlingame, CA) for 10 minutes and rinsed with water.

Finally, coverslips were applied with Fluoromount-G (Southern Biotech, Cat no. 0100-01).

Skin Barrier Analysis

Toluidine Blue

Toluidine blue was used as a marker for skin barrier function in E16.5-E18.5 embryos

as well as neonates. Pups were euthanized and then incubated at room temperature in a

solution of 0.1% toluidine blue in water for 5-60 minutes, depending on embryo/pup age.

Finally, the pups were washed in water and then PBS before imaging.

Alkaline Phosphatase Staining

Sections were first washed in PBS at room temperature for 10 minutes and then

washed in DEPC- H2O with 0.1% Tween-20 for 20 minutes. Next, 200 µl of BM Purple

49 (Roche) with 0.1% Tween-20 was added to each slide and the slides were kept in dark until

the desired signal was observed. Finally, slides were counterstained with Nuclear Fast Red

(Vector Laboratories, Inc., Burlingame, CA) for 10 minutes and rinsed with water.

Coverslips were applied with Fluoromount-G (Southern Biotech, Cat no. 0100-01).

β-galactosidase Staining

Whole Mount

β-galactosidase staining of whole embryos was carried out as follows. Embryos were fixed for 1 hr at 4°C in 4% paraformaldehyde in PBS. Next, they were rinsed 3 x 10-30 minutes at room temperature in lacZ rinse buffer (0.2 M sodium phosphate, pH 7.3; 2 mM

magnesium chloride; 0.02% NP40; 0.01% sodium deoxycholate). They were then rocked

overnight in the dark at room temperature in lacZ staining solution (lacZ rinse buffer

containing 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and 1 mg/ml X-gal).

Finally, the reaction was stopped by transferring the embryos to PBS and fixing in 4%

paraformaldehyde. If bone was to be visualized, skin was removed prior to fixation in

embryos E15.5. and older and after staining the embryos were incubated in 1% KOH until

the surrounding tissues were degraded enough to visualize the bone.

Sections

β-galactosidase staining of frozen sections was carried out as described previously

[281]. In brief, embryos were fixed in 0.2% glutaraldehyde for 30 minutes at room

temperature. They were then soaked at 4°C in 10% sucrose in PBS for 30 minutes, followed

by PBS plus 2 mM MgCl2, 30% sucrose, and 50% OCT for two hours, at 4°C. Next, the

embryos were embedded in 100% OCT on dry-ice. Sections were cut at 10 μm and mounted

on charged glass slides, after which the slides were fixed in 0.2% glutaraldehyde for 10 min

50 on ice, rinsed briefly with 2 mM MgCl2 in PBS, and washed in 2 mM MgCl2 in PBS for 10

min, again on ice. Finally, the sections were incubated in detergent rinse solution (0.005%

NP40, 0.01% sodium deoxycholate in PBS) for 10 min at 4°C and stained in X-gal staining

solution (detergent rinse solution plus 1 mg/ml X-Gal, Invitrogen/Life Technologies,

Carlsbad, CA) for two- three days in the dark at room temperature. Following staining,

sections were counterstained with Nuclear Fast Red (Vector Laboratories, Inc., Burlingame,

CA) for 10 minutes and rinsed with water. Finally, coverslips were applied with

Fluoromount-G (Southern Biotech, Cat no. 0100-01).

RNA Quantification

Realtime PCR (qRTPCR)

Three embryos from each group -- 1) MR26F8 mutants, 2) MR26F8 littermate controls, 3) MCAGF8 mutants, and 4) MCAGF8 littermate controls—were collected (12

total). Skin samples were taken from the region between the eyes and the ears of E18.5

embryos. This region correlates with the region where Msx2-Cre was expressed (Figure

16C) and, thus, the levels of Fgf8 should vary between the controls and mutants. RNA was

extracted from the skin using the RNeasy Fibrous Tissue Mini Kit (Qiagen Cat. No. 74704),

following the manufacturer’s instructions, including the DNAse digestion step. Additionally,

to ensure DNA removal, the extracted RNA was treated using the TURBO DNA-free Kit

(ThermoFisher Scientific Cat. No. AM1907). For each sample, 500 ng of RNA was reverse- transcribed to complementary DNA (cDNA) using SuperScript III First-Strand synthesis kit

(Invitrogen Cat. No. 18080051). Real-time PCR reactions were run using SYBR Select

Master Mix (Applied Biosystems Cat. NO 4472908; Austin, TX) on CFX Connect Real-time

System (BioRad, Hercules, CA). The expression of Fgf8 was normalized to that of

51 corresponding β-actin, a housekeeping gene; primers utilized are found in Table 3. The

wildtype littermate controls from each group were not found to be statistically significantly

different (p=0.678) and thus all six littermate controls were pooled and expression

normalized to one. The expression of Fgf8 in the MR26F8 and MCAGF8 mutants was

normalized to the combined control sample and presented as fold expression change. Error

bars show standard deviation of the fold change.

RNAseq

For the bone and cartilage RNA-seq, MCAGF8 E14.5 embryos were dissected in ice- cold PBS. From the MCAGF8 and control embryos, the skin was peeled away from the skull

and the underlying cranial vault tissue was collected. Subsequently, for the control embryos,

the brain was removed and the cartilaginous sections of the cranial base were carefully

collected as well. All tissues from the MCAGF8 (cranial vault) and control embryos (cranial

vault and cartilaginous cranial base) were stored in RNAlater (Ambion/Life Technologies)

until RNA extraction. RNA was extracted from 9 MCAGF8 and control cranial vaults as

well as 9 cranial bases using RNeasy Lipid Tissue Mini Kit (Qiagen Cat. NO 74804)

following the manufacturer’s protocol. Within each group, the 9 samples were pooled into

three groups, with 3 samples making up each group, so that the RNA concentrations were

similar. Samples were submitted to the University of Colorado Denver Genomic and

Microarray Core and sequenced using the Illumina HiSeq2500 Platform and single end reads

(1x50). Reads generated were mapped to the mouse genome by gSNAP, expression derived

by Cufflinks, and differential expressed analyzed by ANOVA in R, as described previously

52 II[282]. DAVID [283, 284], with default parameters, was used for functional annotation clustering of significantly upregulated and downregulated genes.

II gSNAP, Cufflinks, and ANOVA analysis performed by Kenneth L. Jones. 53 CHAPTER III

CCTSI CLINICAL EXPERIENCE

In addition to my laboratory research, I was given the opportunity to gain clinical experience as part of the Colorado Clinical Translational Science Institute’s (CCTSI) Pre-

Doctoral Fellowship program. One of the requirements of CCTSI is a chapter in my dissertation detailing my experiences. CCTSI afforded me the opportunity to interact with patients under the guidance and mentoring of Dr. Charles Corbett Wilkinson, a pediatric neurosurgeon, at Children’s Hospital of Colorado, as well as his collaborator Dr. Brooke

French, a plastic surgeon. I had the opportunity to observe these doctors both in surgery and interacting with patients and their parents in the craniofacial clinic. The surgical cases included patients undergoing surgery to correct craniosynostosis. In craniofacial clinic, patients had many different craniofacial abnormalities, including, but not limited to craniosynostosis. The clinical shadowing experience provided firsthand knowledge of craniofacial diseases and ideas for future clinical applications to my research.

Craniosynostosis

As described in this dissertation’s introduction, craniosynostosis is the premature

fusion of the cranial sutures, which lie between the cranial bones. This premature fusion

causes distortion of the skull shape due to a combination of lack of growth perpendicular to

the fused suture and compensatory overgrowth at the non-fused sutures [38]. A full-term

infant has nearly 40% of her or his adult brain volume, which rapidly increases to 80% by

three years of age. Correspondingly, the infant cranium is 40% of its adult size; by seven

years, this increases to 90%. [285]. Thus, surgical correction is typically done during infancy

54 to allow for normal brain and skull growth, with the majority of surgical procedures

performed on patients between the ages of 3 to 12 months. Though mortality rates are low,

potential complications include massive blood loss or air embolism during surgery [286].

Follow-up of the patient post-surgery is necessary to ensure that the sutures do not re-fuse

and that there are no symptoms of increased intracranial pressure [285]. The following section describes my experiences observing both craniosynostosis surgeries and the subsequent follow-up.

Interactions with Patients in Clinic

The patients observed during my time in clinic were either 1) undergoing surgery to

correct craniosynostosis or 2) visiting the craniofacial clinic for follow up visits after surgery

or consultations prior to surgery/treatment. I will first describe my experiences observing

craniosynostosis surgeries.

Surgical management of craniosynostosis has three major objectives: to correct the

skull deformity, prevent its progression, and reduce the future risk of raised ICP [38]. There are three types of surgeries that are used to correct craniosynostosis: cranial vault remodeling, endoscopic strip craniectomy, and distraction osteogenesis. I had the opportunity to observe the first two and thus they will be the focus of this section. Cranial

vault remodeling involves the excision of the bones surrounding the affected cranial suture,

which are then remolded and returned to the head. Endoscopic strip craniectomy involves

removal of a 1 to 2 cm wide strip of bone centered on the affected suture followed by

postoperative helmet therapy [287]. I will next describe each of these surgeries in further

detail as well as my impressions from observing them.

55 While watching the cranial vault remodeling surgeries, I was struck by just how invasive these surgeries are on such young children. To perform this type of surgery, the

skin and periosteum is first peeled away from the skull and face to reveal the underlying

bones. Large segments of the cranium are then removed and remodeled by trimming and

reshaping the bones before affixing them back together with absorbable plates. The

discussions between the surgeons involving the artistry of remodeling the bones was interesting to me as I hadn’t previously considered the artistic aspect of surgery. However,

given that the surgeons are determining the future shape of the child’s head, I’m certainly

glad that they were concerned with the aesthetics of their work. Once the bones had been

satisfactorily remodeled, they were placed back into the infant’s head and sutured to the non-

remodeled bone. The skin was then returned to its normal position and also sutured back into

place.

Though this is still the most commonly performed surgical procedure for

craniosynostosis, it has several downfalls. This surgery is associated with significant blood

loss as well as lengthy surgical times (3-8 hours) and hospital stays (4-7 days). To limit

morbidity, this surgery is often delayed until the infant is 6-12 months, but this allows the

deformities to become more severe. In addition to blood loss, other complications of the

cranial vault modeling procedure include: problems associated with blood transfusion

reactions, palpable and visible deformities and asymmetries, inadequate calvarial shape

correction, and improper skull reossification [288]. Many clinicians have noted that the skull

growth patterns are not normal over time and instead revert to the original dysmorphology

[289]. Therefore, despite the surgeons’ best artistic efforts, the cosmetic results of these

surgeries are often not perfect.

56 Endoscopic strip craniectomy, on the other hand, is minimally invasive and involves significantly less blood loss and a shorter hospital stay than cranial vault remodeling. During this surgery, a 1.5-2.5 cm incision is made perpendicular to the affected suture and a bur hole is made directly over the affected suture. An endoscope is then inserted through the bur hole to allow for visualization of the dissection of the dura from the overlying skull. A relatively small (1 to 2 cm wide) strip of bone centered on the affected suture is then cut from the skull and removed. After surgery, a cranial molding helmet is fitted for the child. These helmets mold the skull into the desirable shape as the child grows by contacting all areas of the infant’s cranium, except where growth is desirable. This allows the skull growth to be modified in three dimensions and is adjustable over time [287].

Despite the significant decrease in blood loss and the need for transfusions as well as the shorter operative times and hospital stays associated with the endoscopic strip craniectomy compared to the cranial vault remodeling, the endoscopic strip craniectomy is still less commonly performed than the cranial vault remodeling surgery. The reasons for this are three-fold: physician acceptance/training, parent preference, and infant age. Many physicians do not yet offer the endoscopic strip craniectomy as they have not been trained in the procedure and/or are prejudiced against it for historical reasons. The design and subsequent modifications of the helmet are critical to the success of the procedure, but there can be a steep learning curve in successfully implementing the fitting of the helmet as it requires conceptualization of how cranial growth in certain areas will correct the deformity over time [287]. Even when the procedure is offered as an option to parents, they will often choose the cranial vault remodeling instead because it doesn’t require their child to wear a helmet for an extended period of time (up to a year). Finally, the strip craniectomy is

57 typically considered more appropriate for infants younger than three months of age. While

this offers the advantage of correcting the deformities before they become more severe, not

all cases of craniosynostosis are diagnosed before 3 months of age and parents are also more

unlikely to approve surgery in younger infants. Regardless of the type of surgery chosen, the

patients will need follow-up until brain growth is complete. Next, I will discuss my observations on the follow-up appointments of these patients.

Regular follow-up of craniosynostosis patients throughout childhood is advised so that the clinician can monitor for symptoms of raised ICP, such as headaches, behavior change, or decline in school performance. In addition to asking the children/parents about headaches, behavior, and school performance, the physicians would refer the children to an optometrist to also monitor their ICP as one of the first signs of raised ICP is bulging eyes.

The severity of the scar was variable between children and though there are treatments to lessen the severity of the scars, not all children were interested in diminishing it. In some patients, even older ones, you could see the pulsing in the head in sync with the heart beat where the skull hadn’t completely re-fused. Most patients also participated in a research study that Children’s Hospital of Colorado is conducting using 3D imaging of the head. The head is imaged prior to surgery, as well as after surgery, and as the child grows. This allows the clinicians to examine how the amount of skull removed during surgery correlates with head shape as the child grows older.

I also had the opportunity to see patients with neurological disorders that caused secondary skeletal defects. One such patient had microcephaly that was suspected to have been caused by the zika virus as her mother had visited Columbia when she was 12 weeks pregnant. The mother was concerned that her child had craniosynostosis that was restricting

58 the child’s brain growth and development and was pushing for a CT (computed tomography)

scan. Seeing the professionalism of the doctors as they carefully navigated that situation and

explained that the child’s small skull was a result of a small brain, not the other way around,

was impressive. In general, the sacrifices that families made for their children was eye-

opening. One family said they had moved from California specifically for Children’s

Hospital. Another was about to move to Sweden, but said they would return if the medical

care wasn’t high enough quality. Additionally, some children with syndromes often had

several medical appointments in a week, of which one or both parents brought them to. It

was apparent that this was a particular burden for mothers, as when only one parent attended

the appointment, it was almost always the mother rather than the father.

I also learned that information on the presence of specific mutations helps in

determining prognosis. For example, there was a higher risk of repeat surgeries and

persistent deformities in patients with coronal craniosynostosis if they had the Muenke

syndrome mutation in FGFR3 (C749G) [290, 291]. Additionally, patients with TWIST1 mutations have a higher risk for developing ICP [292] and large TWIST1 deletions are associated with a higher risk of learning disability [293, 294]. Thus, many believe genetic testing should be standard of care as it contributes to both risk assessment for the family and prognostic information for the patient. As causative mutations have only been identified in

25% of cases, the identification of novel craniosynostosis causing mutations could lead to better clinical treatment. While some cases are likely environmentally induced and do not have a causative mutation, whole genome assessment of copy number changes and DNA sequencing are likely to identify further predisposing loci.

59 Mice as a Model to Study Craniosynostosis

There are two main ways craniosynostosis is studied: primary osteoblasts derived

from craniosynostosis patients and the creation of mouse models. Mouse models allow for

the identification of new craniosynostosis causing genes, elucidation of the roles of genes in

craniosynostosis, and experimental analysis of craniosynostosis phenotypes. Thus far, mouse

models of activating Fgfr mutations and of Twist1 LOF have been particularly invaluable to our understanding of coronal craniosynostosis and the role of the neural crest/mesoderm boundary in forming this suture [295, 296]. Additionally, mouse models have allowed researchers to examine gene expression within the sutures; gene expression data is critical to our understanding of the connection between specific mutations and the resulting craniofacial dysmorphologies and pattern of suture fusion. Again, this has been exemplified by the

Fgfr1-3 and Twist1 mouse models which have distinct expression patterns within their sutures [297]. Which sutures develop craniosynostosis may be due to the suture specific expression of certain genes. For example, in the mouse model of Greig

cephalopolysyndactyly syndrome, the Gli3Xt-J/Xt-J mice, which represent a Gli3-null allele,

exhibit craniosynostosis of the lambdoid suture; the lambdoid suture is notably the site of

strong embryonic calvarial Gli3 expression [298]. A more thorough understanding of mouse

sutural gene expression, particularly at embryonic stages, may lead to the identification of

additional human craniosynostosis candidate genes. Furthermore, potential therapeutic

interventions can be tested on the mouse models. Chemical inhibition of FGFR tyrosine

kinase activity as well as activity of effector kinases downstream of FGFRs ameliorates

craniosynostosis in mouse models of Crouzon, Apert, and Beare-Stevenson syndromes [299–

302]. Finally, while the premature suture fusion seen in craniosynostosis is the most

60 commonly studied and treated aspect of craniosynostosis, craniosynostosis also refers to the

abnormal skull bone development that is associated with dysmorphic skull shape. Again,

mouse models are useful in studying this aspect of craniosynostosis as the abnormal skull

bone shapes can often be detected prior to the fusion of the sutures [303, 304]. Thus, mouse

models for craniosynostosis have already provided multiple insights into the mechanisms

behind craniosynostosis and will continue to be a valuable tool in the future.

Conclusions

During my time in clinic, I learned invaluable information about craniofacial disorders, surgical interventions to treat craniosynostosis, and patient care. Seeing the burden craniofacial disorders placed on the families of these children, I was constantly reminded of the critical need for prevention and less invasive treatments for these conditions.

My clinical experience provided inspiration on how my research in craniosynostosis and cranial bone development in mice might be applicable to patients with craniofacial disorders.

Additionally, during the course of my shadowing, I also started the process of planning a

clinical study using tissue samples from craniosynostosis patients. While this study never

reached fruition due to limitations in timing and funding, it provided me with invaluable

insight into the process of planning a clinical study as well as experience serving as a liaison

between basic scientists, clinicians, clinical research associates, and regulatory bodies. My research investigating the role of FGF and HH signaling in cranial bone development has provided new insight into the critical role these pathways play in promoting and maintaining bone and suture development. Understanding the mechanism by which aberrations in FGF and HH signaling lead to craniosynostosis may provide a therapeutic target for preventing repeat surgeries. Although surgery is expected to remain the main treatment for the

61 foreseeable future, the identification of signaling pathways involved in the pathology of the cranial suture raises the possibility for the use of adjuvant medical therapies in the future.

62 CHAPTER IV

INCREASED FGF8 SIGNALING SHIFTS CELL FATE FROM OSTEOGENIC

TO CHONDROGENIC IN THE DEVELOPING SKULLIII

Introduction

Bone forms via two processes: intramembranous or endochondral ossification. The majority of the skeleton, including the long bones, vertebrae, and basicranium, forms via endochondral ossification during which condensed mesenchyme cells first differentiate into chondrocytes that form cartilage tissue. This intermediate cartilaginous template is then replaced by bone, formed through osteogenesis. In contrast, most of the skull, including the jaw and cranial vault, is generated via intramembranous ossification, in which condensed mesenchyme cells directly differentiate into osteoblasts that form bone without any cartilaginous precursor [104].

The development of the cranial vault is a complex process. The anterior bones of the cranial vault (nasal, frontal) are derived from cranial neural crest cells [118, 119] whereas the posterior portion (parietal) is derived from the paraxial mesoderm [119, 120]. Together, these cranial neural crest and paraxial mesoderm derived cells form the intramembranous bones and sutures of the skull [119, 121–123]. Intramembranous ossification of the skull vault involves direct bone matrix deposition to form calvarial plates, which expand during development but do not fuse with other cranial bones during embryogenesis and infancy

[124]. Instead, sutures connect the individual intramembranous bones and serve as growth

III An abridged version of Chapter 4 combined with the first section of Chapter 5 (OC-Cre) is under revision in Disease Models and Mechanisms as “Increased FGF8 Signaling Shifts Cell Fate from Osteogenic to Chondrogenic in the Developing Skull.” 63 centers that regulate the expansive growth of the skull [125]. Because sutures are the major

sites of bone growth during cranial vault development, signaling at the sutures is essential for

the regulation of intramembranous ossification [128, 129]. Several signaling pathways are

implicated in proper skull ossification and growth including fibroblast growth factor (FGF), hedgehog (HH), and WNT signaling [133]. The role of FGF signaling in ossification is of particular interest as mutations in the FGF signaling pathway, composed of four FGFRs

(Fibroblast Growth Factor Receptors) and 22 FGF ligands [101], cause a number of skeletal disorders [125, 128, 305], including those that affect cranial vault ossification; such as craniosynostosis.

Craniosynostosis is the second most common human craniofacial abnormality, occurring in 1 in 2,100 to 2,500 births [36, 37]. It is characterized by the premature fusion of the metopic, sagittal, lambdoid, and/or coronal sutures which causes a distortion in skull shape that frequently requires surgical correction [38]. Activating mutations in the FGF

signaling pathway, mostly heterozygous mutations of FGFRs, account for the majority of the

known causes of craniosynostosis [46]. The etiology of craniosynostosis is complicated,

however, by the fact that activating mutations in the same gene can cause a variety of

different syndromes and/or types of craniosynostosis. For example, mutations in FGF

receptor 2 (FGFR2 ) can cause seven of the eight FGFR-related craniosynostosis disorders,

including Apert, Crouzon, and Pfeiffer syndromes [169].

This diversity in part reflects how each specific mutation alters the FGFR domains

and thus impacts the multiple signaling pathways that are directly regulated by FGFRs.

FGFR mutations leading to craniosynostosis are all considered to be activating mutations.

Classically, these FGFR activating mutations have been thought to lead to ligand-

64 independent receptor dimerization. However, certain mutations in FGFR’s, including one

associated with Apert’s syndrome, do not cause ligand-independent dimerization, but rather

lead to increased affinity for specific FGF ligands [176, 306]. Additionally, ligand- independent and ligand-dependent FGFR mutations are associated with differential regulation of downstream genes [306]. Further complicating matters, identical mutations in

FGFR2 have been found in patients with different diagnoses (Crouzon, Pfeiffer, and Jackson-

Weiss syndromes), suggesting genetic modifiers may play a role [170, 171, 307, 308]. These

genetic modifiers likely contribute to the variation in craniosynostosis phenotypes by altering

the efficiency of the receptor and thus the dose of FGF signaling and/or by altering the timing

or location of the receptors activation.

Thus, the specific mechanisms by which FGF signaling affects skull development and

craniosynostosis remain poorly understood, but the properties of FGF ligands and receptors,

along with dosage, location, and timing, likely all contribute to the complexity of

craniosynostosis etiology. Craniosynostosis-causing FGFR mutations in humans are

heterozygous [46]. Appraising how the degree of FGF signaling activation influences

phenotype is, therefore, difficult. Using mouse models of craniosynostosis, with both

dominant hetero and homozygous mutations, has begun to bridge that gap by providing

evidence that cranial ossification is responsive to the degree of activation of the FGF

signaling pathway [309]. Notably, mice heterozygous or homozygous for activating

mutations in FGFRs have different craniofacial phenotypes, with homozygous gain-of-

function mutations often phenocopying aspects of the loss-of-function (LOF) phenotype

[309, 310]. However, quantitative assessment of the degree of FGF signaling activation,

65 based on expression of FGF responsive genes and signaling pathways, within and between

the different FGFR mouse models is lacking.

Further complicating the study of how FGF activation influences FGF responsive

genes and their downstream signaling pathways, each FGFR is responsive to multiple FGF

ligands. Similarly, FGF ligands can activate multiple FGFR’s. Identifying the role of

individual FGF’s on ossification has mainly been approached via genetic modification

studies of a specific FGF [311–313]. A difficulty with this approach, however, is that

redundancy between ligands makes it difficult to identify their physiologic roles [311]. One

strategy for overcoming this barrier is to overexpress the ligands in regions of known

expression. Thus far, only FGF 2, 9, and 18 have been shown to act in normal mouse cranial

vault ossification [190, 191, 311, 314]. A better understanding of the role of specific FGF ligands in cranial vault ossification is necessary to elucidate the role of FGF signaling on the various presentations of craniosynostosis.

To further investigate the role of FGF ligand dosage in cranial vault development, I utilized novel mouse models that could be induced to overexpress different levels of Fgf8b in the ectoderm of the developing skull. The choice of the FGF8 ligand was based both on its ability to influence development and patterning in multiple contexts as well as its homology to FGF18, which has an important role in chondrogenesis and osteogenesis [190, 191, 311,

315]. Furthermore, FGF8 activates the IIIc splice variants of FGFRs [316, 317] which are expressed in the mesenchyme, where bone formation originates. Given that FGF ligands and

FGFR isoforms interact with each other in a paracrine manner, [309, 318, 319], FGF8 expressed in the developing craniofacial ectoderm and/or brain may impact cranial vault development. Specifically, I employed the Fgf8b isoform since this is more potent than

66 Fgf8a in many developmental processes, a finding which correlates with the higher affinity

of FGF8b for FGF receptors [85, 154, 320]. By comparing skeletal formation in these new

mouse models, I demonstrate that the dosage of Fgf8 expression differentially affects

development of the calvaria and the associated sutures. Moderate Fgf8b expression mimics

the craniosynostosis seen with particular FGFR2 activating mutations. In contrast, higher

Fgf8b expression severely disrupts the process of intramembranous bone formation. Overall,

these results provide insight into the mechanisms and potential treatment of craniosynostosis

as well as the function of FGFs in skeletal development and ossification.

Results

Generation of New Alleles for Differential Expression of Fgf8

Two new alleles were generated to manipulate the dosage of Fgf8 signaling in vivo

IV. The first allele incorporates the Fgf8b cDNA into the Gt(ROSA)26Sor locus under the control of a Lox-Stop-Lox cassette (Figure 14A and Figure 15A). Breeding mice with this

R26LSL Fgf8b allele to mice expressing a Cre recombinase transgene allows the deletion of the

Stop cassette to generate a new allele expressing the Fgf8b cDNA (R26Fgf8b), hereafter termed R26F8. To ascertain the effects of higher levels of Fgf8 overexpression, an additional

construct was generated - R26LSL CAG Fgf8b (Figure 14B and Figure 15A). Following Cre-

mediated recombination this would generate the R26CAG Fgf8b allele, hereafter termed CAGF8.

I initially focused on how elevated FGF8 ligand expression in the ectoderm affected limb development, but in doing these studies uncovered an unexpected skull phenotype.

Therefore, I primary focused on how FGF8 in the overlying ectoderm of the cranium could

IV The R26F8 and CAGF8 mice were generated by former Williams lab member, Dr. Vida Melvin. 67

Figure 14. Detailed map of R26F8 and CAGF8 constructs.

68

Figure 14. Detailed map of R26F8 and CAGF8 . A: Schematic of the R26F8 allele

Top. The targeting vector showing the standard homology arms used for targeting the Gt(ROSA)26Sor locus (ROSA26). The ROSA26 locus provides a promoter and a non- coding first exon to generate mRNA transcripts. The splice acceptor site (SA), LoxP sites (orange triangles), the drug selection cassette (PGK-Neo), and multiple poly A addition sequences to prevent transcription read-through (tpA), along with the IRES-GFP and the bovine growth hormone poly A addition sequences (bpA) are derived from the pBTG vector (see Materials and Methods). DTA indicates the diphtheria toxin A negative selection cassette. The sites of restriction enzymes used for cloning and linearization are shown, but note that these are not necessarily unique sites within vector sequences.

Middle. The allele after homologous recombination with the sites of primers used for ES cell screening and genotyping. Note that the primers R5F and R3R used to screen for appropriate homologous recombination lie outside the homology arms. For standard mouse genotyping ROSA F + ROSA R give an ~165bp wildtype band. ROSA F + BTR give an ~250bp band for the targeted allele.

Bottom. The allele after Cre mediated recombination that allows generation of a functional Fgf8b transcript. Note, this allele is abbreviated as R26F8 in the text. Also, note that in all instances the various DNA elements are not shown to scale.

B: Schematic of the CAGF8 allele

Top. The targeting vector as in A. except that a CAG enhancer/promoter element has been inserted upstream of the positive selection cassette. The sites of restriction enzymes used for cloning and linearization are shown, but note that these are not necessarily unique sites within vector sequences.

Middle. The allele after homologous recombination with the sites of primers used for ES cell screening and genotyping. Note that the primers R5F and R3R used to screen for appropriate homologous recombination lie outside the homology arms. For standard mouse genotyping ROSA F + ROSA R give an ~165bp wildtype band. ROSA F + CMV R1 give an ~230bp band for the targeted allele.

Bottom. The allele after Cre mediated recombination that allows generation of a functional Fgf8b transcript. Note, this allele is abbreviated as CAGF8 in the text. Also, note that in all instances the various DNA elements are not shown to scale.

69

Figure 15. Conditional alleles modulate Fgf8b expression levels.

70

Figure 15. Conditional alleles modulate Fgf8b expression levels. (A) Cartoon diagram of the two Fgf8b conditional alleles placed within the mouse Gt(ROSA)26Sor locus before and after Cre-mediated recombination with exons (Ex), splice acceptors (SA) and LoxP sites (black triangles) illustrated. The abbreviations R26F8 (top) and CAGF8 (bottom) are used to refer to these recombined alleles throughout the text of the manuscript. (B) Relative Fgf8b expression levels of in E18.5 dorsal head skin measured by qRT-PCR, normalized to Actb levels, following Msx2-Cre mediated recombination. Control indicates data from littermates lacking the Msx2-Cre transgene. Bars show average fold change for each group compared to control with error bars indicating standard error. All three groups are statistically significant from each other (Anova: p<0.001; T-test: controls vs R26F8 p= 0.0013, controls vs CAGF8 p= 0.0002, R26F8 vs CAGF8 p=0.022).

71 influence the development of the underlying mesenchyme, particularly the formation and

patterning of the craniofacial skeleton. For this purpose I tested the Cre recombinase, Msx2-

Cre, and determined that it had a spatiotemporal pattern of expression that made it suitable for studying calvarial development ([273, 321] and Figure 16); a more in-depth description of

Msx2-Cre’s expression pattern is included in the following section, 0. I next used qRT-PCR to examine how Fgf8b mRNA levels were altered when Msx2-Cre was employed with either of the new Fgf8 alleles (Figure 15B). As expected, controls (Msx2-Cre negative embryos) from R26F8 and CAGF8 litters had similar Fgf8 expression levels (data not shown). Thus,

Fgf8 expression levels of control embryos from both groups were pooled to form a single control Fgf8 expression value normalized to one. At E18.5, in the presence of Msx2-Cre,

Fgf8b expression was increased ~8 times in R26F8 mutants (Figure 15B, p=0.0013) and ~21 times in CAGF8 mutants compared with controls (Figure 15B, p=0.0002). Therefore, both the R26F8 and CAGF8 alleles increase Fgf8b mRNA expression when activated by a Cre recombinase transgene, with the latter generating ~2.5 fold more transcript than the former

(Figure 15B, p=0.022).

Analysis of Msx2-Cre Expression Pattern During Embryonic Craniofacial

Development

To investigate how elevated FGF8 expression in the overlying ectoderm could influence the development of the underlying mesenchyme, I utilized the Cre recombinase,

Msx2-Cre. To establish the spatiotemporal expression of Msx2-Cre, I examined its expression pattern by crossing Msx2-Cre to the R26R allele (Figure 16). As previously reported [153], Msx2-Cre drove recombination in the limb ectoderm by E10.5 as indicated by

LacZ expression. In addition, our findings revealed that at E10.5, low levels of punctate

72

Figure 16. Embryonic expression of Msx2-Cre during craniofacial development. Msx2-Cre mediated recombination was visualized using β-galactosidase staining (blue) of Rosa26 LacZ Reporter embryos. (A-C) Lateral views of whole mount β- galactosidase staining on E10.5 (A) and E11.5 (B) embryos. (C) At E16.5, almost the entire dorsal aspect of the head showed β-galactosidase staining (lateral view on the left, dorsal view on the right). Red and blue triangles denote regions over part of the parietal and intraparietal bones, respectively, where Msx2-Cre mediated recombination did not occur. Black lines in B and C indicate plane of section in D-G. (D-G) β-galactosidase staining of frozen sections at E11.5 in a frontal plane (D, F) and E16.5 in a sagittal plane (E, G), counterstained with nuclear fast red. F and G are magnifications of the regions outlined by black boxes in D and E, respectively. Sections are shown at both 10x (D, E) and 40x (F, G) magnification. Note in E, bone (arrow) and cartilage (ct), do not show β- galactosidase staining, whereas staining occurs in both ectoderm (boxed) and nasal epithelium (arrowhead). Scale bars: A-C: 500µM; D-E: 100µM; F-G: 50 µM.

LacZ expression began in the head, with the strongest β-gal staining being posterior to the eye (Figure 16A). By E11.5, LacZ expression was more intense throughout the head, particularly posterior and anterior to the eye. Additionally, LacZ expression was seen in the developing limb and, to a lesser extent, the dorsum (Figure 16B). Finally, by E16.5, LacZ

73 was expressed throughout the majority of the skin on the top of the head, with the exceptions

being the apex of the snout, part of the region overlapping the underlying parietal bones, and

striated regions over the interparietal bone (Figure 16C). This LacZ expression domain was

restricted to the ectoderm both early, at E11.5 (Figure 16D, F) and later, at E16.5 (Figure

16E, G). In conclusion, our findings revealed Msx2-Cre drives recombination in the embryonic cranial ectoderm. While this recombination pattern for Msx2-Cre was not originally reported, my data is in concordance with a later study that revealed that Msx2-Cre drives recombination in the cranial skin [321].

Moderately Increased Levels of Fgf8 Cause Craniosynostosis

To investigate how manipulation of Fgf8 expression in the ectoderm could impact development, I first examined mice with a moderate increase in Fgf8b expression by utilizing

R26F8 mice in concert with the Msx2-Cre transgene [153]. These Msx2-Cre;R26F8 mice, hereafter termed MR26F8, were born at normal size and in normal Mendelian ratios, but were easily identifiable from birth onwards by their gross morphology. The MR26F8 mice had both striking craniofacial defects (Figure 17A-F, Figure 18A, B) as well as limb defects

including sirenomelia and postaxial polydactyly (Figure 18C-L), the latter phenotypes

consistent with the expression of the Msx2-Cre transgene in the limb bud ectoderm (Sun et

al., 2000). The facial abnormalities included shortened snout and abnormal skin with raised

rounded protrusions that extended to cover the eye (Figure 17D-F, Figure 18B).

Additionally, the MR26F8 pups had patchy or absent hair on top of the head, limbs,

underbelly, and dorsal midline, reflecting Msx2-Cre’s expression pattern (Figure 17F; Figure

18A-H, Figure 19). Thus, although the MR26F8 mice were viable into adulthood, they

already had significant craniofacial abnormalities by birth. Therefore, I compared the

74

Figure 17. MR26F8 mice exhibit cranial defects including coronal craniosynostosis.

75

Figure 17. MR26F8 mice exhibit cranial defects including coronal craniosynostosis.

(A-F): Gross images of the control (A-C) and MR26F8 (D-F) heads. Neonate pups (P0) are shown in both a lateral (A, D) and dorsal (B, E) view. P12 heads are shown laterally (C, F). (G-L): Dorsal views of bone and cartilage staining of control (G-I) and MR26F8 (J- L) skulls at P0 (G, J), P12 (H, K), and P36 (I, L). G’-L’ show magnification of the coronal and lambdoid sutures of each respective skull. The expected location of the coronal suture is shown by a green arrow. This suture is present in controls, abnormal in MR26F8s at P0, and absent at P12 and P36. The narrower region of the MR26F8 lambdoid suture at P12 and P36 (green arrowhead) and ectopic bone (red asterisk) are shown. The yellow dashed lines in (G, J) outline the edges of ossification. The yellow dashed lines in (K, L) align with the sagittal and interfrontal sutures. Abbreviations: Fr, frontal bone; Lb, lambdoid suture; IP, interparietal bone; Pr, parietal bone; SO, supraoccipital bone. Scale Bars: D-E, J-L: 1mm; F: 5mm. Aged-matched controls and MR26F8s are at the same scale.

76

Figure 18. MR26F8 mice have striking cranial and limb defects. (A-H): Gross morphological view of P12 heads (A-B) and hindlimbs (C-D) of the control (A, C, E, G) and MR26F8s (B, D, F, H). (I-L): Bone and cartilage staining of P16 limbs of controls (I, K) and MR26F8s (J, L). (E-F, I-J): hindlimb; (G-H, K-L): Forelimb. Red arrowheads denote extra digits on the posterior side of the MR26F8 limb (postaxial polydactyly). (M-P): Skeletal staining of P0 control (M, O) and MR26F8 (N, P) mice showing lateral views of skulls and forelimbs (M, N) and cranial base after removal of the mandible (O, P). Red star in N denotes loss of ossification of the parietal. Scale Bars: 1mm.

77

Figure 19. Fur defect in MR26F8 mice. MR26F8 mice are missing fur along the dorsal midline,shown here at 9 months.

underlying skeleton of control and MR26F8 mice during early post-natal development using bone and cartilage staining (Figure 17G-L; Figure 18M-P). In the P0 wildtype skull, the

bony plates of the frontal, parietal, and interparietal bones had not yet met at the dorsal

midline and the coronal and lambdoid sutures were clearly visible as an overlapping region

between adjacent bony plates (Figure 17G, see Figure 17I for suture nomenclature). In

contrast, in MR26F8s, the sagittal and interfrontal sutures were wider except in the region

where the coronal and midline sutures (sagittal and interfrontal) should intersect (Figure 17J).

Here, there was a narrowing of the sagittal and interfrontal sutures (Figure 17J). With

respect to the lambdoid suture, the interparietal and parietal bones did not overlap to the same

extent as in the controls. Most strikingly, though, the coronal suture separating the frontal

and parietal bones, was abnormal (Figure 17J). In some areas, where a coronal suture would

be expected, there were instead wider and irregular regions where the bones did not overlap.

In other regions, particularly nearer the dorsal midline, there was no apparent suture. These

suture defects were accompanied by decreased ossification of the ventral portions of the

78 parietal bones (Figure 18N). Despite the numerous defects in the cranial vault, ossification of the cranial base remained largely unaffected (Figure 18P).

From P12 onwards, some cranial defects in the MR26F8s had resolved, whereas

others had materialized, including ectopic bone development in the base of the eye socket

(Figure 17J-L). With respect to the calvaria, in the P12 and adult wildtype skulls, the bony

plates of the frontal, parietal, and interparietal bones had met at the dorsal midline to form

distinguishable sagittal and interfrontal sutures, and the coronal and lambdoid sutures were

still apparent (Figure 17H, I). Similarly, in the MR26F8s, the skull bones had now met at the midline to form the sagittal and interfrontal sutures. However, craniosynostosis of the coronal suture in the MR26F8s was now complete (Figure 17K, L) and the lambdoid suture

was now narrower than in the wildtype. P12 and older MR26F8s also exhibited

misalignment of the frontonasal region (Figure 17K, L), with equal numbers misaligning to the left and to the right (n=15, Left=8, Right=7 p=0.7963). I speculate this misalignment may be due to the loss of interdigitation of the frontonasal suture in the MR26F8s (data not shown, [54]). Post-weaning, these mice had malocclusions and required teeth trimming to allow for proper feeding. Starting from soon after birth, MR26F8 mice were smaller than their control littermates, and this size difference grew larger with age, likely due to their altered craniofacial shape, misalignment, and malocclusions. Impaired mobility, due to the limb defects, may also contribute, though MR26F8 pups were smaller even when reared apart

from their wildtype littermates, when competition for milk access would play less of a role.

In summary, MR26F8 mice had limb patterning defects and hair defects as well as

craniofacial abnormalities, and in the remainder of this study I focus on the role of Fgf8 in

causing the latter phenotype.

79 Histological Analysis of Craniosynostosis in MR26F8 Mice

To examine the craniosynostosis phenotype in more detail, I next examined sagittal

sections of both P0 and P12 craniums V. Histological stains were used to visualize

mineralization (von Kossa) and bone differentiation (Goldner’s Trichrome) at both timepoints. At the lambdoid suture, von Kossa staining demonstrated that neonate MR26F8 mice showed a wider, more open suture (Figure 20D) than controls (Figure 20C). Examining the same suture using Goldner’s Trichrome stain (Figure 20E, F), however, revealed that these open sutures in the MR26F8s (Figure 20F) were beginning to fill with unorganized mature and immature bone matrix, whereas the narrower suture space in the controls was more defined (Figure 20E). By P12, MR26F8s had a narrower lambdoid suture (Figure 20H,

J) than the controls (Figure 20G, I), though the sutures in neither had completely fused.

Measuring the suture widths at both P0 and P12 (Figure 20K) revealed that at P0, the

MR26F8s’ suture width was ~5 times greater than controls (p<0.0001) whereas at P12 the suture width of MR26F8s was only ~30% of controls (i.e. the controls were 3.26 times as wide as the MR26F8s) (p<0.0003). Thus, the MR26F8s had delayed growth, followed by overgrowth of the bony plates associated with the lambdoid suture.

I next examined the coronal suture similarly using both von Kossa and Goldner’s staining techniques (Figure 21). In common with the lambdoid suture phenotype, the majority of neonate MR26F8 cranial sections stained with von Kossa also showed a wider,

more open coronal suture (Figure 21D) than controls (Figure 21C). Again, the wider suture

area was composed of an unorganized combination of mature and immature bone matrix

(Figure 21F). However, by P12, all MR26F8s exhibited premature fusion of the coronal

V Sectioning and staining was performed by the Yale Orthopædic Histology and Histomorphometry Laboratory. 80

Figure 20. MR26F8 mutants have delayed ossification, followed by over- ossification, of the lambdoid suture.

81

Figure 20. MR26F8 mutants have delayed ossification, followed by over- ossification, of the lambdoid suture. Low magnification images in (A, B) show the approximate position of the lambdoid suture (Lb) from P12 sagittal sections. Sagittal sections from skulls of P0 (C-F) and P12 (G-J) control (C, E, G, I) and MR26F8 (D, F, H, J) pups. Sections (A-D, G-H) were stained with von Kossa, such that mineralized bone appears dark red/black. Sections (E-F, I- J) were stained with Goldner’s Trichome stain such that mature bone matrix appears green whereas immature bone matrix stains red. In the P0 sections, red arrowheads denote the edges of the mature bone and the distance between these arrowheads shows the suture width in (C-E). In (F), only one side of the widened suture is shown and the red rectangle outlines a region of immature bone matrix within this widened suture that is displayed in greater detail in the inset. The bar graph (K) shows the width of the lambdoid sutures at P0 and P12. Comparison of the width between the controls and MR26F8s is significantly different at both P0 (p<0.0001) and P12 (p<0.0003). Bars show average suture width (μM); error bars denote standard error within the group. Scale Bars: A-B: 1mm; C-J: 100µM. Aged-matched controls and MR26F8s are at the same scale.

82

Figure 21. Delayed ossification, followed by craniosynostosis, of the MR26F8 coronal suture.

83

Figure 21. Delayed ossification, followed by craniosynostosis, of the MR26F8 coronal suture. Low magnification images in (A, B) show the approximate position of the coronal suture (Cr) from P12 sagittal sections. Sagittal sections from P0 (C-F), and P12 (G-J) skulls for control (A, C, E, G, I) and MR26F8 (B, D, F, H, J) pups. (A-D, G-H) von Kossa staining with mineralized bone appearing dark red/black. (E-F, I-J) Goldner’s Trichome staining highlighting mature (green) and immature (red) bone matrix. The edges of the sutures (red arrowheads) and immature new bone matrix (red box) are shown, with the red arrow showing greater detail in the inset. Scale Bars: A-B: 1mm; C-J: 100µM. Aged-matched controls and MR26F8 are at the same scale.

84 suture (Figure 21H, J) compared to the control sutures, which were not fused (Figure 21G, I).

Therefore, the coronal suture in the MR26F8s was also characterized by delayed ossification followed by overgrowth and subsequent craniosynostosis. In conclusion, sections confirmed craniosynostosis of the coronal suture and narrowing of the lambdoid suture in the MR26F8s.

Abnormal Cartilage Replaces Intramembranous Bone at Higher Fgf8 Levels

To investigate the impact of Fgf8b dosage on craniofacial and skeletal morphogenesis, I next examined development in mice with a large increase in Fgf8b expression by utilizing

CAGF8 mice in concert with the Msx2-Cre transgene. As with the MR26F8 mice, the Msx2-

Cre;CAGFgf8b mice, hereafter termed MCAGF8, were born at normal size and Mendelian ratios. However, unlike the MR26F8s, the MCAGF8s did not survive beyond the first day, likely because of feeding deficits due to altered facial shape and clefting (Figure 22A-F).

Thus, my analysis was limited to P0 and embryological time points. MCAGF8 neonates exhibited several striking craniofacial phenotypes: a shortened snout, domed skull, cleft secondary palate and loss of the eyes (Figure 22D, E). Other morphological defects included limb deformities, particularly post-axial polydactyly, and more complex fusion phenotypes accompanied by a more rounded body shape (Figure 23).

To ascertain the underlying issues responsible for the craniofacial dysmorphology, I next compared E18.5 skeletons of MCAGF8 and control embryos using standard bone and cartilage staining. These studies revealed that MCAGF8s had a surprising skeletal phenotype. Specifically, while the calvaria of controls were composed of bone (Figure 22C), the bone of the MCAGF8 calvaria was mostly replaced with a matrix that stained with neither alcian blue nor alizarin red, but was still durable enough to persist through a staining protocol that degrades soft tissue (Figure 22F). In comparison, the endochondral bones and cartilage

85

Figure 22. MCAGF8 mice exhibit cranial defects including loss of ossification.

86

Figure 22. MCAGF8 mice exhibit cranial defects including loss of ossification. (A-B, D-E): Gross images of control (A-B) and MCAGF8 (D-E) E18.5 embryos showing lateral views of whole head (A, D) or ventral views of secondary palate after removal of lower jaw (B, E). Note, bluer staining of palate in E versus B results from an examination of skin permeability using toluidine blue staining. (C, F-K): Bone and cartilage staining of the control (C, G-H) and MCAGF8 (F, J-K) skulls, as well as the entire MCAGF8 body (I). (C, F, I): Lateral view; (G, J): dorsal view cranial vault; (H,K): ventral view of cranial base after removal of the lower jaw. Yellow dashed line (J) notes the boundary between bone and non-stained matrix formation. Red stars denote MCAGF8 ossification gains or losses. Abbreviations: BS, basisphenoid; Fr, frontal; IP, interparietal; Md, mandible; Ns, nasal; PPMX, palatal process of the maxilla; PPPL, palatal process of the palatine; Pr, parietal; SO, supraoccipital. Scale Bars: 1mm.

87

Figure 23. MCAGF8 limb Phenotypes. E18.5 gross (A, E, I) and skeletal (B-D, F-H, J-K) limb phenotypes. (A-D): controls, (E-K) MCAGF8s. The latter exhibit a wide range of phenotypes from polydactyly in the forelimb (G, compared to control, C) and hindlimb (H, compared to control, D) to fused forelimbs and hindlimbs (less severe, J; more severe, K). Red arrowheads denote various gross limb phenotypes. Red arrows denote postaxial polydactyly with 1 indicating the anterior most digit and 5 indicating the posterior most digit. Abbreviations: Fb, fibula; Fe, femur; Hu, humerus; Ra, radius; Ti, tibia; Ul, ulna. Scale Bars= 1mm.

throughout the rest of the body stained normally (Figure 22I). Dorsal views of the skull revealed a ring of bone surrounding the midline, with non-stained matrix radiating ventrally

88 from there (Figure 22J). Any sutures, if present, were difficult to identify due to the

prevalence of the non-stained matrix. The skeletal defect was not uniformly distributed throughout the calvaria, as there was a greater amount of ossification in the region where the parietal bones would normally form (Figure 22F, J). One possibility for the presence of bone in this region is that there was reduced activity of Msx2-Cre in this location as shown by fate mapping (Figure 16C). While the primary affected area was the cranial vault, there were also defects within other components of the craniofacial skeleton (Figure 22H, K). Several

skeletal elements were smaller, including the palatal process of the maxilla and palatal

process of the palatine (Figure 22K), consistent with the cleft secondary palate (Figure 22E).

There was also an overall shortening of the snout and its associated skeletal elements in the

MCAGF8s (Figure 22F, J, K) as well as a reduced supraoccipital bone (Figure 22J, K).

Additionally, like the MR26F8s, the MCAGF8s have ectopic bone growth within the eye

sockets (Figure 22J, K). Overall, the MCAGF8s exhibit several abnormalities, including

craniofacial patterning defects, loss of the eye, and limb defects. Most strikingly, however,

was the replacement of the cranial vault bones with an abnormal non-stained matrix.

To determine the origins and composition of the non-stained matrix, I examined the

MCAGF8s at the critical stages in embryological development when bone and cartilage are being formed. At early timepoints I utilized a Bouin’s fixative with alcian blue staining protocol for cartilage, as bone has yet to fully form and mineralize. At later timepoints, I switched to a standard protocol that uses alcian blue and alizarin red to stain cartilage and bone, respectively. Differences between the controls (Figure 24A, C, E, G, I) and MCAGF8s

(Figure 24B, D, F, H, J) were first apparent at E14.5. In the E14.5 wildtype head, cartilage has begun to form in the jaw (Meckel’s cartilage) and snout, as well as a number of posterior

89

Figure 24. Cartilage replaces bone throughout the MCAGF8 cranial vault.

90

Figure 24. Cartilage replaces bone throughout the MCAGF8 cranial vault Lateral (A-H) or dorsal (I, J) views of control (A, C, E, G, I) and MCAGF8 (B, D, F, H, J) skulls. (A-D): Skulls stained for cartilage only (alcian blue) using the Bouin’s Cartilage staining protocol at E14.5 (A-B) and E16.5 (C-D). (E-F): Bone and cartilage staining with alizarin red (bone) and alcian blue (cartilage) at E16.5. Note the difference in alcian blue staining at the same time period between the two methods (D, F). (G-J): Bone and cartilage staining with alizarin red, alcian blue, and toluidine blue (cartilage) at E18.5. The red rectangle (J) outlines a region where cartilage and bone overlap. J’ shows that same region magnified. Abbreviations: Fr, frontal, IP, interparietal; MC, Meckel’s cartilage; Md, mandible; Ns, nasal bone; Pr, parietal; SO, supraoccipital bone. Scale Bars: 1 mm.

91 elements that will develop into the cranial base and ear (Figure 24A). In the MCAGF8s, cartilage is also found in these locations but, in addition, ectopic cartilage forms throughout the cranial vault. At this stage, several of the craniofacial defects observed at E18.5 are already present, including the shorter snout and loss of the eye (Figure 24B, compare to

Figure 22F). The defects in the MCAGF8s seen using Bouin’s stain persisted at E16.5 with

extensive blue staining matrix seen throughout the cranial vault (Figure 24D). Surprisingly,

however, this matrix did not stain when alcian blue and alizarin red were used to stain at

E16.5 (Figure 24F). Instead, the non-stained matrix prevailed throughout the cranial vault,

similar to the E18.5 results (Figure 22F). Given that the staining of the matrix associated

with the skull vault was dependent upon the histological method used, I postulated that this

tissue was an abnormal type of cartilage that was differentially detected depending upon the

pH used for the staining protocol, as the bone and cartilage solution had a pH of 4.7 whereas

the cartilage only solution had a pH of 2.3. Therefore, I included toluidine blue, another

cartilage stain usually employed for sectioned material, to the alizarin red + alcian blue stain

for whole mount analysis of the E18.5 skeletons (Figure 24G-J). Toluidine blue stained the

regions occupied by the non-stained matrix. Alizarin red stained bone similarly in the

presence of toluidine blue in both control (Figure 24G, I) and MCAGF8 embryos, with the

latter displaying limited ossification near the sagittal suture (Figure 24H, J compare to Figure

22F, J). In the transition zone, bone and cartilage overlapped (Figure 24J’), indicating that

the threshold for forming bone or cartilage is not mutually exclusive. Taken together, normal

ossification of the cranial vault appears to be replaced with ectopic abnormal cartilage

formation in the MCAGF8 embryos, starting at the earliest stages of skull development.

92 To examine bone and cartilage formation in more detail in the MCAGF8 cranial vault,

I next identified regions of mineralization using sectioned material from E18.5 embryosVI treated with von Kossa stain (Figure 25A-C). Sections within the cranial vault revealed that while control embryos had abundant mineralization (bone formation) throughout the cranial vault (Figure 25A), MCAGF8 embryos had either no mineralization (Figure 25B) or minimal mineralization (Figure 25C), depending on the cranial region and consistent with the whole mount staining (Figure 22, Figure 24). When minimal mineralization was present in the

MCAGF8s, it was always closest to the brain, with the more dorsal regions forming an overlaying cartilage (Figure 25C). In summary, these results show that the increased levels of Fgf8 expression impaired cranial ossification in the MCAGF8 mice.

Given the absence of calvarial ossification coupled with indications of cartilage

(Figure 24) in the MCAGF8 cranial vault, I also assessed cartilage formation and composition by histochemistry using both toluidine blue (Figure 25D-E) and a dual alcian blue + PAS (Periodic Acid-Schiff) stain (Figure 25F-H). In the controls at E18.5, there was no cartilage in the layer underlying the ectoderm (Figure 25D). However, in the MCAGF8s, there was a thick layer of tissue with a hallmark hyaline cartilage appearance—nuclei embedded in a matrix with lacunae—throughout the cranial vault (Figure 25E, also visible in

Figure 25B-C, H). Additionally, the MCAGF8 tissue stained as a cartilage with toluidine

blue, typified by blue nuclei within a reddish purple matrix of glycosaminoglycans (Figure

25E). However, given that the cartilage did not stain with alcian blue in whole mount

(Figure 22, Figure 24), I still suspected that the cartilage must be abnormal in some way.

VI Sectioning and staining was performed by the Yale Orthopædic Histology and Histomorphometry Laboratory. 93

Figure 25. Histological analysis indicates abnormal cartilage replaces intramembranous bone in MCAGF8 cranial vault.

94

Figure 25. Histological analysis indicates abnormal cartilage replaces intramembranous bone in MCAGF8 cranial vault. (A-H) Frontal sections through the parietal bones of the controls and equivalent location in the MCAGF8 E18.5 embryos; regions shown are midway between the dorsal and ventral most aspects of the parietal. (A-C) von Kossa stain of the control (A) and MCAGF8 (B, C) sections. Mineralized bone is stained in dark red/black. (D-E) Toluidine blue staining of control (D) and MCAGF8 (E) sections. The nuclei stain blue and the glycosaminoglycans that make up the cartilage matrix stain reddish purple. (F-H) Alcian blue + PAS stain of the control cranial vault (F), control cartilage from the cranial base (G), and MCAGF8 cranial vault (H). Acidic mucins stain blue, neutral mucins stain magenta. Abbreviations: Bn, bone; Br, brain; Ct, cartilage; SE, surface epithelium. Scale bars: 20µM. Control and MCAGF8 sections are at the same scale.

95 Thus, I assessed the cartilage composition with dual alcian blue + PAS stain. The addition of

the PAS stain allows for a better understanding of the composition of the cartilage matrix as

it stains neutral mucins (i.e. polysaccharides and mucosubstances) while alcian blue stains

the more acidic mucins (i.e. glycosaminoglycans). As the wildtype showed no cartilage

formation at this stage (Figure 25F), I compared the MCAGF8 cartilage (Figure 25H) to

wildtype cartilage from the primordium of the presphenoid bone (Figure 25G). Both cartilages had a similar morphology, but compared to the wildtype cartilage (Figure 25G),

the MCAGF8 cartilage had less alcian blue and more PAS stain (Figure 25H), which is

indicative of fewer acidic mucins. Thus, cranial osteogenesis is replaced by chondrogenesis

in the MCAGF8 embryos. However, the resulting cartilage is abnormal with an altered ratio

of neutral to acidic mucins.

While the stains chosen were targeted for examination of bone and cartilage

formation, it was also apparent that the MCAGF8 embryos had defects not only in the

underlying skeleton, but also in the ectodermal derivatives. While the wildtype nasal

epithelium was composed of ciliated pseudostratified columnar epithelium (Figure 26A), in

the MCAGF8, the nasal epithelial cells were less elongated and ciliated (Figure 26B). In

contrast, the MCAGF8 surface epithelium (Figure 25B, C, E, H, Figure 26B) was abnormally

thick compared to the controls (Figure 25A, D, F, Figure 26A). Thus, MCAGF8s had defects

not only in the mesenchymal derivatives (bones), but also in ectodermally derived structures

that express the Fgf8 construct.

96

Figure 26. MCAGF8 mice also exhibit epithelial defects. (A, B) Frontal sections through the snout of E18.5 control (A) and MCAGF8 (B) embryos stained using Goldner’s Trichome. Green stain indicates mature bone matrix which is present in the control (A), but not in the MCAGF8 (B). Abbreviations: Bn, bone; Ct, cartilage; NC, nasal cavity; NE, nasal epithelium; NS, nasal septum; SE, surface epithelium. Scale bars: 40µM.

MCAGF8 Mutants Have Impaired Differentiation and Display Dysregulated WNT

Signaling

Given the striking transformation of bone into cartilage, I next explored the mechanistic basis of this phenomenon by comparing gene expression in the developing cranial vault between MCAGF8s and controls using RNA-seq (Figure 27A). I dissected cranial mesenchymal tissue between the skin and brain from the controls and MCAGF8s at

E14.5, when cranial bone development is initiating. In addition, based on the histological staining analysis, I predicted that the MCAGF8 cranial vault might contain gene expression signatures associated with hyaline cartilage. Therefore, I also sampled cartilage tissue fated to become bone by endochondral ossification from the control cranial base at the same age.

97

Figure 27. MCAGF8 cranial vault differentiation shifts from osteogenic to chondrogenic.

98

Figure 27. MCAGF8 cranial vault differentiation shifts from osteogenic to chondrogenic. (A) Schematic of RNAseq experimental design. E14.5 tissue was removed from control (left) and MCAGF8 (center) cranial regions outlined by the red boxes, after skin removal and avoiding underlying brain tissue. Control cranial base cartilage was removed from areas shown by red dots (right). (B) Scatterplot depicting average RPKM values. Colored points represent transcripts that are statistically significant between the groups: MCAGF8 cranial vault (Y-axis), and either control cranial vault (X axis, left) or control cranial base (X axis, right). Yellow represents genes upregulated in the MCAGF8 compared to the control (light yellow = p<0.05, dark yellow= p<0.05 and q<0.1). Blue represents genes downregulated genes in MCAGF8 compared to the control (light blue = p<0.05, dark blue= p<0.05 and q<0.1). (C) Histograms plotting normalized RPKM values of genes associated with collagen (left), cartilage/bone differentiation (middle) and ECM (right) significantly dysregulated in MCAGF8 cranial vaults (orange) as compared to control vaults (black). Error bars represent standard error. (D-E) Histograms plotting expression fold change in MCAGF8 cranial vault samples as compared to controls, where control values are normalized to 1 (black bars) whereas MCAGF8 samples are represented by green or red, respectively, for genes involved in differentiation (D) or Wnt signaling (E). Error bars represent standard error. (F-I) RNA in situ hybridization in the cranial vault for expression (purple) of Sox9 at E12.5 (F, G) or Osterix/Sp7 at E14.5 (H, I) for control (F, H) and MCAGF8 (G, I) embryos. (J-L) Alkaline phosphatase activity (blue) in the cranial vault of E16.5 control (J) and MCAGF8 (K, L) embryos. The cranial vault (red arrows) and surface epithelium (black arrows) are shown. Abbreviations: Br= brain. Scale Bars= 60µM.

99 This allowed comparisons between the MCAGF8 cranial vault and either the wildtype cranial

vault or wildtype cranial base using 9 embryos/group, pooled into 3 groups, with 3 embryos

per replicate, as outlined in Figure 27A. The full results from these comparisons are

presented in the Supplemental Tables of “Increased FGF8 Signaling Shifts Cell Fate from

Osteogenic to Chondrogenic in the Developing Skull” and are summarized in Figure 27B and

Table 4. Principal component analysis (PCA) was utilized to determine the variation

between and within each of the three groups; the wildtype cranial vault had the least variation

between samples, followed by the MCAGF8 cranial vault, and finally the wildtype cranial base (Figure 28, Figure 29) VII. This difference reflects the difficulty of the sample preparations, with wildtype cranial vaults being the easiest to dissect out cleanly and the cranial bases being the hardest.

Subsequently, to identify important categories of differentially expressed genes between the wildtype and MCAGF8 cranial vault samples, I used DAVID [283] functional annotation clustering (all significant genes p<0.05) and functional annotation charting

(p<0.05 and fold change ≥1.5 or ≤ -1.5) to analyze these datasets. Comparison of the

MCAGF8 cranial vault to the control cranial vault yielded differentially-expressed genes

annotated to categories including embryonic morphogenesis, WNT signaling, bone

development, ossification, osteoblast differentiation, regulators of bone mineralization, and

categories likely reflective of the shift to a cartilage backbone- such as extracellular matrix,

glycoproteins, glycosylation, disulfide bonds, and cell adhesion. Note that in the MCAGF8 model, Fgf8b has been activated in the ectoderm, which was removed prior to transcriptional

VII PCA of two way comparisons performed by Kenneth L. Jones; PCA of all three treatment groups performed by Hong Li. 100

Gene comparisons from RNAseq. 4. - Table

101

. (continued) arisons from RNAseq

Table 4. Gene comp .

102

Table 4. Gene comparisons from RNAseq. Genes listed in text, sorted by order of appearance. Left: wildtype cranial vault vs MCAGF8 cranial vault; Right: wildtype cranial base vs MCAGF8. Blue highlighted cells are genes downregulated <-1.5; yellow highlighted cells are genes upregulated >1.5. Significant P-values (p<0.05) highlighted in green. Full results will be deposited to the GEO (Gene Expression Ombibus) repository and are found in “Increased Fgf8 Signaling Shifts Cell Fate from Osteogenic to Chondrogenic in the Developing Skull”, currently under revision at Disease Models & Mechanims.

103 Figure 28. Two-way comparisons of the RNAseq treatment groups. PCA plots. Left: Comparison of the 3 wildtype cranial vault samples (encircled in blue) with the 3 MCAGF8 cranial vault samples (encircled in red). Right: Comparison of the 3 wildtype cranial base samples (encircled in black) with the 3 MCAGF8 cranial vault samples (encircled in red).

104 Figure 29. PCA of all three groups of RNAseq samples. PCA plot comparing the variability seen between the 3 wildtype cranial vault samples (encircled in blue), 3 MCAGF8 cranial vault samples (encircled in red), and 3 wildtype cranial base samples (encircled in black).

profiling of the skull. In this regard, Fgf8 was not found to be significantly different within the developing cranial vault of the MCAGF8 versus controls (Table 4). In contrast, both

Dusp6 and Spry4, which are reporters - as well as feedback regulators - of FGF signaling, were upregulated in the MCAGF8s (Table 4, fold change= 1.74, p=0.0049; 1.47, p=0.02,

105 respectively). Additionally, HtrA1, which directly cleaves FGF8 in the extracellular region

(resulting in decreased activation of FGF signaling) is upregulated in the MCAGF8s (~1.7

fold) and is likely indicative of attempted regulation of FGF8 signaling. These observations

are consistent with FGF8 signaling from the ectoderm affecting gene expression in the

developing underlying mesenchyme. To ask whether gene expression was consistent with a

switch from bone to cartilage, I next sampled expression of transcripts for genes that are

major structural components of bone and cartilage in the calvaria, since these should form

key aspects of the transcriptome differences between the wildtype and MCAGF8 samples.

Here I applied p<0.05 and fold change ≥1.35 or ≤ -1.35, first examining genes with RPKM

values > 100. These analyses showed major increases in the expression of Col2a1 and

Col9a3 (Figure 27C, left). These collagens are particularly notable as type II collagen

(Col2a1) is the major collagen synthesized by the chondrocytes and type IX collagen

(Col9a3) is one of the other major components of hyaline cartilage [322]. Differences were also seen in collagens that were expressed at much lower levels, notably a down-regulation of Col24a1 (Figure 27C, Table 4), which is involved in osteoblast differentiation and is a

marker for embryonic bone formation [323]. Therefore, I also examined an additional set of

highly-expressed genes involved in ossification and determined that many of these were

significantly downregulated in the MCAGF8s (Figure 27C, middle). Of particular note, Ibsp, a major structural protein of the bone matrix that constitutes ~12% of the non-collagenous proteins in human bone, was downregulated ~50 fold in the MCAGF8s (Figure 27C, Table

4). Further, Spp1 transcripts, encoding secreted phosphoprotein 1/osteopontin, a secreted regulator of osteoclast function, were reduced by >30 fold in the MCAGF8s. Other genes involved in bone formation, mineralization, and maintenance downregulated to a lesser

106 extent in the MCAGF8s included Sparc, Ifitm5, Fkbp11 (which interacts with Ifitm5),

Phospho1, Itm2a, Cthrc1, and Clec11a/osteolectin (Figure 27C, Table 4). Together, these findings indicate that the MCAGF8s have shifted from a bone to a cartilage program of gene expression.

Given that the MCAGF8 cartilage has an abnormal cartilage backbone, with more neutral than acidic mucins (Figure 25G, H), I next examined the expression of genes responsible for collagen processing, as well as extracellular matrix (ECM) formation and modification to determine if a subset of these were also dysregulated. Indeed, highly- expressed genes associated with the ECM were dysregulated in the MCAGF8s (Figure 27C, right; Table 4). Upregulated genes included Flna, a cytoskeletal protein involved in collagen remodeling and the ECM structural component Matn4. In addition, expression of Sulf1, encoding an enzyme that modifies heparan sulfate proteoglycans was upregulated, as was

Htra1- previously noted above in reference to Fgfs- encoding a protease that also antagonizes

Tgfb receptor signaling and inhibits bone development. Down-regulated genes included Ost4 and Ostc, which encode components of a membrane oligosaccharyltransferase, and Pcolce, which is involved in collagen processing. Pcolce2 and Loxl4, two additional molecules involved in collagen processing, were also down-regulated in the MCAGF8s although their expression levels are also considerably lower in wildtype tissue (Table 4). The aberrant expression of genes involved in collagen modification, as well as the changes in genes encoding enzymes that alter glycosylation and sulfation of ECM proteins, provides a possible explanation for the altered staining properties of the cartilage found in the MCAGF8 cranial vault.

107 Next, I examined if there were important differences between the collagen and ECM of the MCAGF8 cranial cartilage versus those found in the control cartilage of the cranial base. Comparing gene expression changes between the MCAGF8 cranial vault and control

cranial base samples revealed that more genes were dysregulated, and dysregulated at higher

levels, than between the control and MCAGF8 cranial vault samples (Figure 27B). The most

striking observation from the comparison between MCAGF8 skull vault and the normal

cranial base was that Col10a1 (type X collagen), which is critical for endochondral bone

formation [324], was highly downregulated in the MCAGF8 vault (fold change >100) (Table

4). Several additional genes involved in endochondral ossification were also down-regulated

to varying degrees in the MCAGF8 skull vault compared to the control cranial base including

Mmp13 (>10 fold), 3110079O15Rik /Snorc (>10 fold), Ihh (~10 fold), Spp1 (>5 fold), and

Ibsp (>5 fold). Fewer genes showed significant upregulation in the MCAGF8 skull samples

but these included Flna, Htra1 and Matn4 (1.5-5 fold), which also distinguished this tissue

from the wildtype skull (Table 4). In conclusion, gene expression changes indicate that the

MCAGF8 cranial vault cartilage is distinct from both the normally ossified cranial vault and

normal cartilage, such as found in the cranial base. Thus, although differentiation in the

cranial vault shifts from osteogenic to chondrogenic in the presence of excess FGF8

signaling, it is not undergoing normal endochondral ossification and displays distinct

histological staining properties from other cartilage in the body.

I next examined the balance between regulators of chondrogenesis and osteogenesis

in the control and MCAGF8 skulls to determine how FGF8 was exerting its mechanistic

effects. Since high fold change difference in regulators can affect differentiation even

without high expression levels, for this analysis I studied all differentially expressed genes

108 with a fold change >1.5 or <-1.5 and a significance of p<0.05. Several regulators of

cartilage/bone differentiation were dysregulated in the MCAGF8 vault (Figure 27D). First,

there was a shift towards expression of transcription factors that are linked with a

chondrocyte cell fate and inhibit ossification. In particular, Sox9, which stimulates

chondrocyte differentiation [325–327], Twist2, an inhibitor of osteoblast differentiation

[328–330], and Irx1 and Irx5 - markers of immature chondrocytes - were all upregulated

(~2.4, 3.1, 4.3, and 3.0 times, respectively; Figure 27D and Table 4). Conversely, mRNAs for the Sp7/Osterix, Mef2c, and Satb2, transcriptional regulators that stimulate bone differentiation, were down-regulated. Similar changes in these sets of transcription factors were observed when the MCAGF8 skull was compared to the cranial base, reinforcing that these represent different types of cartilage. I also examined the expression of Sox9 and

Sp7/Osterix using in situ hybridization in the wildtype and MCAGF8 skull at E12.5 and

E14.5 respectively (Figure 27F-I). These results demonstrated that the upregulation of Sox9 occurred in the MCAGF8s before overt ossification normally takes place and this was

mirrored by a later decrease in Sp7/Osterix expression. Together these data indicate that

changes early in the differentiation process may be responsible for shifting the cells from an

osteogenic to a chondrogenic cell fate.

To determine if dysregulation of osteoblast differentiation genes was leading to

impaired differentiation, I next examined alkaline phosphatase activity, a marker of bone

differentiation, at E16.5. These studies revealed that alkaline phosphatase activity was

decreased in the mutants (Figure 27K, L) compared to the controls (Figure 27J). Also,

similar to the mineralization phenotype (Figure 25A), the MCAGF8s had very little alkaline

phosphatase activity in most regions (Figure 27L), but in regions where it was present, it was

109 closest to the brain (Figure 27K). In summary, in the presence of excess Fgf8 during early

development, osteogenic differentiation was impaired and switched to a more cartilaginous differentiation pathway.

Given the significance of the Wnt pathway in the DAVID functional annotation clustering, I also examined how genes involved in cell:cell signaling were altered between the wildtype and MCAGF8 skull samples. The potential involvement of WNT signaling was particularly interesting as loss of β-catenin protein in the skull can also switch development of the calvaria from bone to cartilage [249, 331]. In this respect, multiple inhibitors and negative regulators of WNT signaling were upregulated in the MCAGF8, including Apcdd1,

Axin2, Kremen2, Nkd1, Nkd2, and Prickle1 (Figure 27E and Table 4). At the same time, several genes that positively regulate WNT signaling were upregulated in the MCAGF8s,

including the WNT ligands 4, 5a, 6, 7b. Additionally, members of the canonical WNT

signaling pathway, including frizzled receptors (Fzd9, Fzd10) and transcription factor Lef1

(Figure 27E) were also upregulated. In contrast, only one WNT ligand, Wnt2, and one WNT

pathway inhibitor, Dkk1, were significantly downregulated. In addition to changes in genes

associated with WNT signaling, there were also clear effects of Fgf8 overexpression on other

signaling pathways, notably the Hedgehog, Insulin-like Growth Factor, and TGF-β/BMP

pathways that might all be predicted to alter bone growth and development (Figure 30).

Reducing Axin2 Gene Dosage Partially Rescues the Cranial Skeleton Phenotype

While the expression of many signaling genes are altered in the MCAGF8 model, the

RNAseq data alone do not provide a clear indication of how these alterations could impact

intramembranous ossification. However, given the many alterations in the Wnt signaling

pathway shown in the RNA-seq, as well as the connections between β-catenin and

110

Figure 30. Additional pathways altered in the MCAGF8 skull. Histograms plotting expression fold change in MCAGF8 cranial vaults samples (orange) as compared to control vaults (blue), where control values are normalized to 1, for significantly dysregulated genes involved in BMP/TGFβ (A) and HH (B) signaling. Error bars represent the standard error.

intramembranous ossification, I further probed the interaction between FGF and WNT signaling in vivo. Specifically, I bred an Axin2lacZ allele into the MCAGF8 background 111 (Figure 31). Axin2 expression is stimulated by the canonical WNT signaling pathway and

also acts as a feedback regulator of the pathway via degradation of β-catenin [243]. The

Axin2lacZ allele replaces the normal gene with NLS-lacZ transgene [332], so that normal

Axin2 expression decreases in heterozygotes, increasing Wnt signaling.

Initial gross morphological analysis of the MCAGF8;Axin2+/- indicated that several

MCAGF8 phenotypes still persisted, including the missing eye, the shortened snout, and the

cleft palate (Figure 31C and data not shown). However, subsequent skeletal staining

indicated clear difference in skull ossification between the two models. Thus, in contrast to

the MCAGF8 skulls, which had some ossification near the midline suture with non-stained matrix throughout most of the skull (Figure 31E, H), the MCAGF8;Axin2+/- mice showed an expansion of ossification midway down the skull, with non-stained matrix only on the lateral part of the skull, nearest the cranial base (Figure 31F, I). The expansion of bone development in MCAGF8;Axin2+/- mice also led to the reappearance of the lambdoid suture

(Figure 31I), a feature not apparent in the MCAGF8s due to the absence of bone (Figure

31H). However, increased bone development did not lead to rescue of all sutures, as the

MCAGF8;Axin2+/- mice displayed craniosynostosis of the coronal suture (Figure 31I).

Indeed, the suture pattern in the MCAGF8;Axin2+/- mice, with an open lambdoid suture and

craniosynostosis of the coronal suture, was similar to that of the MR26F8 mice (Figure 17).

Interestingly, the reduced Axin2 gene dosage also rescued the hypoplasia of the

supraoccipital bone seen in MCAGF8s, but consistent with the gross morphology, it did not

alter the skeletal defects associated with the eye, snout and secondary palate (Figure 31L).

Thus, targeting the WNT signaling pathway by regulating Axin2 gene dosage partially

112 Figure 31. Reduced Axin2 gene dosage improves cranial vault ossification.

113

Figure 31. Reduced Axin2 gene dosage improves cranial vault ossification. (A-C): Lateral view of control (A), MCAGF8 (B), and MCAGF8;Axin2lacz +/- heads. (D-L): Bone and cartilage staining of control (D, G, J), MCAGF8 (E, H, K), and MCAGF8;Axin2lacz +/- mutants (F, I, L). D-F show a lateral view, G-I show dorsal view of the cranial vault, and J-L show the cranial base. The boundary between bone and non-stained matrix (yellow dashed line), the presence or absence of the coronal suture (green arrow), and lambdoid suture (green arrowhead) are shown. Red stars denote regions of the cranial base that differ from the control; the green star in (L) shows the rescue of the supraoccipital bone. Abbreviations: BS, basisphenoid; Fr, frontal bone; IP, interparietal bone; Md, mandible Ns, nasal bone; PPPL, palatal process of the palatine; PPMX, palatal process of the maxilla; Pr, parietal bone; SO, supraoccipital. Scale Bars= 1mm.

114 rescued the shift from osteogenesis to chondrogenesis in the skull vault of MCAGF8s, but did not rescue the coronal craniosynostosis or craniofacial shape phenotypes.

The Axin2lacZ allele also serves as a means to follow WNT-signaling via transgene expression. Thus, I next examined the levels of WNT signaling during both early embryogenesis and at birth in the MCAGF8s. At E12.5 and E13.5, Axin2lacZ expression in

the MCAGF8s was increased in an ectopic band of tissue extending from the snout, but there

were no obvious differences in expression observed in the developing cranial vault (Figure

32A-J). In the control limbs, Axin2lacZ was expressed on the lateral edges of the forelimbs

and, by E13.5, in a pattern that outlined the digits (Figure 32K, M). While Axin2lacZ was also

expressed in the lateral edges of the MCAGF8 forelimbs, Axin2lacZ expression did not

demarcate the digits, as in the controls (Figure 32L, N). Therefore, the expression pattern of

Axin2lacZ during early development suggests WNT signaling dysregulation is correlates more

with facial dysmorphology than ossification. Thus, I next examined Axin2lacZ expression in

the neonate cranial vaults. In the control head, Axin2lacZ was mostly strongly expressed in the

vibrissae, eyes, and developing sutures, with more diffuse expression throughout the cranial

vault bones (Figure 33A-B). In contrast, Axin2lacZ was not clearly expressed in the sutures of

the MCAGF8 cranial vault, but rather was more strongly expressed throughout the cranial

vault, particularly in the lateral part of the bones, where cartilage rather than bone

predominates. As MCAGF8 mice lack eyes, there was no Axin2lacZ in the eyes, but there was

expression in the vibrissae, as in the controls. In summary, Axin2lacZ expression in the

cranial vault suggests that there was decreased WNT signaling in the MCAGF8 sutures, but

increased WNT signaling in the MCAGF8 cranial vault cartilage. Notably, the region in the cranial vault where ossification was rescued in the MCAGF8;Axin2+/- mice had less Axin2lacZ

115

Figure 32. Axin2lacZ expression during early embryogenesis.

116

Figure 32. Axin2lacZ expression during early embryogenesis. Axin2lacZ expression as visualized with β−galactosidase staining at E12.5 (A, C, D, G, H, K, L) and E13.5 (B, E, F, I, J, M, N). (A-B): Lateral views of the whole embryo. (C- F): Lateral view of the head. (G-J): Front of the face. (K-N): Forelimbs. As labeled, the 1st and 3rd columns are control;axin2lacZ/+, the 2nd and 4th columns are MCAGF8. Red stars indicate regions of ectopic expression. Red arrowheads indicate regions of expression in the limb. 1 denotes the 1st digit, 5 denotes the 5th digit.

117

Figure 33. Axin2lacZ expression at birth in the cranial vault. Lateral (A) and dorsal (B) views of Axin2lacZ expression in the cranial vault after skin removal, as visualized with β−galactosidase staining. Top images are controls, bottom images shows MCAGF8s. Scale Bars=2mm.

expression than the cartilaginous regions. Thus, the regions with normalized levels of WNT

signaling in the MCAGF8 mice correlates with the regions where ossification is restored; this suggests that Fgf8 induced increases in WNT signaling shifts cell fate from osteogenic to chondrogenic.

Summary

Cranial ossification responds to Fgf8 overexpression in a dose dependent manner with moderate levels leading to craniosynostosis and higher levels shifting cranial vault ossification to abnormal cartilage formation. Expression analysis demonstrated that abnormal skull chondrogenesis was accompanied by changes in genes required for WNT signaling. Moreover, the reduction in ossification could be partially rescued by manipulating 118 Axin2 gene dosage, indicating a role for altered WNT signaling in the pathology. Taken

together, these findings indicate that mesenchymal cells of the skull are not fated to form

bone but can be forced into a chondrogenic fate via manipulation of FGF8 signaling.

Discussion

Craniosynostosis is a relatively common human craniofacial defect with clear

involvement of FGF signaling pathway mutations, particularly the FGF receptors. Here I

have probed this pathway further by manipulating the dosage of an FGF ligand during

mammalian development. My characterization of how increased Fgf8 expression impacts skeletal formation, particularly within the cranial vault, yields several interesting and previously undocumented findings. First, I note the cranial vault has a dose-dependent response to FGF8 signaling, with moderate levels of Fgf8 overexpression leading to coronal craniosynostosis and high levels of Fgf8 overexpression leading to ectopic cartilage formation. Second, I report that a balance between FGF and WNT signaling is critical for driving the decision between an osteoblast or chondrocyte cell fate in vivo for the mesenchymal cells of the calvaria that would normally form bone via intramembranous ossification. Finally, I note that the effect of ectopic Fgf8 on craniofacial development is partially dependent on the timing and location of expression.

Moderate Overexpression of Fgf8 Leads to Craniosynostosis

Cranial suture patency is normally maintained via a delicate balance of proliferation and differentiation. Here I report that when Fgf8 is moderately overexpressed in the cranial ectoderm using Msx2-Cre;R26Fgf8b I observed coronal craniosynostosis and narrowing of the lambdoid suture. However, premature suture closure was not simply caused by accelerated ossification at the sutures. Instead, in these MR26F8s, the coronal and lambdoid sutures were

119 wider at birth than in the controls. However, by P12, the coronal suture had completely

fused and the lambdoid suture was narrower than in the controls. Therefore, in this model

there is a pattern of slower bone maturation accompanied by a failure to develop appropriate

suture architecture that subsequently results in overgrowth and craniosynostosis. Notably,

such delayed ossification followed by catch-up growth and subsequent obliteration of the sutures has also been observed in two mouse models with either loss or gain of function alterations in FGFR2 [309, 333]. Like these latter mouse models, the majority of such craniosynostosis models resulting from FGF signaling aberrations have been generated in mice using activating mutations in FGF receptors, particularly Fgfr2 and Fgfr3 [39, 163].

Such mutations may cause ligand independent dimerization and activation of the receptors with downstream intracellular signaling consequences dependent on the position and nature of the mutation [172–175]. Alternatively, some FGFR activating mutations, including ones that cause Apert Syndrome, do not lead to ligand independent receptor binding, but instead cause increased affinity for different FGF ligands [176]. This is striking as patients with

Apert Syndrome, as well as those with Crouzon and Pfeiffer syndrome, often exhibit craniofacial phenotypes similar to those found in my mouse model of moderate Fgf8 overexpression - a characteristic appearance that includes a shortened face and premature fusion of the coronal sutures. Additionally, like in my model, those with Apert syndrome frequently exhibit other symptoms such as cleft palate and/or limb defects, including syndactyly as well as occasional polydactyly [47, 133]. Therefore, the availability of alleles

expressing different combinations or amounts of FGF ligands may help tease apart the

combinatorial manner by which specific ligand:receptor interactions can lead to the different

forms of craniosynostosis, with potential therapeutic applications.

120 Dose-dependent Response to Fgf8 Signaling in the Cranial Vault

The phenotypes observed when comparing mice with heterozygous or homozygous activating mutations in FGFR2 have demonstrated that the effect of increased FGF signaling on cranial vault ossification can be dose-dependent [309]. Here, using two novel Fgf8 constructs, I describe an FGF ligand having a dose-dependent effect on cranial ossification.

When Fgf8 is overexpressed at moderate levels (MR26F8), the mutants exhibit coronal craniosynostosis, as predicted by numerous studies linking an overexpression of FGF signaling with craniosynostosis phenotypes [104]. However, when Fgf8 is overexpressed at higher levels (MCAGF8), an unexpected phenotype emerges: cartilage formation replaces ossification extensively across the cranial vault. This is a novel phenotype associated with altered FGF8 signaling, although previous studies have shown ectopic expression of Fgf9 in the cranial vault also lead to cartilage formation, though in this case, prior to ossification

[334]. Additionally, certain Fgfr2 mutations can result in more limited ectopic cartilage formation. In particular, mice homozygous for an activating mutation in FGFR2 (W290R), mimicking human Crouzon Syndrome, exhibit thickened cartilage underlying the cranial bones [309]. Similarly, in a mouse model for Apert syndrome, a heterozygous activating mutation in FGFR2 (S252W) leads to coronal synostosis and ectopic cartilage at the midline sagittal suture [335]. Beyond the FGF signaling pathway, ectopic cartilage formation has also been observed due to aberrations in WNT signaling [111, 249, 331, 336]. Thus, removal of β-catenin from cranial bone progenitors results in near complete transformation of the skull bones to cartilage – a phenotype that is more akin to the MCAGF8 mutants than that seen with the above mentioned W290R and S252W Fgfr2 alleles.

121 One potential complication with the interpretation of the dose-dependent effects of

Fgf8 signaling is that highly increased levels of Fgf8 may not lead to greater FGF signaling than moderately increased Fgf8. In contrast, highly increased Fgf8 (CAGF8) may lead to less FGF signaling than moderately increased Fgf8 (R26F8) by inducing negative regulation of FGF signaling. I do not favor this hypothesis as there are several phenotypes in addition to the cartilage phenotype that are more severe in the MCAGF8 mice than in the MR26F8

mice. The MCAGF8 mice had more severe eye defects, craniofacial patterning defects, and

limb patterning defects than the MR26F8 mice and, unlike the MR26F8 mice, exhibited cleft

palate and did not survive past birth. Additionally, both Dusp6 and Spry4, reporters of FGF

signaling, were upregulated in the MCAGF8 mice, showing that there is upregulation of FGF

signaling in the MCAGF8 model. However, to verify that FGF signaling is increased in the

MCAGF8 mice more than in the MR26F8 mice, further experiments would need to be done,

such as measuring Dusp6 and Spry4 expression in the MR26F8 mice or looking at the

amount of FGFR phosphorylation between the two models.

How Does Fgf8 signaling Provoke the Shift from Bone to Cartilage Formation?

My studies using both in situ hybridization and RNA-seq analysis indicate that

upregulation of chondrogenic differentiation markers during early bone development shifts

cell fate from osteogenic to chondrogenic. As such, later osteogenic differentiation markers

and genes critical for osteoblast differentiation are downregulated whereas inhibitors of

osteoblast differentiation are upregulated. Impaired osteoblast differentiation in MCAGF8s

was confirmed via alkaline phosphatase activity. This shift in differentiation led to increased

expression of the collagens involved in cartilage formation, Col2a1 and Col9a3, decreased

expression of genes involved in ossification, and dysregulation of genes involved in

122 extracellular matrix formation. Together, these shifts reflect the formation of cartilage, rather than bone in the MCAGF8s.

Strikingly, the gene expression data indicate that dysregulation of WNT signaling in the Fgf8 induced cartilage is one potential mechanism for impaired differentiation – especially given the similarity between the skull vaults of the MCAGF8 embryos and mice with tissue-specific loss of β-catenin from cranial bone progenitors [111, 249, 331, 336].

Previous studies have shown that, in conjunction with Sox9 [337, 338], Wnt signaling promotes chondrocyte differentiation [338, 339]. As noted earlier, however, downregulation of WNT signaling also induces ectopic cartilage in the cranial vault. Interestingly, this suggests that both upregulation and downregulation of WNT signaling in the developing cranial vault shifts cell fate from osteogenic to chondrogenic. This could be due to the developmental stage at which WNT is overexpressed with WNT signaling regulating the chondrocyte life cycle by balancing positive and negative influences during embryogenesis, whereas in postnatal cells, it has been implicated as a positive regulator of differentiation

[340] In vitro, activation of WNT signaling in mesenchymal stem cells (MSCs) inhibits osteoblast differentiation. However, once the MSCs have committed to the osteoblast lineage, WNT signaling promotes their differentiation [341].

In this study, I found that both positive and negative regulators of the WNT signaling pathway were upregulated making it difficult to predict if the influence of increased FGF ligand expression was to activate, repress, or else leave the overall throughput of the canonical WNT pathway unaffected. To help distinguish between these possibilities, and determine if WNT dysregulation was contributing to the cranial ossification phenotype, I utilized mice containing Axin2lacZ, an allele that inactivates the endogenous copy of the gene.

123 Since Axin2 is an inhibitor of the WNT pathway, mutants with this allele should show

increased Wnt signaling compared with controls. I determined that MCAGF8, Axin2lacZ

mutants had a less severe phenotype than the MCAGF8 mutants with significant rescue of ossification of the cranial vault. This suggests that WNT signaling is downregulated in the skulls of MCAGF8 mutants, as decreased expression of an inhibitor (AXIN2) partially rescues the ossification phenotype. Indeed, Axin2lacZ expression was higher in the lateral cartilaginous regions of the skull than it was in the dorsal cranial vault where ossification was

rescued. In support of this hypothesis, previous studies in tissue culture using an osteoblast

cell line also demonstrated that FGF signaling was antagonistic to the role of the WNT

pathway in driving bone differentiation [342]. In addition to the ossification defects in the

calvaria, the Msx2-Cre mutants had patterning defects—a shortened snout, and in the

MCAGF8 mice, a domed skull and missing eye. Unlike the ossification defect, these

patterning defects are not rescued by a reduction in Axin2 gene dosage, and thus are not as

responsive to, or do not rely on, Wnt signaling as an intermediate.

Overall, my findings using the two new Fgf8 alleles suggest a model in which

moderate overexpression of Fgf8 in the MR26F8 model slows the differentiation of

osteogenic progenitors leading to wider sutures at birth, while also inhibiting normal function

of the sutures, ultimately causing craniosynostosis. However, as the dose of Fgf8 is

increased in the MCAGF8 model, the skeletogenic progenitor cells are shifted from an

osteogenic to a cartilaginous cell fate. Alternatively, the ectopic Fgf8 may block a set of

cells that would normally form the bony calvaria and instead simulate adjacent underlying

cells that can adopt a cartilage fate. Although I cannot currently distinguish between these

models, it is apparent that these decisions are at least partly as a result of altered Wnt

124 signaling. Thus, the Wnt signaling pathway inhibits cartilage formation in the skull via its

action in the mesenchyme, and this function is repressed by the significant overexpression of

Fgf8.

Previous studies suggest that osteoblast location within the skull may impact the

degree of FGF signaling activation. Osteoblasts from neural crest-derived bones, such as the

frontal bone, express FGF osteogenic ligands and their receptors at higher levels than

osteoblasts from paraxial mesoderm-derived bones, such as the parietal bones, suggesting the

frontal bone is more sensitive to FGF signaling than the parietal bones [127, 343]. My

results support this conclusion as the parietal bones, while not immune to Fgf8’s effects, are

less severely affected than the frontal bones, particularly in the MCAGF8 embryos. This

phenomenon may be due, in part, to Msx2-Cre’s expression pattern. In addition, the limited expression pattern of Msx2-Cre did not allow an assessment of how ectopic Fgf8 expression could impact the development of additional skeletal structures, such as the lower jaw. Thus,

I will further address the effect of Fgf8 overexpression in these locations in the following chapter.

During endochondral, but not intramembranous ossification, there is a cartilage intermediate on the way to producing bone. While cartilage forms in the MCAGF8 mice, this

cartilage template does not subsequently appear to undergo endochondral ossification, but

remains cartilaginous. Since these mice do not live into the post-natal period, it is possible

that the cartilage might form bone if a longer developmental window was available.

However, I do not favor this possibility due to the unusual properties and gene expression

profile of the ectopic cartilage. Although the cartilage has a normal hyaline morphology, the

extracellular matrix is abnormal as shown by its altered response to alcian blue staining

125 indicative of a switch from acidic to neutral mucins. One potential reason for this abnormal

extracellular matrix formation is that FGFs normally require heparan sulfate proteoglycans

for optimal interaction with the FGFRs [344]. Therefore, the abnormal balance between

acidic and basic mucins in the ECM may be a feedback control mechanism to alter the

ECM’s ability to interact with the unusually high levels of Fgf8 in my model. In addition,

my RNAseq data indicate that there are significant differences between the abnormal

MCAGF8 skull cartilage and the normal cartilage of the cranial base that is fated to undergo

endochondral ossification. In particular, Col10a1, which is critical for chondrocyte

differentiation during endochondral ossification, is significantly downregulated in the Fgf8

induced cranial vault cartilage compared to cartilage in the cranial base. Taken together,

these results indicate that Fgf8 induced cartilage is not equivalent to the cartilage formed

during endochondral ossification.

Outside the context of the craniofacial skeleton, the MCAGF8 and MR26F8 mice

have defects in limb patterning, similar to those previously observed with other models that

overexpress the Fgf ligands FGF4 and FGF8 in the limb ectoderm [345, 346]. These defects

include postaxial polydactyly and forelimb hindlimb fusions, but in general - although limb

outgrowth and patterning are affected - the formation of bones via endochondral ossification

still occurs relatively normally. In concordance with my results, Fgf8 has been shown to be required for outgrowth and patterning, but not ossification, in the limbs [89, 347], while generalized transgenic expression of Fgf2 also causes shortening and bending of the radius and ulna [314]. However, it is possible that the developing limb bones are outside the range of Fgf8 influence when the source of the increased Fgf8 expression is the limb ectoderm, as it is with Msx2-Cre.

126 Additionally, the supraoccipital, which forms via endochondral ossification, was severely hypoplastic in the MCAGF8 mice, but was rescued in the MCAGF8, Axin2+/- mice.

I speculate that the ossification defect in this endochondral bone may be due to the decrease in Mef2c expression, as previous studies have shown that a loss of Mef2c results in loss of ossification of the supraoccipital [348]. Nevertheless, ossification of the endochondrally forming supraoccipital is impaired and is Wnt responsive. Thus, from these data alone, it is difficult to say if intramembranous ossification is more sensitive to Fgf8 overexpression than endochondral ossification. As the molecular mechanisms responsible for the early transition from progenitor cell to chondrocyte or pre-osteoblast are not well understood, the role of

Fgf8 signaling in intramembranous and endochondral ossification will be further examined in the following chapter.

The availability of the new Fgf8 alleles will allow for further analysis of how different FGF expression levels as well as different FGF ligands can impact cell fate decisions during skeletogenesis. It has not escaped my notice that this method—using a non- bone specific Cre recombinase to study ossification—sets a precedent for studying the role of other FGF ligands or additional paracrine signaling pathways in cranial ossification. While in this study, I examine the effects of paracrine signaling from the ectoderm, paracrine signaling from other nearby tissues, such as the dura matter, would presumably yield similar results. Given the important role the dura mater plays in dictating overlying bone formation

[349], this method could have particular implications for studying craniosynostosis. Finally, my results on the dose-dependent effects of Fgf8 on the cranial vault have implications for disease treatment. Some craniosynostosis patients require repeat surgeries due to the re- fusion of the sutures and knowledge of the molecular pathways downstream of FGF

127 signaling could lead to the rational design of treatments to prevent re-fusion. Unraveling the role of FGF signaling in cranial ossification and its downstream molecular consequences greatly expands our understanding of human craniofacial disorders and provides the possibility of novel treatments for those pathologies.

128 CHAPTER V

INCREASED FGF8 SIGNALING DRAMATICALLY IMPAIRS OSSIFICATION

OF INTRAMEMBRANOUS, BUT NOT ENDOCHONDRAL BONESVIII

Introduction

There are two types of ossification: intramembranous or endochondral. Bone forms

through intramembranous ossification in most of the skull, including the cranial vault, jaw, and most of the facial bones. Most of the rest of the skeleton, including the long bones, vertebrae, and basicranium, forms via endochondral ossification. The main distinguisher between intramembranous and endochondral ossification is the presence of an intermediate step of cartilage formation. During endochondral ossification, condensed mesenchyme cells first differentiate into chondrocytes that form cartilage tissue. This intermediate cartilaginous template is then replaced by bone, formed through osteogenesis. In contrast, in intramembranous ossification, condensed mesenchyme cells directly differentiate into osteoblasts that form bone without any cartilaginous precursor [104, 105].

While the basic processes of intramembranous and endochondral ossification are well established, the molecular signals distinguishing the two processes have not been well differentiated. In intramembranous ossification, differentiating cells (preosteoblasts) are marked by a sequence of transcriptions factors as they develop into osteoblasts – first Sox9, followed by Runx2, and finally Osterix [105]. However, Sox9+ cells have the potential to differentiate into osteoblasts or chondrocytes [107]. Additionally, co-expression of Sox9 and

VIII An abridged version of Chapter 4 combined with the first section of Chapter 5 (OC-Cre) is under revision in Disease Models and Mechanisms as “Increased FGF8 Signaling Shifts Cell Fate from Osteogenic to Chondrogenic in the Developing Skull.” 129 Runx2 typically causes cartilage formation, not bone, and the deletion of Osx causes Runx2+

cells to undergo chondrocyte rather than osteoblast differentiation [108, 350, 351]. Finally,

histological and most molecular markers of cartilage and bone are conserved during skeletal

differentiation in intramembranous and endochondrally forming bone [350].

Thus, the regulation of intramembranous vs endochondral cell fate is not well defined

and likely includes contributes from multiple signaling pathways. The canonical WNT

signaling pathway has been shown to be critical for osteoblast differentiation [109, 110, 352].

Deletion of β−catenin in the WNT signaling pathway can shift cell fate from osteogenic to chondrogenic in both Runx2+ and Runx2+Osx+ cells [109–111] . In turn, WNT signaling

has been shown to function downstream of the FGF and HH signaling pathways in

development of the osteoblast lineage [342, 352, 353]. However, the distinct role of each

signaling pathway in intramembranous vs. endochondral ossification is not well understood.

Fgf and Fgfr gene expression is critical for the formation of both endochondral and

intramembranous bones. Though the expression pattern of certain FGF signaling members

differs between the two types of developing bone, it has not yet been delineated how or if

these differences impact cell fate of the osteogenic mesenchyme [128]. In the previous

chapter, increased Fgf8 expression had a striking impact on the development of the

intramembranous bones of the cranial vault, shifting cell fate from osteogenic to

chondrogenic. Thus, I wondered if its impact was as severe on the development of the

endochondral bones. While the expression pattern of the Msx2-Cre transgene was suitable

for studying development of the calvaria, it was not efficient at directing Fgf8b to other

elements of the craniofacial skeleton that form by intramembranous ossification, such as the

mandible. At the same time, the expression of Msx2-Cre in the limb ectoderm did not appear

130 to affect ossification of endochondral bones in the limb (Figure 18, Figure 23), whereas the supraoccipital bone in the skull, which forms via endochondral ossification, was greatly reduced in the MCAGF8s (Figure 22 and Figure 31). However, it is possible that the developing limb bones are outside the range of Fgf8 influence when the source of the increased Fgf8 expression is the limb ectoderm. Therefore, to ascertain the sensitivity of endochondral and intramembranous forming bone to Fgf8 overexpression, I utilized three different mesenchymally expressed Cre-recombinases – Osteocalcin-Cre, Col2a1-Cre, and

Creface -- in conjunction with the Fgf8 overexpressing alleles, R26F8 and CAGF8. Each of these Cre-recombinases has a specific spatiotemporal expression pattern that bestows it with unique advantages and disadvantages, each of which will be addressed in turn. Though the

Analyzed together, this approach allows me to further examine the role of FGF8 signaling in intramembranous and endochondral bone formation.

Results

Osteocalcin-Cre

Given the striking impact increased Fgf8 expression has on the development of the intramembranous bones of the cranial vault (Chapter 4), I wondered if its impact was as severe on the development of the endochondral bones. Thus, to ascertain how increased

Fgf8b expression affects ossification more globally throughout the embryo, I first utilized

OC-Cre (osteocalcin-Cre), which is expressed in osteoblasts in both endochondral and

intramembranous forming bone [354]. While the creators of OC-Cre didn’t observe expression until E17.0 [354], in my hands, OC-Cre based recombination began in endochondral and intramembranous forming bones at E14.5 and marked the cranial vault, jaw, ribs and limbs by E16.5 (Figure 34). Thus, using OC-Cre, along with either the R26F8

131

Figure 34. Osteocalcin-Cre expression from E14.5-E16.5. Osteocalcin-Cre (OC-Cre) mediated recombination was visualized using β- galactosidase staining (blue) of Rosa26 LacZ Reporter embryos. (A-E) Whole embryo β- galactosidase staining was performed at E14.5 (A), E15.5 (B), and E16.5 (C-E). Skin was removed prior to staining in the E15.5 and E16.5 embryos. (C): Lateral view of the ribs and limbs. (D): Lateral view of the head. (E): Dorsal view of the cranial vault. Scale Bars= A, D-E: 1mm; B, C: 2mm.

or CAGF8 allele, should increase Fgf8 expression from E14.5 onwards in the endochondral and intramembranous bones.

Once again, the increase in Fgf8b expression resulted in developmental defects, with the resulting OR26F8 (OC-Cre;R26Fgf8b) and OCAGF8 (OC-Cre;CAGFgf8b) mice being

132 readily distinguishable from controls by gross morphology at E18.5, due to shortened

mandibles and limbs (Figure 35 and Figure 36). As expected given the expression pattern

and timing of OC-Cre, neither OR26F8 nor OCAGF8 embryos had the ectodermal-associated

defects seen in the MR26F8 and MCAGF8 mice — abnormal skin and loss of the eye. Both

OR26F8 and OCAGF8 mice had a number of similar skeletal defects, which were more

severe in the latter (Figure 37). The calvaria were mostly replaced by an unstained matrix, in

common with Msx2-Cre-based mice, although there were some notable differences in the most affected regions. Thus, whereas Msx2-Cre-based mice had some ossification

surrounding the midline sutures, this top most region of the cranial vault consisted solely of

non-stained matrix in the OC-Cre mutants (Figure 37E, F). In contrast, there was more

lateral ossification in both OC-Cre mutants (Figure 37B, C), than there was in the MCAGF8s

Figure 35. Gross Morphology of E18.5 OCAGF8 embryos. (A-F) Gross Morphology of control (A-C) and OCAGF8 (D-F) E18.5 embryos. (A, D) Lateral view of head. (B, E) Dorsal view of the head. (C, F) Lateral view of the body. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars= A-B, D-E: 1 mm; C, F: 2 mm.

133

Figure 36. OR26F8 mice have craniofacial shape and skeletal defects. (A-B): Gross morphology of P0 control (A) and OR26F8 (B) heads, lateral view. (C-D): Bone and cartilage staining of E16.5 control (C) and OR26F8 (D) heads, lateral view. Scale bars= 1mm.

134

Figure 37. Intramembranous forming bones are more severely impacted by increased Fgf8 expression than endochondral forming bones.

135

Figure 37. Intramembranous forming bones are more severely impacted by increased Fgf8 expression than endochondral forming bones. (A-R): Bone and cartilage staining of E18.5 control (A, D, G, J, M, P), OR26F8 (B, E, H, K, N, Q), and OCAGF8 (C, F, I, L, O, R) samples. (A-C): lateral view of head, (D- F): dorsal view of cranial vault (G-I): ventral view of cranial base after mandible removal, (J- L): dorsal view of mandibles, (M-O): forelimbs, (P-R): hindlimbs. Blue arrow shows parietal bones. Abbreviations: Fb, fibula; Fe, femur; Fr, frontal; Hu, humerus; IP, interparietal; Md, mandible; Ns, nasal; PG, pelvic girdle; Pr, parietal; Ra, radius; Sc, scapula; SO, supraoccipital, Ti, tibia; Ul, ulna. Scale Bars = 1mm. Controls, OR26F8, and OCAGF8 samples are shown at the same magnification.

136 (Figure 22F), likely due to the later initiation of OC-Cre expression. One notable similarity,

though, was that the parietal bones had the most ossification and were least affected when

either Msx2-Cre or OC-Cre were employed. There were also facial patterning differences between the OC-Cre and Msx2-Cre models, likely due the difference in expression domains between the two Cre transgenes. While the snouts of the Msx2-Cre mutants were severely shortened (Figure 22F), this shortening was not as apparent in the OC-Cre mutants (Figure

35D, E; Figure 36B, D; Figure 37A-C). With respect to additional craniofacial bones that

form via intramembranous ossification, both OR26F8 and OCAGF8 mandibles were shorter

than the controls and did not ossify normally, but instead were partially replaced by non-

stained matrix (Figure 37J-L). Similarly, the bones of the maxilla and palate had severely decreased ossification, and were mostly replaced by non-stained matrix, particularly in the

OCAGF8s (Figure 37D-I). In marked contrast, the bones of cranial base, as well as the supraoccipital, formed by endochondral ossification, were less affected as were the bones of the appendicular and axial skeleton (Figure 37G-I, M-R, Figure 38). Thus, while the

endochondral forming limbs and ribs of both the OM26F8 and OCAGF8 mice show

patterning defects and have reduced growth compared with the controls, they appeared to

ossify normally, staining strongly with Alizarin red with no unstained matrix present.

Therefore, increased Fgf8b expression, derived from either adjacent tissues or from within the developing bone itself, had a greater impact on the process of intramembranous ossification than it did on endochondral bone differentiation when Fgf8b overexpression began at E14.5.

137

Figure 38. OCAGF8 ribs have patterning defects, but ossify. (A-B): Bone and cartilage staining of E18.5 control (A) and OCAGF8 (B) ribs. Images are shown at the same magnification.

Col2a1-Cre

Col2a1-Cre is under the control of the Col2a1 promoter. Col2a1 is one of the principal markers of chondrocyte differentiation and thus Col2a1-Cre is expressed in a chondrocyte-specific pattern. Previous studies have shown that Col2a1-Cre expression begins at E8.75 in the mouse; by E9.5 expression is detected in the somites and the cranial mesenchyme destined to give rise to cartilage. Finally, by E11.5, Col2a1-Cre expression is also detected in a cartilaginous pattern in the long bones [355]. To validate these findings and analyze the expression pattern of Col2a1-Cre mice at later timepoints, I bred Col2a1-Cre mice with Rosa26 (R26R) reporter mice. Embryos were then stained with X-gal to detect

β−galactosidase activity. As expected, Col2a1-Cre drove recombination in the somites, cranial mesenchyme destined to give rise to cartilage, and the long bones at E12.5 and E13.5

138 (Figure 39). By E13.5, Col2a1-Cre was detected in the developing digits and ribs as well

(Figure 39C-D). Thus, Col2a1-Cre’s expression pattern was consistent with the known chondrocyte-specific pattern. Because cartilage is a precursor in endochondral, but not intramembranous, forming bones, Col2a1-Cre can therefore be used to study the effects of early Fgf8 overexpression on endochondral, but not intramembranous forming bones.

The gross morphology of Col2a1-Cre;CAGF8 embryos, referred to as ColCAGF8, was first visibly different at E13.5. Compared to the controls, ColCAGF8 embryos were elongated

(Figure 40A) and had an altered craniofacial shape, with a smaller jaw and less distinctive boundary between snout and cranium (Figure 40B). Cartilage staining revealed that the cartilage development was also severely impaired in the ColCAGF8 embryos. At E12.5,

Figure 39. Embryonic expression of Col2a1-Cre. (A-D): Col2a1-Cre mediated recombination was visualized using β- galactosidase staining (blue) of RosaLacZ E12.5 (A-B) and E13.5 (C-D) embryos. (A, C): Lateral view of the embryos. (B-D): Dorsal view of the embryos. Separation between head and body shown in the E13.5 embryos is from handling during imaging. Scale bars=1 mm.

139

Figure 40. ColCAGF8 embryos have severe cartilage formation defects. (A-C, G-H): Lateral view of E13.5 (A-C) and E18.5 (G-H) control and ColCAGF8 embryos, as labelled. (D-F): Cartilage staining of E12.5 (D-E) and E14.5 (F) embryos. Scale bars= (A, F-H): 2mm; (B-E):1 mm.

controls had well-organized cartilage throughout the vertebral column and limbs (Figure

40D). In contrast, almost no cartilage was visible in the ColCAGF8 embryos and the little

140 cartilage staining that was apparent was not organized into any recognizable structure (Figure

40E). At E13.5, while there was some cartilage staining, it was still not organized into

recognizable structures (Figure 41C, F). By E14.5 the defect was even more striking; though the controls had distinguishable cartilage elements throughout the vertebrae, limbs, cranial base, and nasal region, ColCAGF8 embryos had no cartilage whatsoever (Figure 40F). Thus, the cartilage visible at E12.5 and E13.5 in the ColCAGF8 embryos did not continue to develop and either was eliminated or made so diffuse as the embryo grew that it was no longer apparent. Interestingly, though ColCAGF8 embryos were longer than the controls at

E12.5 and E14.5, they did not continue to grow past E14.5, but were not reabsorbed (Figure

40G). Rather, by E18.5 they had adopted various shapes that seemed to be based on the pressures exerted upon them by their surroundings (Figure 40G, H). This was likely because

without a skeleton they had no structural integrity. In summary, ColCAGF8 embryos had

severe defects in growth and cartilage development, suggesting that early overexpression of

Fgf8 can impair endochondral bone formation.

However, this conclusion is confounded by Col2a1-Cre’s early expression throughout

much of the embryo due to the paracrine signaling ability of Fgf8. Since the embryo is

smaller at early stages, higher levels of Fgf8 overexpression from the somites and ribs may

affect the development of critical internal organs. Therefore, I next examined the effect of

moderate levels of Fgf8 overexpression using R26F8. Here, I hypothesize that more

moderate levels may not impact the developing cartilage or other organs as severely and thus

will allow a more thorough evaluation of Fgf8’s role in cartilage and endochondral bone

development. Indeed Col2a1-Cre;R26F8 embryos, referred to as ColR26F8, had a less severe

phenotype than ColCAGF8 embryos. At E13.5, control embryos had well-defined cartilage

141 in the limbs, ribs, vertebral column, and cranial base (Figure 41A, D). While cartilage had developed in the limbs and ribs of ColR26F8 E13.5 embryos, there was less cartilage in the vertebral column. (Figure 41B, E). Further examination of ColR26F8 cartilage and

Figure 41. E13.5 ColR26F8 and ColCAGF8 embryos have ectopic cranial cartilage formation. Cartilage staining in E13.5 control (A, D), ColR26F8 (B, E), and ColCAGF8 (C, F) embryos. Red arrows point to regions of ectopic cranial cartilage. Scale Bars= 1mm. All embryos shown at the same scale.

142 endochondral bones at later timepoints is necessary to better tease out this phenotype and the

Fgf8’s role in endochondral bone development. Strikingly, ColR26F8 embryos also had

cartilage formation in the cranial vault, as the cranial base cartilage extended into the lateral

regions of the vault (Figure 41B, E). Similarly, in the E13.5 ColCAGF8 embryos, there were

also traces of cartilage in the cranial vault, though this cartilage was less defined structurally

(Figure 41C, F). Given that Col2a1-Cre is not expressed in the cranial vault, this phenotype is likely the result of paracrine signaling and is reminiscent of the ectopic cartilage formation seen in Chapter 4. Studies using the ColR26F8 and ColCAGF8 models are preliminary, but these initial results suggest these models could offer interesting insights into the early stages of cartilage and bone development. One caveat is that certain studies may be confounded by off-target effects of Fgf8, as it can signal to other surrounding tissues through paracrine signaling. Thus, to further my studies on the sensitivity of bone formation to Fgf8 overexpression during early development, I next utilized a Cre recombinase expressed in the exclusively in the limb and frontonasal prominence mesenchyme.

Creface

The lack of growth in the ColCAGF8 embryos and the potential for confounding Fgf8 signaling to other critical tissues made the results utilizing the Col2a1-Cre difficult to interpret. Thus, to examine the effects of early overexpression of Fgf8 on specific endochondral and intramembranous forming bones, I next utilized a Cre-recombinase created in the Williams lab, AP-2Cre [54]. AP-2Cre (referred to as Creface) is expressed early in the limb and frontonasal prominence (FNP) mesenchyme, but with a more limited expression pattern than Col2a1-Cre. Thus, Creface could be used to examine the effects of early overexpression of Fgf8 on endochondral forming bone, without the potential complication of

143 negatively impacting critical abdominal organs. Previous studies have described the

generation of the Creface transgenic line, in which Creface expression is driven by the minimal human AP-2α promoter and a frontonasal prominence (FNP) and limb enhancer element (Figure 42A). Creface expression begins at E9.0 in the cranial neural crest cells

(Figure 42B); by E9.5 expression is strong in the FNP mesenchyme, but faint in forelimb buds (Figure 42C). Finally, from E10.5 onwards Creface expression is strong in both the

FNP and limb bud mesenchyme (Figure 42D-G). This results in strong Creface-mediated recombination in the ulna, radius, digits, nasal bones and suture, nasal cartilage, frontonasal suture, and interfronal suture in P1 and P15 mice (Figure 43) [54]. Thus, using Creface in combination with CAGF8 will result in overexpression of Fgf8 in both endochondral and intramembranous forming bones associated with the limbs and face, respectively.

Creface;CAGF8 mice, referred to as FaceCAGF8, were born at the same size as controls and in normal Mendelian ratios. Although after birth the FaceCAGF8 mice grew slower than their littermate controls and thus were smaller (Figure 44), unlike the MCAGF8,

OCAGF8, or ColCAGF8 mice the FaceCAGF8 mice could survive birth. Nevertheless,

FaceCAGF8 mice were immediately recognizable by their craniofacial and limb defects. I’ll first focus on the defects in the head.

FaceCAGF8 mice had craniofacial shape abnormalities, with a shortened snout and hypertelorism (qualitative observations, Figure 44C, F and Figure 45A-B, D-E). Though the

FaceCAGF8 pups facial shape shared similarities with the MCAGF8 mice, unlike the

MCAGF8 mice they did not have cleft palate (Figure 45F). As with the MCAGF8s,

FaceCAGF8 also had skeletal defects. At P0, control mice had fully developed nasal bones and the frontal bones were nearly meeting at the midline (Figure 45G). In contrast, the

144

Figure 42. Characterization of AP-2CRE (Creface) transgene activity in mid- embryogenesis using the R26R reporter strain.

145

Figure 42. Characterization of AP-2CRE (Creface) transgene activity in mid- embryogenesis using the R26R reporter strain. (A) Schematic representation of AP-2CRE (Creface) transgene that contains the human AP-2α promoter and intron 5 frontonasal prominence (FNP) and limb enhancer elements. In whole-mount β-galactosidase assays, Creface activity begins in the cranial neural crest cells (nc) in the lateral regions of the head at approximately (B) E9.0. (C) At E9.5, activity is strong in the FNP, mesonephros (m), and is faint the forelimb buds (arrowhead). The arrow indicates the migratory path of cranial neural crest cells. (D) By E10.5, activity is strong in both FNP and limb bud mesenchyme. Vascular tissue (v) in the head and a small portion of the first branchial arch (*) are also positive. (E) A majority of distal cells have been marked by Creface transgene recombination in sections through the middle region of the FNP (upper) and forelimb bud (lower). (F) At E12.5, Creface activity is most robust around nasal pits in the center of the face, as seen in a sagittal section through FNP (G), as well as in the distal regions of fore- and hindlimbs (F). Note: Numbered lines in D and F designate planes of sections in E and G, respectively, and fb indicates the forebrain in E (top). Size bars represent 1.0 mm. Figure used with permission from Nelson and Williams, 2004 [54].

146

Figure 43. Creface transgene activity in skull and limb derivatives. Whole-mount β-galactosidase assays of Creface/R26R positive skulls at P1 (A) and P15 (B) show strong activity is present in nasal bones (n) and suture, the nasal cartilage (nc), the frontonasal suture (arrowheads), and the metopic (interfrontal) suture (ms). In serial sections at P1 (at the level of the arrow in A), Creface mediated β- galactosidase activity is robust in (C) the nasal bones and suture (ns), and (D) the septal cartilage (sc), and nasal cavity mucosa (m). Activity is not present in the olfactory epithelia (oe), or mucosal arteries (a). (E) In skinned P5 limbs, the distal two-thirds of the zeugopod (z) is positive, including digits (d) and skeletal muscle (sm). (F) Sagittal section through the handplate at E15.5 to illustrate Creface-mediated β-galactosidase activity is present throughout the mesenchyme, including cells that give rise to the skeletal condensations of the digits (d). Note that the overlying ectoderm (e) lacks β-galactosidase activity. (G) Creface activity has also targeted the cells giving rise to the radius (r) and ulna (u). Size bars represent 1.0 mm. Figure used with permission from Nelson and Williams, 2004 [54]

147

Figure 44. FaceCAGF8 can survive past birth, but are grow slower than their littermate controls. (A-F): Gross morphology of control and FaceCAGF8 mice at P0 (A), P11 (B- C), and P39 (D-F). Images of FaceCAGF8 and their littermate controls are shown at the same magnification.

148

Figure 45. FaceCAGF8 mice exhibit craniofacial shape and skeletal defects. (A-F): Gross morphology of P0 control (A-C) and FaceCAGF8 (D-F) heads from a lateral (A, D), dorsal (B, E), and ventral (secondary palate (C, F)) view. (G-L): Bone and cartilage staining of control (G-I) and FaceCAGF8 (J-L) heads at P0 (G, J) and E15.5 (H-I, K-L). Dorsal view shown in G-H, J-K; ventral view shown in I, L. Yellow dotted lines note the boundary of the bone. Yellow arrowheads mark regions where there is less ossification in the FaceCAGF8 embryos than in the control. Yellow stars denote skeletal damage from processing - there is no true difference between the skeletons in this region. Scale Bars= 1 mm.

149 FaceCAGF8 mice had severely underdeveloped nasal bones and the frontal bones were less developed as well, with more space between them than seen in the controls (Figure 45J). The

underdevelopment of the nasal, frontal, palatal, and even some bones of the cranial base was

apparent as early as E15.5 (Figure 45K-L). As Creface is not expressed in the bones of the

cranial vault and base, defects in the growth of these bones are likely due to paracrine

signaling of Fgf8 from the regions where it is expressed. The development of the bones

where Creface was not expressed was only delayed, not permanently impaired, as by P15 the

frontal, parietal, and cranial base bones had fully developed in the FaceCAGF8 skulls

(Figure 46A, B, D). Additionally, by P15, the defects in the FaceCAGF8 palatal and nasal

bones were no longer as prominent as they were in the neonates (Figure 46B, D).

Interestingly, by P12 the lateral part of the nasal bones had formed; however, the midline of

the nasal bones developed into cartilage, rather than bone (Figure 46C). Over time, this

cartilage no longer stained with alcian blue (Figure 46A, D, E) and appeared similar to the

non-stained matrix seen in the MCAGF8 mice (Figure 22 and Figure 24). Thus, I examined

the FaceCAGF8 nasal cartilage for defects during early development. At E14.5, the nasal

cartilage in the controls stained strongly with distinct boundaries (Figure 47A, C). While the

nasal cartilage was present in the E14.5 FaceCAGF8 embryos, the cartilage did not stain as

strongly and the boundaries of its formation were less distinct Figure 47B, D). In summary,

Fgf8 overexpression led to craniofacial shape differences, delayed growth of cranial skeletal

elements, and impaired cartilage development followed by ectopic cartilage formation in the

nasal bones.

FaceCAGF8 also had limb patterning and skeletal defects, namely syndactyly and

defects in distal outgrowth. Control mice had five distinct digits (Figure 48A, D-E, J-K) in

150

Figure 46. Post-natal cranial skeletal staining. Skeletal staining of P12 (C), P15 (B, D), and P20 (A, E) control (A, top) and FaceCAGF8 (A, bottom, B-E) skulls. Dorsal view (A, C-D); ventral view (B). Abbreviations: Fr, frontal; Pr, parietal. Scale Bars= 1 mm.

151

Figure 47. Cartilage staining in the E14.5 FaceCAGF8 head. Cartilage staining in E14.5 control (A, C) and FaceCAGF8 (B, D) heads. Red stars indicate regions of impaired cartilage formation. X indicates damage done during processing. Scale bars= 1 mm.

both the forelimb and hindlimb. In contrast, during embryonic development and at birth,

FaceCAGF8 mice had completely penetrant polysyndactyly, with no identifiable digits, but rather an nonpatterned region of cartilage/bone (Figure 48A, G-H). Thus, it was not possible to determine the number of digits for the FaceCAGF8 mice. Postnatally, appearance of the limb varied somewhat in the FaceCAGF8 mice. All of the mice had syndactyly; however, in some limbs, certain digits were identifiable, even if partially or mostly fused (Figure 48M-

N). The limbs were also shorter in FaceCAGF8 mice (Figure 48C, G-I, L-N) than in the

controls (Figure 48B, D-F, J-K) and had ectopic bones in the glenohumeral joint connecting

the humerus and scapula (Figure 48L,O). The elbow joint was also fused (Figure 48L)

Ossification of the radius and digits was delayed; in the E15.5 and neonate forelimb, the 152

Figure 48. Limb defects in FaceCAGF8 mice.

153

Figure 48. Limb defects in FaceCAGF8 mice. Wholemount and skeletal staining of E15.5 (A-C), P0 (D-I), and P15 (J-O) limbs. (A): Ventral view of E15.5 control (left) and FaceCAGF8 (right) embryos. (B-C, F-I, L-O): Bone and cartilage staining of control (B, F) and FaceCAGF8 (C, I, L, O) forelimbs. O shows the glenohumeral joint in the black box in L in greater detail. (D, G, J, M): Control (D, J) and FaceCAGF8 (G, M) forelimbs. (E, H, K. N): Control (E, K) and FaceCAGF8 (H, N) hindlimbs. 1 denotes the 1st digit (anterior), 5 denotes the 5th digit (posterior). Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars= 1mm.

154 radius and limbs had only developed as a cartilage, but by P15 had developed into bone

(Figure 48C, I, L). Though development was delayed, the cartilage of the radius and digits

stained normally with alcian blue, unlike the unstained matrix observed in the cranial vault of

the OCAGF8, OR26F8, and MCAGF8 mice or in the P15 and P20 nasal bones of the

FaceCAGF8 mice. The only possible exception to this was that preliminary evidenceshowed

there was some unstained tissue alongside the stained bone in the P15 FaceCAGF8 forelimb digits (Figure 48L). Further bone and cartilage analysis of postnatal FaceCAGF8 limbs would need to be done to determine if this nonstained matrix is a regular feature of the postnatal limbs and, if so, whether this tissue more closely resembles bone or cartilage histochemically. In summary, FaceCAGF8 limbs had patterning defects including defects in limb outgrowth and digit identity. Additionally, while ectopic bones were observed in the joints, other limb bones, including the radius and digits had delayed ossification.

In the MCAGF8 mice, Fgf8 overexpression led to aberrations in normal WNT signaling which contributed to the loss of ossification and shift in cell fate (Figure 27 and

Figure 31). Therefore, I next examined WNT signaling in the FaceCAGF8 heads using the

Axin2lacZ reporter to determine if WNT signaling was also perturbed in this model, and thus

possibly contributing to the formation of the unstained matrix in the nasal bones. Compared

to the controls (Figure 49A, C) at E12.5, the FaceCAGF8 embryos have decreased WNT

signaling in the FNP (Figure 49B, D). This aligns nicely with the model from the MCAGF8

mice, where Fgf8 overexpression leads to decreased WNT signaling which results in a shift

from osteogenic to chondrogenic cell fate. To further delineate this hypothesis, I also

examined Osx expression, a marker of ossification, in the FaceCAGF8 heads during early

osteogenesis. At E14.5, Osx expression in the FaceCAGF8 primordial frontal bone was

155

Figure 49. FaceCAGF8 E12.5 embryos have decreased WNT Signaling in the FNP. Axin2lacZ expression as visualized by β−galactosidase staining, in Ctrl (A, C) and FaceCAGF8 (B, D) E12.5 heads. Red arrows denote regions of reduced Axin2lacZ expression.

decreased compared to the control (Figure 50). Thus, this preliminary work supports the hypothesis that FaceCAGF8 skull bones have impaired osteogenic differentiation, like in the

MCAGF8s. Further work examining other other markers of osteogenesis and chondrogenesis would better elucidate the similarities and differences between these two models.

156

Figure 50. Decreased Osterix expression in the FaceCAGF8 frontal bone. Osterix expression, visualized through ISH, in E14.5 control (A) and FaceCAGF8 (B) primordial frontal bone. Scale Bars= 200µΜ.

Given that WNT signaling has been shown to function downstream of HH signaling

pathways during development of the osteoblast lineage [352], I next explored the possibility

that Fgf8 overexpression could be downregulating WNT signaling through modulating HH signaling. To examine HH signaling expression and patterning, I utilized a PatchedLacZ allele, referred to as PtcLacZ in conjuction with FaceCAGF8, as Patched (Ptc) is a reporter of

HH signaling. At E12.5, PtcLacZ expression in the control heads was mainly limited to the developing vibrisaae, with minimal expression in the nasal pits and surrounding the oral cavity (Figure 51A, C). Similarly, the FaceCAGF8 embryos had expression in those locations, with increased expression in the lateral regions surrounding the oral cavity (Figure

51B, D). Larger differences in expression patterning were seen in the limbs. PtcLacZ expression in the control forelimbs was clearly demarcated in the three most posterior digits at E12.5 (Figure 51A, E). However, in the FaceCAGF8 forelimbs, PtcLacZ expression was evenly spread across the posterior side of the forelimb, without any clear demarcation of digits (Figure 51B, F). Thus, Fgf8 overexpression increases the intensity (face) or range

157

Figure 51. PtcLacZ expression in E12.5 control and FaceCAGF8 embryos. (A-F): Expression of Patched, a target gene of Hh signaling, was visualized using β-galactosidase staining (green-blue) of E12.5 FaceCAGF8;Ptci (B, D, F) and littermate PtcLacZ controls (A, C, E). (A-B): Lateral view of the embryos. (C-D): Front of the face. (E-F): Forelimbs. White arrowheads point to regions of increased PtcLacZ expression in the face. 1 marks the anterior side of the limb, 5 the posterior side. Scale bars=500µM.

(limbs) of HH signaling in E12.5 FaceCAGF8 embryos. Given the small range of increased

PtcLacZ signaling in the FaceCAGF8 face compared to the larger range of the decreased WNT

158 signaling, it is unlikely that increased HH signaling is the sole contributor to the decreased

WNT signaling in the head. However, the loss of HH signaling patterning in the limbs may

play a role in the loss of digit patterning or delayed ossification observed in the FaceCAGF8

mice. Further examinations of these pathways at additional timepoints would be helpful to

better understanding their connections.

Discussion

Fgf and Fgfr gene expression is critical for the formation of both endochondral and

intramembranous forming bones. Though the expression pattern of certain FGF signaling

members differs between the two types of developing bone, it has not yet been delineated

how or if these differences impact cell fate of the osteogenic mesenchyme [128]. In the

previous chapter, increased Fgf8 expression had a striking impact on the development of intramembranous bones of the cranial vault, shifting cell fate from osteogenic to chondrogenic. Here, I have probed this pathway further by manipulating the dosage of Fgf8 in the developing mesenchyme of both intramembranous and endochondral bones using three mesenchymally expressed Cre-recombinases – Osteocalcin-Cre, Col2a1-Cre, and Creface -- in conjunction with the Fgf8 overexpressing alleles, R26F8 and CAGF8. Together, these data suggest that Fgf8b overexpression has a more severe impact on intramembranous ossification than endochondral ossification. Additionally, the effect of ectopic Fgf8b on bone development is partially dependent on the timing and location of expression.

In Chapter 4, the expression of the Msx2-Cre transgene in the ectoderm overlying the calvaria beginning around E10.5 enabled me to uncover how ectopic Fgf8 expression could impact skull development. However, since this transgene was not uniformly expressed throughout the surface ectoderm of the whole embryo it was difficult to generalize how this

159 ligand might impact endochondral versus intramembranous ossification. Therefore, to

address this issue, I utilized the OC-Cre transgene that is expressed in all bone precursors, although this activity begins later, at around E14.5. With respect to the calvaria, both of these transgenes resulted in the transformation of bone to abnormal non-stained matrix - cartilage in Msx2-Cre and presumptive cartilage in OC-Cre - indicating that this effect can occur whether Fgf8 is expressed from an earlier timepoint in the ectoderm or later in the bone progenitors. Interestingly, both transgenes also had similar differential effects on particular components of the skull vault. In this respect, previous studies suggest that osteoblast location within the skull may impact the degree of FGF signaling activation. Osteoblasts from neural crest-derived bones, such as the frontal bone, express FGF osteogenic ligands and their receptors at higher levels than osteoblasts from paraxial mesoderm-derived bones, such as the parietal bones, suggesting the frontal bone is more sensitive to FGF signaling than the parietal bones [127, 343]. My results using the OR26F8 and OCAGF8 mice support this conclusion as the parietal bones, while not immune to Fgf8’s effects, are less severely affected than the frontal bones. While the parietal bones were also less affected in the Msx2-

Cre models, there was concern that this phenomenon may be partly due to Msx2-Cre’s expression pattern. The additional of the OR26F8 and OCAGF8 mice thus adds additional support to the hypothesis that overexpression of Fgf8 is more detrimental to cranial neural crest-derived bones than to paraxial mesoderm-derived bones. Future gene expression

studies on individual sutures and their associated progenitors may help to further elucidate

why particular bones and sutures respond differentially to alteration of specific signaling

pathways [356, 357].

160 Outside the calvaria, there were clear differences between the phenotypes obtained

with Msx2-Cre and OC-Cre, presumably due to differences in timing and location of ectopic

Fgf8 activation. In the context of the head, Msx2-Cre mutants had additional patterning

defects – a shortened snout, and in the MCAGF8 mutants, a domed skull and missing eye that

were either not present or as severe in the OC-Cre mutants. The eye defects are likely

present in the Msx2-Cre, but not OC-Cre, mutants because the eye is derived from the

ectoderm where Msx2-Cre but not OC-Cre is expressed. FaceCAGF8 mice (to be discussed

in greater detail in the latter half of this discussion) also had a shortened snout; as Creface is

expressed earlier than OC-Cre, this suggests that this defect is sensitive to the timing of Fgf8

overexpression. Other defects present in the Msx2-Cre mutants, but not the mesenchymal

Cre mutants OC-Cre and Creface, such as the loss of the eye, are more likely sensitive to the

location of Fgf8 overexpression, rather than the timing. The limited expression pattern of

Msx2-Cre also did not allow an assessment of how ectopic Fgf8 expression could impact the

development of additional skeletal structures, such as the lower jaw. Here, the more

widespread expression of Fgf8 in the OC-Cre mutants revealed that additional

intramembranous bones within the palate and lower jaw were also altered to form a non-

stained tissue, which I suspect, given the similarity to the non-stained tissue seen in Msx2-

Cre mutants, is a shift to a cartilaginous cell fate. Though the OC-Cre mutants did not survive long enough to assess the fate of this non-stained tissue, I postulate that like the cartilage in the MCAGF8 mice, it would not subsequently undergo endochondral ossification, but remain cartilaginous.

There were also patterning defects in the endochondral forming bone of the OR26F8 and OCAGF8 mice. The hindlimbs and forelimbs of both OR26F8 and OCAGF8 mice were

161 shorter than in the controls. Similarly, the ribs and other bones of the trunk axial skeleton in

the OC-Cre mutants were also smaller. However, there was ossification in all of the

endochondral forming bones and, thus, no general transformation of bone to cartilage.

Similarly, the MCAGF8 limbs had patterning defects, but underwent ossification, indicating that endochondral bone phenotype does not result from Cre expression in a unique tissue or timepoint. In concordance with my results, Fgf8 has been shown to be required for outgrowth and patterning, but not ossification, in the limbs [89, 347], while generalized transgenic expression of Fgf2 also causes shortening and bending of the radius and ulna

[314]. Taken together, these results indicate that intramembranous ossification is more sensitive to Fgf8 overexpression than endochondral ossification.

While the skeletal results from OR26F8 and OCAGF8 mice suggest that increased

Fgf8 expression has a more severe impact on the development of intramembranous bone than endochondral bone, one caveat is that OC-Cre is not expressed until late in development, after mesenchymal precursor cells have begun to ossify. Thus, it is possible that endochondral forming bones are only less sensitive to high levels of Fgf8 when Fgf8 is expressed at later timepoints. Therefore, I utilized two different Cre-recombinases that are active during earlier stages in development; the first of which was Col2a1-Cre.

Col2a1-Cre is expressed in all cartilage precursors starting during early development.

Given the lack of a cartilage intermediate during intramembranous ossification, it is thus only suitable to study the effects of Fgf8 overexpression on endochondral forming bone. The other disadvantage of this Cre is that because it is expressed so broadly during early development, off-target effects may occur when its used in combination with ligands that signal in a paracrine fashion, such as Fgf8. This is because during early development the

162 embryo is smaller and thus the range of Fgf8 may extend to critical organs outside of the developing cartilage. This phenomenon may be responsible for the ColCAGF8 phenotype,

where the ColCAGF8 embryos had very little cartilage development, but also do not develop

beyond E14.5. Preliminary studies in the ColR26F8 mice, where Fgf8 is overexpressed at

more moderate levels, supports this conclusion as these mice both survived longer and had

more normal cartilage. Also, in concordance with the idea that paracrine signaling can result

in Fgf8 having a wide range of expression or influence in the early embryo, overexpression

of Fgf8 in both the ColR26F8 and ColCAGF8 E13.5 embryos led to ectopic cartilage

formation in the cranial vault, where Col2a1-Cre is not expressed. Thus, while further work using the ColR26F8 and ColCAGF8 mice may lead to interesting insights into the first stages of endochondral development, the off-target effects of this model led me to explore the effects of Fgf8 overexpression in the endochondral and intramembranous bones using another early mesenchymally expressed Cre-recombinase, Creface.

Like Col2a1-Cre, Creface is expressed in the early mesenchyme. However, its expression is limited to the FNP and limb mesenchyme, and thus, the paracrine actions of

Fgf8 signaling are less likely to impair critical organs. Indeed, the FaceCAGF8 mice survived into adulthood. Similar to the Msx2-Cre and OC-Cre mutants, the FaceCAGF8 mice had unstained matrix replacing intramembranous bone formation in the nasal bones.

This unstained matrix actually did stain with alcian blue during the early stages of its development, but no longer stained at latter stages, suggesting that it is an abnormal cartilage, as seen in the MCAGF8 mice. However, further histochemistry at multiple stages of development are needed to better characterize this tissue. As with the MCAGF8 embryos, the shift from bone formation to unstained matrix (presumptive cartilage) in the FaceCAGF8

163 mice may be due to decreased WNT signaling impairing differentiation of the nasal bones, as

WNT signaling and Osterix expression were decreased in the FaceCAGF8 mice. The unstained matrix in the FaceCAGF8 mice prevailed postnatally, providing further evidence that in the intramembranous bones, overexpression of Fgf8 leads to a shift from bone to cartilage formation and that this cartilage does not later ossify. In contrast, the endochondral forming bones in the FaceCAGF8 limbs had delayed ossification. While certain limb elements were cartilaginous at birth, they later ossified during postnatal development. Thus, the FaceCAGF8 data revealed that while early overexpression of Fgf8 can delay ossification in endochondral forming bones, the bones do ossify, unlike the FaceCAGF8 intramembranous forming bones. Taken together with the ossification seen in the OR26F8 and OCAGF8 endochondral, but not intramembranous forming bones, this suggests that intramembranous ossification is more sensitive to increased levels of Fgf8 than endochondral ossification. This may reflect a difference in the evolutionary history of the two types of ossification. Further work on the role of Fgf8 signaling in both types of ossification may reveal if Fgf8 is a critical regulator in the decision of early mesenchymal cells to undergo endochondral or intramembranous ossification.

164 CHAPTER VI

INCREASED HEDGEHOG SIGNALING FROM THE ECTODERM AND

MESENCHYME IMPAIRS CRANIOFACIAL AND LIMB DEVELOPMENT

Introduction

Reciprocal interactions between the epithelia and underlying mesenchymal cells

regulate the morphogenesis of many organs, including the limb and head, which will be the

main focus of this chapter. In both the head and limbs, tissue-tissue interactions between the

ectoderm and mesenchyme are regulated, in part, by signaling centers.

In the head, the developing surface ectoderm houses signaling centers that are

important for directing craniofacial development, such as the frontonasal ectodermal zone

(FEZ) and the nasal pits [52, 358–361]. The FEZ regulates outgrowth and dorsoventral

patterning in the face, particularly of the frontonasal prominence (FNP). The FEZ is defined

by the juxtaposition of a Sonic hedgehog (Shh) expression domain and a Fibroblast growth factor 8 (Fgf8) expression domain. While the FEZ was first discovered in chick, a similar

region has since been identified in both mice and humans [52, 53]. However, the impact of

alterations in HH signaling from the ectoderm in mice has not been well examined, despite the fact that dysregulation in HH signaling is known to cause clinical disorders in humans, including holoprosencephaly, hypotelorism, hypertelorism, cleft lip/palate, and other skeletal and craniofacial deformities [197, 200, 201, 362]. In regard to cranial bone ossification it is clear that HH signaling is an important regulator of both osteogenesis and cranial suture biology. However, studies examining the role of HH signaling in cranial bone ossification and cranial suture morphogenesis are relatively new and thus far have produced conflicting evidence to even basic questions such as whether HH signaling acts as an inhibitor or inducer

165 of cranial ossification/suture fusion [197]. Nevertheless, it is apparent that HH signaling in

mesenchymal derived tissue (bone) also is important for normal cranial morphogenesis.

In the limb, patterning is regulated by two major signaling centers: the apical ectodermal ridge (AER) and zone of polarizing activity (ZPA). The AER is an epithelium

that runs anterior to posterior (A-P) along the limb bud, separating the dorsal and ventral

sides of the limb bud. Proximal-distal (P-D) growth is regulated by the AER, which secretes

FGF ligands that signal to the underlying limb mesenchyme [74, 77]. The underlying

mesenchyme contains the other major limb signaling center, the ZPA. The ZPA regulates A-

P growth and is found in the posterior mesenchyme where it secretes SHH, which is

responsible for the ZPA’s polarizing activity [78–80]. Epithelial -mesenchymal feedback

loops between the ZPA and AER are required for normal limb development.

Dysregulation of SHH in the limb results in defects in A-P limb patterning. In mice,

loss of SHH leads to a severe limb phenotype characterized by the loss of posterior digit

development (2-5); in contrast, grafting posterior (SHH containing) grafts onto the anterior

limb leads to extra digits displayed as mirror image duplications [90–93]. Similarly, in

humans, defects in hedgehog signaling cause several inherited limb disorders. Specifically,

these malformations include both pre-axial and post-axial polydactyly as well as syndactyly

[222]. However, these defects are not actually due to alterations in the coding sequence of

Shh, as mutations within Shh have not been associated with malformations. This is

presumably because mutations in Shh are not compatible with life. However, point

mutations in the ZRS enhancer, which lead to ectopic Shh expression specifically in the limb

bud, are found in patients with pre-axial polydactyly and triphalangeal thumb polydactyly

(TPTPS). TPTPS patients can have either pre-axial or post-axial polydactyly [223]. This

166 suggests that point mutations in the ZRS perturb normal regulation of SHH in the posterior,

as well as the anterior, limb bud. Additionally, polydactyly is found in patients with Gli3

mutations, likely through altering the domain of SHH so it is no longer restricted to the

posterior limb buds [224, 225]. However, it is not understood whether ectopic Shh

expression directly causes polydactyly or if increased HH signaling dysregulates the

expression of other genes/pathways and those genes/pathways act as secondary messengers

to cause polydactyly.

While the role of ectopic expression of Shh in the anterior limb bud mesenchyme on

A-P limb patterning has been established, the consequence of increased HH signaling in the ectoderm is yet unclear. While Shh is classically thought of as being solely expressed in the

ZPA of the limb, newer studies have shown that Shh is also expressed in the AER and that loss of Shh in the AER results in ectopic cartilaginous condensations in the autopod [363].

However, the phenotypic effects of increased HH signaling in the AER has not been determined.

Therefore, though the HH signaling pathway has been extensively studied, there are still gaps in our understanding of its specific roles in the facial and limb ectoderm, as well as mesenchyme. Thus, I utilized two Cre-recombinases, Crect and Creface, in conjunction with the SmoM2 allele, to specifically increase HH signaling in the limb and facial ectoderm and mesenchyme. The SmoM2 mice express a constitutively active form of smoothened. Given that smoothened is inhibited in the absence of HH, constitutive activation of smoothened reproduces the effect of having continuous HH signaling. However, given that the HH ligands are not involved in this model, I can differentiate between the effects of increased HH ligand and increased HH signaling.

167 Results

In this chapter, I will investigate how increased HH signaling from a variety of tissues and timepoints impacts head and limb development. This will be done using SmoM2 mice in combination with a variety of Cre-recombinases. The SmoM2 allele contains a point mutation, W539L, which renders it constitutively active [364]. A cDNA encoding SmoM2 with a COOH-terminal yellow fluorescent protein (YFP) tag was targeted into the R26 locus

3’ to a LoxP flanked polyadenylation stop sequence cassette [204] (Figure 52). Constitutive

Figure 52. SmoM2 allele in mice. Depiction of the mouse SmoM2 allele. Orange triangles represent the LoxP sites. Upon Cre-mediated combination at the LoxP sites, the stop cassette (shown in black) will be removed, allowing for transcription of the SmoM2 allele (green) with a polyadenylated tail (yellow) from the Rosa26 promoter (blue).

activation of smoothened (SMO) leads to up-regulation of the HH signaling pathway [365].

Thus, breeding mice carrying the SmoM2 allele to various Cre lines resulted in tissue specific, cell-autonomous up-regulation of the HH signaling pathway. In the following, I will describe the consequences of upregulation of HH signaling in the mesenchyme compared to upregulation in the ectoderm.

Mesenchymal Upregulation of HH Signaling Results in Cranial and Limb Defects

I first examined the effects of increased HH signaling in the mesenchyme. To examine how manipulation of HH signaling in the mesenchyme could impact development, I utilized SmoM2 mice in concert with the frontonasal prominence (FNP) and limb

168 mesenchyme specific Creface transgene (Figure 42, Figure 43). As described in Chapter 5,

Creface expression begins at E9.5 in the FNP and E10.5 in the limb mesenchyme. The

Creface;SmoM2 mice, referred to as FaceSMO, were born at normal size and in normal

Mendelian ratios, but were easily identifiable from E11.5 onwards by their gross morphological defects in the face and limb. I will first focus on the malformations in the face.

The FaceSMO mice had craniofacial patterning defects including shortened snout and hypertelorism (Figure 53, Figure 54). The most striking deformity, however, was a protrusion of tissue in the region between the eyes and snout, present in ~70% (16/23) of the

FaceSMOs (Figure 53G-I). This protrusion showed a high degree of variability from a single large protrusion (Figure 53G) to a very small protrusion (Figure 53H), and occasionally even a double protrusion (Figure 53I). However, the single large protrusion was the most common phenotype. Finally, the variability in protrusion presence/size was not litter dependent, as all four phenotypes could be found within the same litter.

The 30% of FaceSMOs without a protrusion still exhibited a shortened snout and hypertelorism (Figure 53F), as well as the limb defects (to be discussed later). To examine the degree of craniofacial shape differences in the FaceSMOs, I overlayed arrows on a wildtype skull (Figure 54A) and then placed those same arrows on a FaceSMO skull (Figure

54B), taken from the same litter and imaged at the same magnification as the wildtype. White arrows were placed between the front and back of the eye sockets and red arrows were placed from the tip of the snout to the back white arrow. This allowed me to visualize the degree of hypertelorism and the shortening of the snout as, in the FaceSMOs, the front white arrow did

169

Figure 53. Gross morphology of FaceSMO heads. (A-I) Gross morphology of control (A, C, E) and FaceSMO (B, D, F-I) neonate heads. (A-B): Lateral view; (C-D): Front of face; (E-I): Dorsal view. Note various sizes of the protrusions in G-I and lack of protrusion in F.

170 Figure 54. FaceSMO mice exhibit shortened snout and hypertelorism. (A-B) Bone and cartilage staining of wildtype (A) and FaceSMO (B) neonate skulls, dorsal view. White arrows denote the distance between the eye sockets in the wildtypes. The same sized white arrows are overlayed on the mutant to show respective size difference. The red arrow denotes the distance from the apex of the snout to the back white arrow in the wildtype; the same sized red arrow is overlayed on the FaceSMO head to show respective size difference. Tickmarks on the ruler indicate mm. (C) Graph depicting the average distance between the eyes (mm) in the wildtype (black) and FaceSMO (blue). Blue star indicates significance, p ≤0.01.

not reach the other eye socket and the red arrow extended past the back white arrow (Figure

54B). However, the back white arrow fit perfectly on the FaceSMO head, showing that in the mid-cranium, the width of the head was the same as in the wildtype

(Figure 54B). To better quantify the degree of hypertelorism, I measured the distance between the eyes in the wildtypes and FaceSMOs. The distance between the eyes in the

FaceSMOs was ~15% wider than the controls, increasing significantly (p= 0.0003) from a mean of 5.05±0.10mm to 5.80±0.21mm (Figure 54C). In summary, at birth, FaceSMO mice

171 had craniofacial abnormalities including shortened snout, hypertelorism, and, in 70% of the

FaceSMOs, a protrusion between the snout and eyes.

I was not able to examine the FaceSMOs after P0 due to neonatal lethality. This lethality was likely due to malnutrition as the FaceSMOs had little to no milk in their stomach (Figure 55B, D), though their littermate controls (Figure 55A, C) fed normally.

Additionally, the FaceSMO abdomens became distended and discolored (Figure 55B, D), suggesting the FaceSMOs were swallowing air rather than milk. Thus, I hypothesized that the FaceSMOs may have a cleft palate. However, examinations of the palates revealed that, while the FaceSMOs had minor shape differences in the primary palate and palatine rugae, they did not have cleft palate (Figure 56D compared to Figure 56A). Thus, the feeding difficulties in the FaceSMOs were likely due to the alterations in craniofacial shape.

Despite the lack of major defects on the surface of the palate, bone and cartilage staining revealed that the FaceSMOs did have underlying palatal defects. The bones in and surrounding the palate were underdeveloped and/or under ossified compared to the controls, including the trabecular basal plate, maxilla, premaxilla, palatal process of premaxilla, and the nasal capsule (Figure 56B, E).

I next examined the bones throughout the rest of the cranium. While from a lateral view, the craniums were similar, except for the nasal capsule (Figure 56C, F), a dorsal view revealed there were major defects in ossification in the midline of the nasal bone and extending into the frontal bone (Figure 56K-L). This obliterated the frontonasal suture and the rostral part of the interfrontal suture and resulted in a “hole” in the nasal region of the cranium. This lack of ossification was present irrespective of the presence of a tissue protrusion, suggesting the protrusion was a secondary defect to the primary skeletal defect

172

Figure 55. FaceSMO neonates develop distended abdomens and lack milk sacs. (A-D): Gross morphology of control (A,C) and FaceSMO (B, D) P0 pups. (A-B): Lateral view; (C-D): Ventral view. Note the absence of milk and the swelling and discoloration in the abdomen of the mutants.

173 Figure 56. FaceSMO neonates have cranial and cerebral defects.

174

Figure 56. FaceSMO neonates have cranial and cerebral defects. (A, D) Palates of control (A) and FaceSMO (D) neonates. Note the lack of cleft palate in the FaceSMOs, but the change in shape of the primary palate, outlined in yellow. (B-C, E-F, J-L): Bone and cartilage staining of control (B-C, J) and FaceSMO (E-F, K-L) neonate skulls shown from a ventral (B, E), lateral (C, F), and dorsal (J-L) view. In the ventral and dorsal view the jaw has been removed. Red star denotes defects in the palatal bones; red arrow denotes changed shape of the nasal cartilage. (G-I): Deskinned control (G) and FaceSMO (H, I) heads. (M-O): Control (M) and FaceSMO (N-O) brains. Red “X” marks region damaged by processing. Abbreviations: Ce, cerebrum; mx, maxilla; nc, nasal capsule; pmx, premaxilla; ppmx, palatal process of the maxilla; pppmx, palatal process of the premaxilla, Ol, olfactory lobe.

175 (Figure 56H, K compared to Figure 56I, L). To better understand the composition of this tissue protrusion, I dissected the heads of mice with and without a protrusion. Removal of the skin and skeletal tissue in mice with a protrusion revealed that the protrusion consisted of brain matter (Figure 56O). The rostral region of the brain was greatly expanded and appeared to be an extension of the cerebrum. No distinct olfactory bulb was present. However, further analysis would have to be done to determine the exact composition of the brain protrusion.

In contrast, in mice without a protrusion (Figure 56N), the brains were similar to those found in the control (Figure 56M), but with a slightly misshapen cerebrum and a smaller olfactory bulb, both changes likely due to the constraints put on the brain by the abnormal craniofacial shape. In summary, FaceSMOs exhibited loss of ossification in the nasal region of the

cranial skeleton; in ~70% of the FaceSMOs, brain protruded through the skeletal hole formed by this lack of ossification.

FaceSMO limbs are obviously distinguishable at birth due to shortened limbs and syndactyly in both the hindlimbs and forelimbs (Figure 57A-B, E-F). Bone and cartilage staining revealed additional defects in the limbs: polydactyly, ectopic bone in the joints and distal most tips of the digits, and shortened and thickened bones (Figure 57C-D, G-H). The shortening and thickening of bones was particularly prominent in the bones of the zeugopod

(Figure 57D, H), with the tibia and fibula being so thick they either were fused together

(Figure 57H) or had just a very narrow space between them (data not shown). In contrast, the distal regions of the radius and ulna were farther apart (Figure 57D), likely to accommodate the expanded autopod. The autopod was so impaired that the digits were hard to individually identify, but the additional digits mostly seemed to be added to the anterior side, thus, pre-axial polydactyly. These additional digits did not ossify as well as the normal

176

Figure 57. FaceSMOs exhibit multiple limb patterning defects. Neonate gross (A-B, E-F) and skeletal (C-D, G-H) limb phenotypes. (A, C, E, G): controls, (B, D, F, H): FaceSMOs. Red stars denote ectopic bone in joints; red arrow indicates fused tibia and fibula in the hindlimb. FaceSMOs exhibit polydactyly with 1 indicating the anterior most digit and 5 indicating the the posterior most digit. Abbreviations: Fb, fibula; Fe, femur; FL, forelimb; HL, hindlimb; Hu, humerus; PG, pelvic girdle; Ra, radius; Sc, scapula; Ti, tibia; Ul, ulna.

177 digits and mostly consisted of cartilage, especially in the hindlimb (Figure 57D, H). In summary, the FaceSMO mice exhibited several limb patterning defects including pre-axial polydactyly, shortened limbs, ectopic joints, and thickened zeugopod bones.

Having established the gross and skeletal phenotypes of the FaceSMO mice, I next investigated the molecular underpinnings that could be contributing to the phenotypes. The first signaling pathway that I examined was HH signaling. Given the expression pattern of

Creface and properties of SmoM2, HH signaling should be upregulated in the structures derived from the FNP and limb mesenchyme. To visualize HH signaling in both controls and

FaceSMO mice, I bred a Patched LacZ allele, PtcLacZ, into the FaceSMO background.

Patched is a target gene of hedgehog signaling, and thus serves as a readout of HH signaling activity. At E12.5, PtcLacZ was mainly expressed in the vibrissae, nasal pits, and the posterior limb in the controls (Figure 58A, E-G). In the FaceSMO embryos, these regions of expression were expanded with PtcLacZ expressed throughout the nasal region and limb

(Figure 58A, E-G). Thus, HH signaling was upregulated in the areas derived from the frontonasal prominence and limb mesenchyme in the FaceSMO mice compared to the controls, as expected based on Creface’s expression pattern (Figure 42, Figure 43). Given the brain protrusion seen in ~70% of the mutants, I also checked the degree of HH signaling in the brain at E12.5 and confirmed that while there was some expression of PtcLacZ in the brain, this expression was not expanded or increased in the FaceSMO brain (Figure 58B, D).

Additionally, to better understand the molecular basis for the variability in the brain protrusion phenotype, I examined PtcLacZ expression in FaceSMO neonates with and without a protrusion. While PtcLacZ expression was increased in all FaceSMO neonates compared to the controls (Figure 59), PtcLacZ was more highly expressed in the primary palate and nasal

178

Figure 58. Increased hedgehog signaling in E12.5 FaceSMO embryos corresponds with regions of Creface expression.

179

Figure 58. Increased hedgehog signaling in E12.5 FaceSMO embryos corresponds with regions of Creface expression. A-F): Expression of Patched, a target gene of HH signaling, was visualized using β-galactosidase staining (blue) of E12.5 FaceSMO, PtcLacZ (C-D, H-J) and littermate PtcLacZ controls (A-B, E-G). (A, C): Lateral view of control (A) and FaceSMO (C) embryos. (B, D): Control (B) and FaceSMO (D) heads cut into 2 halves along the sagittal plane. Red arrows point to regions in the FaceSMO embryos with greater Patched expression than in the controls. These areas are shown in greater detail in E-J. (E, J): Control (E) and FaceSMO (H) hindlimbs (HL). 1 denotes the anterior side of the limb, 5 denotes the posterior side. (F-G. I-J): Control (F-G) and FaceSMO (I-J) heads. Scale bars= (A,C): 1 mm; (B, D-J): 500mM..

180

Figure 59. Hedgehog signaling is increased in FaceSMO neonates. (A-F): Expression of Patched, a target gene of HH signaling, was visualized using β-galactosidase staining (blue) of deskinned P0 FaceSMO;PtcLacZ (B-C, E-F) and littermate PtcLacZ controls (A, D). (A-C): Dorsal view of the cranial vault. (D-F): Ventral view of the palate. FaceSMO mice without a brain protrusion are shown in B, E; FaceSMO mice with a brain protrusion are shown in C, F. Red arrows point to regions in the FaceSMO mice with greater Patched expression than in the controls. Scale bars=1 mm.

bones of FaceSMO mice with a brain protrusion than in those without a brain protrusion

(Figure 59B-C, E-F). Thus, variable levels of increased HH signaling in the head may be responsible for the variability in the severity of the FaceSMO skeletal and brain protrusion phenotypes.

To further confirm the increased hedgehog signaling in the FNP and limb of

FaceSMO mice, I next examined the expression pattern of Gli1, another target gene of HH signaling. At E11.5 in the controls there was very little Gli1 expression in the face (Figure

60A) and Gli1 expression was restricted to the posterior half of the limb (Figure 60C). In

181

Figure 60. Increased Gli1 expression in the FaceSMO limb and face mesenchyme. WMISH of Gli1 expression in the face (A-B) and limbs (C-D) of control (A, C) and FaceSMO (B, D) E11.5 embryos. Red arrows indicate regions of increased Gli1 expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars= 500µΜ.

contrast, in the FaceSMO embryos, Gli1 expression was increased in the nasal processes and expanded to the anterior part of the limb (Figure 60B, D). Taken together, these results showed that HH signaling was increased in the anterior limb as well as the region derived from the FNP in the FaceSMO mice.

182 During the establishment of A-P patterning in the normal limb, the mutual antagonism between the transcription factors GLI3 and HAND2 restricts SHH to the posterior limb bud and thus serve as important contributors to digit identity [93–95]. While the Gli3 and Hand2 expression are classically believed to be upstream of Shh [366, 367], given the polydactyly observed in the FaceSMO limbs, I next examined the balance of Gli3 and Hand2 expression. In the head, preliminary evidence suggested that while controls had

Gli3 expression in the FNP, FaceSMOs did not (Figure 61A-B). In the limbs, Gli3 expression was expanded early, and then diminished at later timepoints. At E10.5, in the controls, Gli3 was expressed in the distal mesenchyme of the limb, with slightly more expression in the anterior than the posterior region (Figure 61C). There was a similar expression pattern in the FaceSMO limbs, except with slightly more expression in the posterior region than in the controls (Figure 61D). By E11.5, Gli3 expression in the

FaceSMO limbs had diminished compared to the controls (Figure 61F). At E12.5, Gli3 was expressed between each developing digit in the forelimbs and hindlimbs of controls (Figure

61G, I). While there was also expression of Gli3 between the normal FaceSMO digits, the expression was shallower and there was little to no expression in between the extra digits

(Figure 61H, J). In summary, Gli3 expression in the FaceSMO limbs was increased in the posterior region at E10.5 and then diminished by E11.5. At E12.5, Gli3 is specifically reduced in the mesenchyme separating the digits, and thus represents one possible mechanism behind the syndactyly.

At E10.5 and E11.5, Hand2 expression was restricted to the posterior in the controls

(Figure 62A, C). In the FaceSMOs, Hand2 expression was also restricted to the posterior at

E10.5 (Figure 62B). However, by E11.5, Hand2 was also expressed in the anterior part of

183

Figure 61. Gli3 expression in the FaceSMO face and limb.

184

Figure 61. Gli3 expression in the FaceSMO face and limb. WMISH of Gli3 expression in the face (A-B) and limbs (C-J) of E10.5 (C-D), E11.5 (A-B, E-F), and E12.5 (G-J) control (left column) and FaceSMO (right column) embryos. Blue arrows indicate regions of decreased Gli3 expression; red arrow indicates region of increased Gli3 expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars= 500µΜ.

185

Figure 62. Domain of Hand2 expression expanded in FaceSMO limbs. WMISH of Hand2 expression in the limbs of E10.5 (A-B), E11.5 (C-D), and E12.5 (E-H) control (left column) and FaceSMO (right column) embryos. Red arrows indicate regions of ectopic Hand2 expression. Blue arrows indicate regions of decreased expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars= 500 µΜ.

186 the limb, in a striated, or Turing pattern (Figure 62D). At E12.5, Hand2 was expressed throughout much of the control forelimbs and hindlimbs, particularly in the distal region and on the posterior side (Figure 62E, G). In the FaceSMO limbs, Hand2 had a similar expression pattern, but expression was greatly diminished (Figure 62F, H). Additionally,

expression was more uniform throughout the A-P axis and thus extended further into the

anterior side, particularly in the forelimb (Figure 62F, H). In summary, Hand2 expression in

the FaceSMO limbs was diminished and extended into the anterior, rather than being

restricted to the posterior, as in the controls. Thus, increased HH signaling in the limb mesenchyme can regulate the expression of Gli3 and Hand2, suggesting that there may

reciprocal signaling between these two genes and SHH, rather than one-way signaling, as

previously proposed.

Expression of Bmps has been shown to be important in the regulation of interdigital

cell death [259, 265] as well as the formation of bone [267]. Given the syndactyly in the

FaceSMO limbs and impaired ossification in the FaceSMO nasal bones, I next examined the

expression of Bmp2 and Bmp4. At E11.5, Bmp2 had low expression in the ridges of the

facial prominences (Figure 63A). However, in the FaceSMOs, there was no expression of

Bmp2 in the face at this timepoint (Figure 63C). By E12.5, Bmp2 was expressed in the

vibrissae and surrounding the nasal pits and mouth (Figure 63B). In the FaceSMO E12.5

embryos, similar expression was seen in the vibrissae and surrounding the mouth, but there

was greatly diminished expression surrounding the nasal pits (Figure 63D). Similarly, Bmp2

expression was also reduced in the limbs. While similar expression of Bmp2 in the distal

mesenchyme was observed in both controls and FaceSMO limbs at E11.5 (Figure 63E, H),

by E12.5, the range and intensity of Bmp2 expression was reduced. While Bmp2 was

187

Figure 63. Bmp2 expression is decreased in the FaceSMO face and limb. WHISH of Bmp2 expression in the face (A-D), forelimbs (E-F, H-I), and hindlimbs (G, J) of control (A-B, E-G) and FaceSMO (C-D, H-J) E11.5 (A, C, E, H) and E12.5 (B, D, F-G, I-J) embryos. Red arrows point to regions of decreased expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars- 500 µΜ.

188 strongly expressed in the interdigital mesenchyme in the control forelimbs and hindlimbs at

E12.5 (Figure 63F-G), expression was decreased in both intensity and range of expression

(depth into the mesenchyme) in the E12.5 FaceSMO forelimbs and hindlimbs, particularly in the anterior, where the extra digits were forming (Figure 63I, J). In summary, Bmp2 expression was decreased in both the FaceSMO nasal pits and interdigital mesenchyme, suggesting BMP2 is downregulated by HH signaling.

Unlike Bmp2, Bmp4 expression was increased in the FaceSMO face. In the controls at

E11.5, Bmp4 was mostly expressed in the lateral edges of the FNP (Figure 64A). However, in the FaceSMOs, Bmp4 was strongly expressed in the medial FNP as well (Figure 64B).

This pattern continued into E12.5; there was stronger Bmp4 expression in the nasal prominence of the FaceSMOs than the controls (Figure 64C-D). Bmp4 expression was also upregulated in the limbs. At E11.5, Bmp4 was evenly expressed across the A-P axis in the distal mesenchyme of control forelimbs and hindlimbs (Figure 64E, I). This expression pattern was similar in the FaceSMO limbs, but with stronger Bmp4 expression in the anterior of the limbs (Figure 64F, J). By E12.5, there was very little Bmp4 expression in the control limbs (Figure 64G, K). In the FaceSMO limbs, however, Bmp4 was still expressed in the distal mesenchyme (Figure 64H, L). In summary, Bmp4 expression was increased in the

FaceSMO face and limbs.

Given the dysregulation of BMP expression in the FaceSMO limbs, I next examined the expression of BMP antagonist, Gremlin. Preliminary evidences suggested that while there was very little Gremlin expression in the E11.5 control hindlimb, Gremlin was strongly expressed in the FaceSMO hindlimb, particular in the anterior (Figure 65). This may

189

Figure 64. Increased BMP4 expression in the FaceSMO face and limb.

190

Figure 64. Increased BMP4 expression in the FaceSMO face and limb. WHISH of BMP4 expression in the face (A-D), forelimb (E-H), and hindlimb (I- L) of control (left column) and FaceSMO (right column) E11.5 (A-B, E-F, I-J) and E12.5 (C- D, G-H, K-L) embryos. Red arrows point to regions of increased expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Scale Bars- 500 µΜ.

191

Figure 65. Increased Gremlin expression in the FaceSMO hindlimb. WMISH of Gremlin expression in E11.5 control (A) and FaceSMO (B) hindlimbs (HL). 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Scale Bars= 500µΜ.

represent the reduction in Gli3 caused by increased HH signaling as Gremlin expression is

normally restricted to the posterior limb by GLI3R. GLI3R has also been shown to restrict

Hoxd13 to the posterior and mutations in Hoxd11 and Hoxd13 cause defects in zeugopod and

autopod development [368]. Thus, I next examined the expression of Hoxd11 and Hoxd13 in the FaceSMO limbs.

In the E11.5 controls, Hoxd11 was expressed in the distal mesenchyme, with the

strongest expression in the medial mesenchyme and weakest expression in the anterior and

posterior regions of the limb (Figure 66A). Hoxd11 expression was increased in the E11.5

FaceSMO limbs, particularly in the anterior and posterior regions (Figure 66B). Similarly, at

E11.5, Hoxd13 expression was also increased in the FaceSMO limbs, particularly in the

anterior and posterior (Figure 66D compared to Figure 66C). Interestingly, by E12.5,

Hoxd13 was strongly expressed in the interdigital mesenchyme of the control hindlimb

(Figure 66E), but was greatly reduced in the FaceSMO hindlimb (Figure 66F). In summary,

192

Figure 66. Increased, followed by decreased, expression of Hox genes in FaceSMO limbs. WMISH of Hoxd11 (A-B) and Hoxd13 (C-D, E-F) expression at E11.5 (A-D) and E12.5 (E-F) in control (A, C, E) and FaceSMO (B, D, F) limbs. Red arrows indicate regions of increased expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior) of the limb. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars- 500 µΜ.

193 Hoxd11 and Hoxd13 expression was increased at E11.5, but Hoxd13 expression was greatly decreased by E12.5, demonstrating the importance of timing in the effects of HH signaling regulation.

Finally, WNT signaling is critical to both normal facial and limb development, including osteogenesis in the skull and limb bud outgrowth and dorsoventral patterning.

Thus, I examined WNT expression in the limbs and face by utilizing the Axin2lacZ reporter allele in conjunction with the FaceSMO mice. Axin2lacZ was expressed in the facial prominences at E11.5 in the controls (Figure 67A). In the FaceSMOs, Axin2lacZ expression was increased in the FNP (Figure 67B). By E12.5, Axin2lacZ was strongly expressed in the nasal region of the control and FaceSMO heads (Figure 67C-D). However, in the FaceSMOs

this region of strong expression was expanded laterally (Figure 67D). In the control limbs,

Axin2lacZ was expressed in the anterior mesenchyme and AER at E11.5 (Figure 67E). In the

FaceSMO limbs, Axin2lacZ was also expressed in the AER and anterior mesenchyme (Figure

67F). However, the expression in the anterior mesenchyme was decreased with a more

striated expression pattern extending more distally into the limb. Therefore, Axin2lacZ

expression was decreased in the medial anterior mesenchyme, but increased in the distal

anterior mesenchyme. In summary, WNT signaling patterning and expression levels were

altered in the FaceSMO face and limbs.

Ectodermal Upregulation of HH Signaling Results in Cranial and Limb Defects

Next, I wondered if the phenotypes seen when SmoM2 was expressed in the

mesenchyme would be recapitulated if it was expressed in the ectoderm. Since SmoM2 is a

mutation in the SMO effector, when expressed in the ectoderm, increased HH signaling

should be specific to the ectoderm. Therefore, any effects on tissues of mesenchymal origin

194 Figure 67. Axin2lacZ expression in FaceSMO embryos. Axin2lacZ expression in E11.5 (A-B, E-F) and E12.5 (C-D) control (A, C, E) and FaceSMO (B, D, F) embryos. Red arrow points to region of striated Axin2lacZ expression. 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior). Abbreviations: FL, forelimb; HL, hindlimb. Scale bars= 500µΜ, Α−Β, Ε−F; 1 mm, C-D.

195 must occur via non cell-autonomous mechanisms, including paracrine signaling pathways downstream of HH signaling. Therefore, I hypothesized that if the phenotypes were similar, it would indicate similar paracrine signaling pathways were activated by increased HH signaling from the ectoderm or mesenchyme. If they were different, HH signaling in the ectoderm or mesenchyme may work through different mechanisms.

Crect;SmoM2 (EctSMO)

To investigate the specific role of hedgehog signaling from the ectoderm, I first utilized an ectodermal Cre recombinase transgenic mouse line, Crect, developed by the

Williams lab. The Crect driver recombines in the surface ectoderm, including in the ectoderm of the developing limbs and head, as visualized by R26R β-galactosidase function

[271, 272] (Figure 68). Crect recombinase activity can be detected in the ectoderm and its derivatives as early as E8.5 and is fully expressed in the head by 9.5 [272]. Thus, Crect mediated recombination of SmoM2 leads to constitutive activation of SMO in the ectoderm and its derivatives, beginning at E8.5. As SMO is an effector of HH signaling, the resulting

Figure 68. Crect drives recombination in the surface ectoderm. E11.5 wholemount (A) and E14.5 sectioned (B) Crect;R26R embryos stained for β-gal. Abbreviations: T, tongue; OC, oral cavity. Adapted from Schock & Struve et al, 2017, open access [271].

196 upregulation of HH signaling in the ectoderm will be cell-autonomous. Thus, any defects observed in the ectodermal derivatives could be the direct result of increased HH signaling or the result of the subsequent dysregulation of other signaling pathways downstream of HH in the ectoderm. In contrast, any defects observed in mesodermal derivatives can only be the result of non cell-autonomous defects, including paracrine signaling pathways downstream of

HH.

At E18.5, Crect;SmoM2IX embryos, referred to as EctSMO, had several striking gross morphological defects of the torso, head, and limbs (Figure 69). While all the EctSMO embryos had defects in these regions, the severity of the defects was variable. First, I’ll focus on the torso defects. During early embryogenesis, the ventral body wall must close to envelop the growing internal organs. While in the controls (Figure 69A, C, E) body wall closure occurred normally, about 20% of the EctSMOs had a body wall closure defect, which first became apparent at E13.5 (Figure 69B, D). In the EctSMO embryos with normal body

wall closure, the torso was still abnormal, as it had an atypical shape and swollen appearance

(Figure 69F). EctSMOs with body wall closure defects also frequently had exencephaly

(Figure 69B, D). Overall, about 30% of the EctSMOs had exencephaly, starting at E13.5 and

persisting until birth (Figure 69G, H). The brains of EctSMO mice without exencephaly

appeared fairly normal, though there were indistinct boundaries between brain lobes (Figure

70). While body wall closure defects and exencephaly are both serious birth defects that warrant further study, the rest of this section will focus on the craniofacial and limb defects

IX Experiments using the Crect;SmoM2 (EctSMO) mice were performed by former Williams lab member, Jian Huang, with the exception of the data presented in Figure 69G,H and Figure 71A-D, performed by Linnea Schmidt. 197

Figure 69. EctSMO embryos exhibit torso defects and exencephaly. (A-H): Lateral (A-B) and ventral (C-D) views of control (A, C, E) and EctSMO embryos (B, D, F-H). (A-B): E13.5; (C-D): E14.5; (E-F, H): E18.5, (G): E16.5. Note the body wall closure defects (arrows) and exencephaly (arrowheads). A-F: 1mm, G-H: 2mm.

198

Figure 70. EctSMO brains without exencephaly have indistinct boundaries between sections. Control and EctSMO P0 brain. Black arrows point to regions that have a distinct boundary in controls, but not in EctSMOs. Scale bar= 2mm.

of the majority of the EctSMO embryos—those without body wall closure or exencephaly defects.

One obvious craniofacial defect of the EctSMO embryos was their abnormal skin.

The head, particularly from a lateral view, had a distinctive pattern of coloration, with lighter skin rostrally and darker skin more caudally (Figure 69F). Additionally, the head, as well as the rest of the body, was frequently covered with skin tags and protrusions (Figure 69H and data not shown). Given the abnormal skin appearance in the EctSMO embryos, I next hypothesized that the EctSMO skin may have a barrier defect. To test this hypothesis, I used toluidine blue stain as a marker of skin barrier function. While toluidine blue did not permeate the skin of control (Figure 71A, C) E18.5 embryos, EctSMO embryos (Figure 71B,

D) stained blue throughout the body and permeated particularly well through the limbs and head, indicating a loss of barrier function in those areas. To better understand the differences in skin morphology that contributed to the loss of barrier function, skin sections were labelled by immunofluorescence with markers for the spinous layer (K10), basement

199

Figure 71. EctSMO embryos have skin barrier defects.

200

Figure 71. EctSMO embryos have skin barrier defects. (A-D): E18.5 control (A, C) and EctSMO (B, D) embryos stained with toluidine blue as a marker of skin barrier function. Immunofluorescence of skin sections from E18.5 control (E, G, I) and EctSMO (F, H, J) embryos labelled with K10 (E-F), Laminin (G-H), and K14 (I-J). Note the thicker spinous layer (red line) in F compared to E, the loss of the basement membrane in H compared to G (marked by red arrow), and the loss of hair follicles (marked by red arrowheads) in J compared to I. (K-L): BrdU staining in control (K) and EctSMO (L) E16.5 skin sections. Note the increased black staining in the EctSMOs, indicative of an increase in proliferation. Scale Bars= A-B: 2mm; C-D: 1mm; E-H: 40µM; I- L: 200 uM.

201 membrane (laminin), and basal layer (K14) of the skin [369]. Probing with K10 revealed

that the spinous layer of the skin was thicker at E18.5 in the EctSMOs (Figure 71F) than in their littermate controls (Figure 71E). Additionally, laminin labelling showed that unlike the controls (Figure 71G), the EctSMO skin did not have a basement membrane (Figure 71H).

While the basal layer of the skin was similar between E18.5 controls and EctSMOs (Figure

71I-J), K14 labelling showed that, unlike the controls (Figure 71I), EctSMOs did not have hair follicles (Figure 71J). One potential mechanism that may contribute to these defects is a change in proliferation. To test if proliferation was altered in the EctSMO skin sections,

E16.5 embryos were stained with BrdU. Proliferation increased in the EctSMO skin (Figure

71L) compared to the control skin (Figure 71K). In summary, the skin of EctSMO embryos had decreased barrier function, likely due to the loss of the basement membrane.

Additionally, the skin did not develop hair follicles and had a thicker spinous layer, possibly due to increased proliferation at early stages.

In addition to the cranial skin defect, EctSMO embryos also had patterning and skeletal defects. In the face, EctSMO mice had eye defects, hypertelorism, and cleft

lip/palate (Figure 72E-G compared to Figure 72A-C). Early cartilage staining at E14.5

showed that the EctSMO heads had decreased paranasal cartilage (Figure 72H). By P0 even

more cranial skeletal defects were apparent. The shape of several cranial bones was altered

in the EctSMOs. First, the supraoccipital had a kink not present in the controls (Figure 72I-J,

L-M). Second, the apex of the nasal bones was narrower in the EctSMOs than the controls

(Figure 72I-J, L-M). Finally, the bones of the cranial vault were slightly rounder than in the

controls (Figure 72K, N). There were also defects in bone formation in the EctSMOs. In

202

Figure 72. EctSMO mice exhibit multiple craniofacial defects. (A-C, E-G): Gross morphology of P0 control (A-C) and EctSMO (E-G) craniofacial features, including the eye (A, E), snout (B, F), and palate (C, G). Black arrow indicates eye defect, white arrowhead indicates cleft lip, white arrow indicates secondary cleft palate, and blue arrow indicates primary cleft palate. (D-H): Cartilage stain of E14.5 control (D) and EctSMO (H) snouts, ventral view. Black arrowhead indicates loss of paranasal cartilage. (I-N): Bone stain of P0 control (I-K) and EctSMO (L-N) skulls, viewed dorsally (I, L), ventrally (J, M), and laterally (K, N). Ventral view shown without jaw. Blue stars mark changes in bone shape, yellow stars mark loss of bone/incisor, and the green star marks bone loss found together with ectopic bone formation. Abbreviations: In, incisor; Ns, nasal bone; PPMX, palatal process of the maxilla; SO, supraoccipital; TR, tympanic ring. Scale bars=A, C-E, G-N: 1 mm; B, F: 2 mm.

203 terms of lost bones, the tympanic ring as well as the maxillary and mandibular incisors were absent in the EctSMO skulls (Figure 72M, N and Figure 73). In contrast, there was ectopic bone formation in the palatal process of the maxilla (Figure 72M). In summary, EctSMO heads had defects in both ectodermal derived structures, such as the skin and eyes, as well as mesenchyme derived structures, such as the cranial bone.

The EctSMOs also had limb defects. Similar to the FaceSMO limbs, the EctSMO forelimbs and hindlimbs were shorter than in the controls (Figure 74A-B). While digit patterning appeared normal in the EctSMO forelimbs (Figure 74A), 80% of the EctSMO hindlimbs had polydactyly, with either six or seven digits (Figure 74B, C, E). The extra

Figure 73. Incisors are missing in EctSMO mice. (A-B): Lateral view of E18.5 control (A) and EctSMO (B) mandibles. (C-D): Ventral view of E18.5 control (C) and EctSMO (D) maxillas. Black arrows indicate normal incisors, note their absence in B, D. Scale Bars= (A-B): 1 mm; (C-D): 500 µM.

204

Figure 74. EctSMO mice have limb defects. (A-B): Bone and cartilage staining of E18.5 forelimbs (A) and hindlimbs (B). Black arrow indicates decreased formation of the zeugopod bones. (C): Gross morphology of E18.5 EctSMO hindlimb. (D-E): Cartilage staining of E16.5 control (D) and EctSMO (E) hindlimbs. Green arrows indicate extra digits/tarsal; red arrows indicate extra distal phalange. 1 indicates the 1st digit (anterior), 5/6 indicates the 5th//6th digit (posterior) of the limb. Scale Bars= 1mm.

digits appeared to be added to the anterior side of the limb. Additionally, extra distal phalanges were present in some of the EctSMO digits (Figure 74E). Finally, the zeugopod

205 bones of the hindlimb (tibia and fibula) were severely underdeveloped in the EctSMO mice; this loss was the primary source of the overall shortening of the hindlimb (Figure 74B). In summary, EctSMO limbs were shorter than controls; in addition, the hindlimbs had underdeveloped zeugopod bones and digit patterning defects.

Having established the gross and skeletal phenotypes of the EctSMO mice, I next investigated the molecular underpinnings that could be contributing to the phenotypes, with particular emphasis on the hindlimb as EctSMO and FaceSMO hindlimbs shared several

defects, including polydactyly and shortened limbs. The first signaling pathway that I

examined was HH signaling. Given the expression pattern of Crect and properties of

SmoM2, HH signaling should be upregulated in the ectoderm, but not the mesenchyme. To

visualize HH signaling in both controls and EctSMO mice, I bred PtcLacZ into the EctSMO

background as Patched is a target gene of HH signaling, and thus serves as a readout of HH signaling activity. Beginning at E9.5, PtcLacZ expression in the EctSMO hindlimb was upregulated in the ectoderm, but not the mesenchyme (Figure 75A, B). This same expression

pattern, with increased ectodermal HH signaling, but no change seen in the mesenchyme,

persisted through E14.5, though at E13.5 and E14.5 PtcLacZ expression was visible in the

developing skin, which is derived from the ectoderm (Figure 75).

Given the polydactyly in the EctSMO hindlimbs, I next examined whether expression

of Gli3 or Hand2 was altered, as they were in the FaceSMO limbs. In the control E11.5

hindlimb, Gli3 was expressed throughout the distal mesenchyme (Figure 76A). Gli3 in the

EctSMO hindlimb had a similar expression pattern except with slightly decreased expression,

particularly in the anterior limb (Figure 76B). In contrast, Hand2 expression in the controls

206

Figure 75. Hedgehog signaling is increased in the EctSMO ectoderm, but not mesenchyme. PtcLacZ expression, visualized by β−galactosidase staining, in E9.5 (A-B), E10.5 (C-D), E11.5 (E-F), E12.5 (G-H), E13.5 (I-J), and E14.5 (K-L) control (1st and 3rd columns) and EctSMO (2nd and 4th columns) embryos. Red arrows indicate increased PtcLacZ expression. Note that the increased β−galactosidase staining outside of the AER in the E13.5 and E14.5 embryos most likely represents expression in the skin, not underlying mesenchyme. Scale Bars=500 µΜ.

207

Figure 76. Expression pattern of Gli3 and Hand2 is altered in the EctSMO E11.5 hindlimbs. (A-B): Gli3 expression in the control (A) and EctSMO (B) E11.5 hindlimb. (C- D): Hand2 expression in the control (C) and EctSMO (D) E11.5 limbs. Black arrow indicates region of decreased expression, red arrow indicates region of increased expression. 1 and 5 denote the anterior (1) and posterior (5) of the limb bud. Abbreviations: FL, forelimb; HL hindlimb. Scale Bars= 500 µM.

was restricted to the posterior mesenchyme (Figure 76C). In the EctSMO hindlimbs,

however, Hand2 was expressed in the posterior, but also extended into the anterior

mesenchyme (Figure 76D). In summary, Gli3 expression was decreased in the anterior and

Hand2 expression was increased in the anterior of EctSMO E11.5 hindlimbs. These expression changes are similar to those observed in the FaceSMO limbs, though the decrease in Gli3 was more dramatic in the FaceSMOs.

208 Given that ectodermal expression of SmoM2 was impacting patterning of

mesenchymal signals, I next examined the expression of several FGF ligands expressed in

the AER that are known to signal into the mesenchyme to determine if increased HH signaling in the ectoderm was altering FGF levels. If so, the altered ectodermal signals may

be causing an impaired epithelial-mesenchymal feedback loop that could contribute to the

hindlimb defects seen in the EctSMO mice. In the control E11.5 hindlimbs, Fgf4 was

expressed in the medial AER (Figure 77A). In the EctSMO hindlimbs, Fgf4 was also

expressed in the AER, but with a wider range of expression, extending toward both the

posterior and anterior sides of the limb (Figure 77B). Fgf9 was also expressed in the AER of

the control hindlimbs (Figure 77C). However, in the EctSMO hindlimbs, Fg9 expression in the AER was lost (Figure 77D). In contrast to the severely changed expression patterns seen

with Fgf4 and Fgf9, Fgf8 expression was similar in the control and EctSMO hindlimbs with

only a small increase in Fgf8 in the anterior ectoderm (Figure 77E-F). Thus, expression

changes in Fgf4, Fgf8, and Fgf9 in the EctSMO AER may be contributing to the distal

outgrowth defect. Additionally, these changes may contribute to signaling differences in the

mesenchyme that lead to the other EctSMO limb phenotypes, like polydactyly. Alternatively,

the expansion of Fgf4 and Fgf8 in the AER, may contribute to a wider footplate that causes

polydactyly.

Given the role of Bmp expression in both A-P patterning as limb bone shape, as well

as the differences seen in the Bmp expression in the FaceSMO limbs, I next examined the

expression of Bmp2 and Bmp4 in the control and EctSMO limbs. In the control forelimb,

Bmp2 was expressed in the distal mesenchyme as well as the anterior AER (Figure 78A).

Comparatively, in the EctSMO forelimb, Bmp2 expression was increased in the distal

209

Figure 77. FGF ligand expression in EctSMO E11.5 hindlimbs. (A-B): Fgf4 expression in control (A) and EctSMO (B) E11.5 hindlimbs. (C- D): Fgf9 expression in control (C) and EctSMO (D) E11.5 hindlimbs. (E-F): Fgf8 expression in control (E) and EctSMO (F) hindlimbs. Black arrow indicates region of decreased expression, red arrow indicates region of increased expression. 1 and 5 denote the anterior (1) and posterior (5) of the limb bud. All images are shown in the same orientation, with the anterior at the top. Scale Bars= 500 µM.

210

Figure 78. Bmp expression is increased in E11.5 EctSMO limbs. WMISH of Bmp2 (A-D) and Bmp4 (E-F) in E11.5 control (A, C, E) and EctSMO (B, D, F) forelimbs (A-B) and hindlimbs (E-F). Red arrows indicate regions of increased expression in the ectoderm; blue arrows indicate regions of increased expression in the mesenchyme. Limbs are all shown in the same orientation- 1 indicates the 1st digit, 5 indicates the 5th digit. Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars- 500 µΜ.

211 mesenchyme, but decreased in the anterior AER (Figure 78B). In the control hindlimbs,

Bmp2 was expressed in the AER, but not the distal mesenchyme (Figure 78C). However, in

the EctSMO hindlimbs, Bmp2 expression was lost in the AER and increased in the distal

posterior mesenchyme (Figure 78D). The similarity between the changes in Bmp2

expression in the AER and mesenchyme (lost in the AER and gained in the mesenchyme in

the EctSMOs) in the forelimbs and hindlimbs, suggests that these expression differences

likely result in a phenotype shared by the EctSMO forelimb and hindlimb – for example, the

loss of distal outgrowth, but not the polydactyly phenotype only observed in the hindlimb.

Bmp4 expression was also altered in the EctSMO hindlimbs. In the control hindlimbs, Bmp4

was exclusively expressed in the distal mesenchyme; however, in the EctSMO hindlimbs,

Bmp4 was expressed in both the distal mesenchyme and the AER (Figure 78E, F). Thus,

altered expression of both Bmp2 and Bmp4 expression in the AER, or increased expression of

Bmp2 in the mesenchyme may also be how ectodermal SmoM2 expression exerts its affects.

Interestingly, the direction of the expression change of the Bmps in the AER in the EctSMOs mirrored the expression change of the Bmps in the mesenchyme in the FaceSMOs. As such,

Bmp2 expression was decreased in the AER of EctSMOs and mesenchyme of FaceSMOs;

Bmp4 expression was increased in the AER of EctSMOs and mesenchyme of FaceSMOs.

Finally, given the alterations in expression of the Hox genes and WNT signaling in the FaceSMO limbs, I also examined the expression of Hoxd11 and Hoxd13 in the EctSMOs.

In the E11.5 control hindlimb, both Hoxd11 and Hoxd13 were expressed in the distal mesenchyme, with more expression in the posterior than anterior sides (Figure 79A, C).

EctSMO hindlimbs had a similar expression pattern, but with greater Hoxd11 and Hoxd13 expression on the anterior side of the limb (Figure 79B, D). This altered expression pattern –

212

Figure 79. Increased expression of Hoxd11 and Hoxd13 in the anterior of EctSMO hindlimbs. WMISH of Hoxd11 and Hoxd13 in control (A, C) and EctSMO (B, D) limbs. Red arrows indicate regions of increased expression. Limbs are shown in the same orientation- 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior). Abbreviations: FL, forelimb; HL, hindlimb. Scale Bars- 500 µΜ.

increased anterior expression of Hoxd11 and Hoxd13 – is also identical to the altered expression seen in the FaceSMO E11.5 limbs, suggesting a similar mechanism regardless of whether SmoM2 is expressed in the ectoderm or mesenchyme.

As with the Hox genes, WNT signaling was also altered in the EctSMO hindlimbs.

β−catenin-activated transgene (BAT) was bred into the EctSMO line so that WNT signaling

213 could be visualized in the hindlimbs through β−galactosidase staining. At E11.5 there was

WNT signaling throughout the control hindlimbs; a similar pattern was seen in the EctSMO hindlimbs, but with additional WNT signaling in the AER (Figure 80A-B). By E12.5, WNT signaling in the control and EctSMO hindlimbs was somewhat diminished; however, in the

EctSMO hindlimbs, there was still additional WNT signaling in the posterior and anterior regions of the AER (Figure 80C-D). In summary, WNT signaling was increased in the

Figure 80. WNT signaling is upregulated in the EctSMO ectoderm. Expression of β−catenin-activated transgene (BAT) visualized by β−galactosidase staining at E11.5 (A-B) and E12.5 (C-D) control (A, C) and EctSMO (B, D) hindlimbs. Red arrows indicate regions of increased expression. Limbs are shown in the same orientation- 1 indicates the 1st digit (anterior), 5 indicates the 5th digit (posterior). Scale Bars- 500 µΜ

214 E11.5 and E12.5 EctSMO hindlimb ectoderm. However, it is difficult to accurately compare

WNT signaling between the FaceSMO and EctSMO hindlimbs because of differences

between the Axin2lacZ and BAT-GAL reporter. In the control E11.5 limbs, Axin2lacZ reporter showed WNT signaling in the anterior mesenchyme and ectoderm, but the BAT-GAL reporter showed WNT signaling distributed more evenly across the limb mesenchyme and not in the ectoderm. WNT signaling was increased in the anterior mesenchyme in the

FaceSMO limbs, but increased in the ectoderm of the EctSMO hindlimbs; however, this difference may not reflect actual variation between the FaceSMO and EctSMO limbs, but rather the disparity between two different reporters. Thus, for a better comparison of WNT signaling between the FaceSMO and EctSMO limbs, it would be necessary to use the same reporter. X

Msx2-Cre;SmoM2

To determine the effect of timing on the EctSMO phenotypes, I also examined used another ectodermal Cre that is later expressed, Msx2-Cre. Like CRECT, Msx2-Cre also exhibits ectodermal expression (Figure 16), but that expression begins at E10.5, rather than the E8.5 as seen in Crect. Therefore, I next examined if Msx2-Cre;SmoM2 mice; referred to as MSMOs, would also exhibit patterning and skeletal defects. MSMO P0 pups were born at normal size (Figure 81) and in normal Mendelian ratios, but did not survive beyond the first day. MSMOs did not have cranial or limb patterning defects as in the EctSMOs, but were identifiable from their control littermates by aberrant pigmentation. While the controls were generally pink all over (Figure 81A), MSMOs exhibited black pigmentation on the rostral

X Different reporters were used due to the timing of the experiments. Earlier studies by former Williams lab member Jian Huang, prior to my arrival in the lab, were done using BAT-GAL. I used the Axin2lacZ reporter for later studies, at which time the available reporter was Axin2lacZ reporter. 215

Figure 81. MSMO neonates have pigment variation and skin barrier defects.

216

Figure 81. MSMO neonates have pigment variation and skin barrier defects. A-C: P0 pups. Note the darker pigmentation and wrinkly skin in the MSMO mice. D-I: P0 pups stained with toluidine blue as a marker for skin barrier function. Note areas of toluidine blue staining on back are similar to areas of darker pigmentation in wholemount pups.

217 region of the head and along the midline of the back, with some darker pigmentation radiating outwards from the midline (Figure 81B). Additionally, the MSMO skin, particularly on the head, was more wrinkled than in the controls (Figure 81C). Given the early mortality and no other gross morphological defects, I next stained the P0 pups with toluidine blue as a marker for skin barrier function. While toluidine blue did not permeate through the controls skin (Figure 81D, G), toluidine blue did permeate through the MSMO skin at the dorsal and ventral midline (Figure 81D, right and Figure 81E, respectively), paws

(Figure 81E), eyes (Figure 81F, H, I), and parts of the head and face (Figure 81F, H, I).

This distinctive pattern of permeability is likely reflective of Msx2-Cre’s expression pattern

(Figure 16, Figure 19). The defect in barrier function in the MSMOs likely leads to the

MSMOs’ desiccation which contributes to the wrinkled skin phenotype. This desiccation may cause or be a contributing factor in the pups early mortality. Thus, both EctSMO and MSMO mice had skin barrier defects, but MSMO mice did not have cranial or limb patterning defects.

Though there were not obvious patterning defects in the MSMO face or limbs, given the skeletal defects seen in the EctSMOs, I next examined if the MSMOs had similar underlying skeletal defects. However, there was no difference in the skeleton of MSMOs compared to controls (Figure 82). Based on Msx2-Cre’s expression pattern (Figure 16), any skeletal defects would likely occur in the head and limbs, but no defects were observed in the cranial vault (Figure 82A), cranial base (Figure 82B), jaw (Figure 82D), forelimbs (Figure

82E, F) or hindlimbs (Figure 82G). Thus, if SmoM2 expression does not begin until E10.5, increased HH signaling in the ectoderm does not lead to patterning or skeletal defects.

218

Figure 82. MSMO neonates exhibit no skeletal defects. A-B, D-F: Alizarin red stained bone of A) Cranial vault, B) Cranial base (mandible removed), D) jaw, E) Right forelimb, F) Right paw, and G) Right hindlimb in P0 pups. C: Alcian blue stained cartilage in E16.5 embryos. Abbreviations (in purple): A, anterior; P, posterior.

Discussion

In this chapter, I described the consequences of specific mesenchymal or ectodermal upregulation of HH signaling in the mouse face and limb. In addition to the tissue layer specificity, these studies are novel in that they utilize SmoM2, a constitutively active form of

SMO. This allows for the delineation of the effects of ectopic or increased HH ligand expression and overall increased expression of the pathway. The origin of increased HH signaling (ectoderm vs mesenchyme) in the head and limb was important for some phenotypes, but not for others. In general, the origin affected ossification, but not patterning.

219 In the following I will address the consequences of increased HH signaling in the head and

limb separately, starting with head.

Both the FaceSMO and EctSMO heads displayed mediolateral widening of the FNP

and widening between the eyes, together known as hypertelorism. This phenotype was not unexpected as excess SHH in the mesenchyme causes hypertelorism in chick [203] as does ectopic SHH in the mouse ectoderm [370]. My studies thus confirm that increased HH signaling causes hypertelorism regardless of ectodermal or mesenchymal facial expression.

Additionally, this hypertelorism is a result of activation of the pathway, not simply the SHH or IHH ligand, and thus other signaling molecules activated by HH signaling likely play a role in hypertelorism development. Later in this section, I will discuss some putative targets.

In contrast, HH mediated cleft lip/palate appears to be caused by the specific activation of HH signaling in the ectoderm. Given that cleft lip/palate is a common birth defect and that ectodermal signaling centers are important for craniofacial patterning,

EctSMO mice could serve as a great model for orofacial clefting as HH is increased from the natural domain of expression. On the other hand, formation of a shortened snout and the loss of nasal and frontal bone ossification appears to be dependent upon mesenchymal specific

HH signaling, though the paranasal cartilage is shorter in the EctSMO mice at E14.5.

Additionally, the forebrain protrudes out from the hole in the skull in FaceSMO mice. This is notable as a rare congenital malformation, Frontonasal dysplasia (FND), is characterized by hypertelorism, a broad nasal bridge, and in severe cases, encephalocele (an opening of the skull with protrusion of the brain) (Figure 83). While it has been suggested that HH signaling may play a role in FND, the etiology of FND is unknown though multiple cases

220

Figure 83. Facial dysmorphology of a child with frontonasal dysplasia. Used with permission from Sharma et. al, 2012 [371].

were reported within families, suggesting a genetic inheritance [372]. Given the phenotypic facial similarities between FaceSMO mice and FND patients, FaceSMO mice could serve as a model for FND. Though an enhancer region like the ZRS in limb for HH signaling has not yet been identified for the face, I speculate that FND patients may have mutations in a ZRS like regulatory region. While their function is unknown, certain chromosomal regions, most of which are on different chromosomes than Ihh or Shh, are associated with FND [372]. I propose that these regions should be examined for their potential ability to serve as facial HH signaling regulatory regions. Finally, solely the EctSMO mice had a collection of other

221 defects in both ectodermal and mesenchymal derived structures. Defects in ectodermal- derived structures included ocular and dermal defects. Defects in mesenchymal derived structures included shape alterations in the nasal bones, cranial vault, and supraoccipital as well as loss of the tympanic rings and incisors, but ectopic bone formation in the palatal process of the maxilla. One explanation for the additional phenotypes observed in the

EctSMO is that Crect has a more general expression pattern than Creface, which in the head is only expressed in the FNP. As such, some of these defects are outside of the range of the

FNP (i.e. tympanic rings, supraoccipital) and so would not be expected in the FaceSMO mice. However, others (i.e. ectopic bones in the palatal process of the maxilla) may reflect a true requirement for HH signaling from the ectoderm. In summary, in the face, the

development of hypertelorism was not dependent on the specific tissue layer activation of

HH signaling, but the majority of the defects varied depending on ectodermal or mesenchymal expression of SmoM2.

The role of HH signaling in cranial suture and bone morphogenesis is a relatively new area of study, yet one of importance. While it is clear that HH signaling is an important regulator of both osteogenesis and cranial suture biology, studies examining the role of HH signaling in cranial bone ossification and cranial suture morphogenesis have thus far produced conflicting results [197]. For example, Ihh is expressed in mice during calvarial osteoblast development and during osteoblast proliferation in the osteogenic fronts [210], as well as in the sagittal suture [209]. If IHH is disrupted osteoblastogenesis is impaired, which leads to a reduction in cranial bone size [209] . Increased HH signaling has been shown to lead to craniosynostosis both mice and humans [213–215], suggesting that HH signaling promotes calvarial ossification and suture closure. Specifically, duplications of Ihh in mice

222 and humans result in craniosynostosis [213]. Additionally, mutations in Rab23 and Ptch1, which lead to increased HH signaling, also cause craniosynostosis [214, 215]. However,

IHH has also been shown to repress osteogenic lineage differentiation, with loss of IHH resulting in premature osteoprogenitor cell differentiation [216]. Ossification in the midline of the nasal and frontal bones was impaired in FaceSMOs, resulting in wider frontonasal and

interfrontal sutures which suggests that HH signaling represses, rather than promotes,

osteogenesis. In contrast, the EctSMO heads had both ectopic and lost bone formation.

Together, this suggests that the role of HH signaling in ossification may be complicated by both the origin of expression (ectodermal vs mesenchymal) as well as the specific location/bone within the skull that is receiving the signal. Interestingly, while disrupted IHH signaling impairs osteoblastogenesis likely via downregulation of Bmp expression [209], in situ data in the FaceSMO heads suggested that Bmp4 expression was increased. This could

imply that decreased HH signaling leads to decreased BMP signaling whereas increased HH

signaling promotes BMP signaling. However, Bmp2 was downregulated in the FaceSMO

nasal pits; therefore, there was differential regulation of BMP ligands in the head. Given that

BMP signaling promotes bone formation [373] and that Bmp2 expression was only altered in

the nasal pits, it seems likely that the altered expression of Bmp2 and Bmp4 are responsible

for the patterning, rather than the ossification, defects in the FaceSMOs, at least at E11.5.

Additionally, WNT signaling was increased in the FaceSMO prominences. Previous work

has shown WNT signaling promotes osteogenesis [374]. Given the decreased ossification in

the FaceSMOs, it seems likely that the upregulation of WNT in the FaceSMO prominences

was responsible for the patterning defects of the head rather than the ossification defects.

Therefore, further research controlling for the timing, location, and tissue of HH expression

223 is required to elucidate the role of HH signaling in ossification and potential mechanisms of action.

As in the face, EctSMO and FaceSMO limbs had both distinct and shared phenotypes as well as potential signals driving those phenotypes. Both the FaceSMO and EctSMO limbs had defects in distal outgrowth and syndactyly; additionally, the FaceSMO limbs as well as the EctSMO hindlimb had polydactyly. The difference in polydactyly presentation in the

EctSMO forelimb and hindlimb may be due to timing of Crect expression. As the forelimb begins development prior to the hindlimb, it may be that by the time Crect expression begins only the hindlimbs are receptive to A-P patterning regulation by SmoM2 in the ectoderm.

This hypothesis is supported by the absence of any limb defects in the MSMO mice, where

Cre expression begins later, at E10.5, suggesting the ability of SmoM2 in the ectoderm to affect limb development diminishes over time. Alternatively, the lack of limb defects may reflect Msx2-Cre’s expression in the AER, rather than the entire ectoderm. In contrast,

Creface expression also begins at E10.5 and the FaceSMO limbs have both A-P and P-D defects. Thus, to affect limb development, HH signaling must be increased earlier in the ectoderm than in the mesenchyme.

There were also differences in limb development between the two models. The

EctSMO hindlimb and FaceSMO limbs both had ectopic bone formation, but the location was variable between the two models. While the EctSMO hindlimbs had extra distal phalanges, the FaceSMO limbs had ectopic bone at both the joints and in the distal tips of the digits.

Interestingly, when SMO is lost in the AER using Msx2-Cre, the digits have ectopic cartilaginous condensations [363], suggesting that both increased and decreased HH signaling leads to ectopic bone/cartilage. Additionally, while both the EctSMO hindlimb and

224 FaceSMO limbs had severe defects in the zeugopod bones, the defects were on polar ends of

the spectrum. The zeugopod bones in the EctSMO hindlimbs were severely underdeveloped, contributing to the overall shortness of the limb. In contrast, the zeugopod bones in the

FaceSMO limbs were thicker and the individual bones were either too close together

(hindlimb) or too far apart (forelimb). As the EctSMO and FaceSMO limbs had both similar and distinct defects, I hypothesized that ectodermal and mesenchymal activation of HH signaling would lead to similar alterations in certain downstream signals, but not others. If a signal was similarly dysregulated between the two models, that signal may be responsible for the shared defects between the models; in contrast, dysregulated signals specific to either

FaceSMO or EctSMO may be responsible for the distinct defects.

Figure 84 shows a graphical summary of the expression patterns of dysregulated genes in the FaceSMO and EctSMO E11.5 hindlimbs as well as postulated connections between those genes. The expression analysis yields several interesting potential mechanisms for the phenotypes observed in the FaceSMO and EctSMO limbs. First, the targeted

expression of SmoM2 in the ectoderm and mesenchyme led to specific upregulation of HH signaling in the mesenchyme of FaceSMO and ectoderm of EctSMO limbs as shown by

PtchLacZ expression, as well as Gli1 expression in the FaceSMOs. This is striking as it

demonstrates that 1) Increased HH signaling in the ectoderm can induce A-P limb patterning

defects in the absence of increased HH signaling in the mesenchyme and 2) Given that the

SmoM2 allele increases HH signaling, rather than the SHH morphogen as in previous studies,

the formation of extra digits is not dependent on the presence of SHH morphogen, but rather

the resulting increase in HH signaling. Alternatively, Shh and/or Ihh may be direct or

indirect targets of increased HH signaling. This explanation is not favored as Ptch was only

225

Figure 84. Model of FaceSMO and EctSMO signaling in the E11.5 hindlimb. Summary of dysregulated genes in the FaceSMO (left) and EctSMO (right) E11.5 hindlimbs. Red indicates upregulated genes, green indicates downregulated genes. The pattern of mis-expression for each gene on the A-P axis is represented; genes that are dysregulated in the anterior are shown near the top, genes that are dysregulated in the posterior are shown near the bottom, and genes that are dysregulated medially or throughout the tissue are shown near the middle. Arrows show postulated relations between the genes. Circles surround genes normally restricted to the posterior by Gli3R. The unfilled arrow from Gli3 to the circled genes represents the likely activation of these genes in the anterior due to loss of restriction by Gli3R.

upregulated in the ectoderm of the EctSMO hindlimbs -- if Shh or Ihh expression was

increased in this model, they would likely diffuse into the mesenchyme and upregulate Ptch expression there as well. However, at this time I cannot exclude this possibility; future work should include examination of the pattern of Ihh and Shh expression in these models.

During normal limb development, prior to HH signaling [366], antagonism between

HAND2 and GLI3R prepatterns the limb bud so that the eventual expression of Shh is restricted to the posterior mesenchyme. In the presence of HH signaling, GLI3R is converted into GLI3A and thus GLI3A predominates in the posterior whereas GLI3R predominates in

226 the anterior limb bud (Figure 85A). Interestingly, both Hand2 and Gli3 expression were

regulated by HH signaling in the FaceSMO and EctSMO hindlimbs. Specifically, increased

HH signaling in both models led to upregulation of Hand2 expression in the anterior mesenchyme whereas it was restricted to the posterior in the controls. While I did not examine the balance between GLI3R and GLI3A, based on previous studies [227] I’d strongly predict that the increased HH signaling in the FaceSMO mesenchyme led to conversion of GLI3R into Gli3A, resulting in very little GLI3R remaining (Figure 85B).

Additionally, total Gli3 expression was diminished in both the FaceSMO and EctSMO

Figure 85. Model of A-P patterning in the FaceSMO limb. In the controls, Shh expression is restricted to the posterior. This causes a gradient in the forms of Gli3 with more Gli3A in the anterior and more Gli3 R in the posterior. Gli3R restricts Hand2 to the anterior. In the FaceSMO limbs, HH signaling is upregulated throughout the limb mesenchyme. Thus, Gli3A prevails, rather than Gli3R. Since Gli3R is diminished, Hand2 is no longer restricted to the posterior limb bud, but rather is expressed throughout the limb mesenchyme.

227 models, which would further limit the amount of GLI3R available. As HH signaling was

increased in the ectoderm of EctSMO limbs, I’d expect the balance of GLI3A and GLI3R to

be altered in the EctSMO ectoderm, rather than the mesenchyme. While this may be the

case, ectodermal HH signaling was still able to regulate Gli3 expression in the mesenchyme

(and thus there was likely less GLI3R), though it is unknown at this time how this regulation

occurs (Figure 84).

Thus, while Shh expression is regulated by Gli3R and HAND2, HH signaling can

also regulate the expression of Hand2 and Gli3. Therefore, SHH, GLI3, and HAND2 may be

part of a feedback loop, rather than simple one-way regulation from HAND2 and GLI3 to

SHH. In addition to Hand2, two additional genes normally restricted to the posterior by

GLI3R, Hoxd13 and GremlinXI, were also upregulated in the anterior in FaceSMO and

EctSMO hindlimbs, furthering the hypothesis that GLI3R is diminished in the mutants.

Future work should determine if the predicted decrease in GLI3R in both EctSMO and

FaceSMO hindlimb reflects reality.

As Hoxd13 was upregulated in both the FaceSMO and EctSMO anterior mesenchyme, the expansion of Hoxd13 may be responsible for the shared defects between the models, particularly polydactyly. Indeed, expression of Hoxd13 throughout the mouse limb bud, via inversion of the mouse Hoxd cluster, led to polydactyly and an ectopic Shh domain [375], suggesting that regulation of HH and Hoxd13 is bidirectional. Additionally, mutations in Hoxd13 are associated with digit abnormalities, including syndactyly,

XI Gremlin was increased in the FaceSMO anterior mesenchyme; preliminary data shows Gremlin was also increased in the EctSMO anterior mesenchyme (EctSMO data not shown). 228 polydactyly, and brachydactyly. Thus, HH mediated increase in Hoxd13 may be responsible for those phenotypes in the FaceSMO and EctSMO limbs [376, 377].

Hoxd11 was also upregulated in both the FaceSMO and EctSMO anterior

mesenchyme. Mutation in Hoxd11 are also known to induce polydactyly [368] and thus may be another signal responsible for the polydactyly in the FaceSMO and EctSMOs.

Additionally, mutations in Hoxd11 are known to affect the formation of the zeugopod [378].

As mentioned previously, the zeugopod was severely affected in both the EctSMO and

FaceSMO hindlimbs, but with different phenotypes. Thus, differential signals between the

two models are likely responsible for the phenotypes. There was differential expression of

two gene families, BMP and FGF, between the FaceSMO and EctSMO hindlimbs.

Bmp4 was upregulated in the FaceSMO anterior mesenchyme, but not ectoderm. In

contrast, expression of the Bmps was altered in the ectoderm, but not mesenchyme of the

EctSMOs, where in the ectoderm Bmp4 was upregulated and Bmp2 was downregulated. As

loss of Bmp2 and Bmp4 in the limbs has previously been shown to result in syndactyly due to

a lack of interdigital cell death, the decreased Bmp2 expression in the FaceSMO and EctSMO

limbs may be contributing to the syndactyly phenotype [259, 265]. Upregulation of Bmp2

and Bmp4 in the chick, via retroviral vectors, led to a dramatic increase in the volume of the

cartilage elements, altered their shapes, and led to joint fusions [267]. Therefore,

upregulation of Bmp4 in the FaceSMO mesenchyme may be leading to the thicker zeugopod

and ectopic bones in the joints. It is not yet clear how increased HH signaling differentially

regulates Bmp2 and Bmp4, but based on the phenotypes, it seems likely that Bmp2 has a more

prominent role in interdigital cell death and Bmp4 has a more prominent role in bone

patterning. In the chick, upregulation of Hoxd13 led to upregulation of Bmp4 in the

229 mesenchyme and ectoderm [379]. Thus, Hoxd13 upregulation may be responsible for Bmp4 upregulation, with other genes modifying the location of expression in the EctSMO and

FaceSMO limbs.

While the effect of increased HH signaling on FGF signaling in the FaceSMO limbs

was not examined, increased HH signaling in the limb ectoderm led to increased Fgf4 and

Fgf8 expression, but decreased Fgf9 expression. As AER-FGFs contribute to outgrowth, these alterations likely account for the shortening of the limb in the EctSMOs and may be

particularly critical for zeugopod formation. Interestingly, loss of Smo in the mouse

ectoderm leads to longer AERs, and in chick, increased SHH (beads) leads to shorter AERs.

Thus, it has been postulated that loss of ectodermal HH signaling leads to longer AERs and

gain of ectodermal hedgehog signaling leads to shorter AERs [363]. However, the range of

Fgf8 in the EctSMO AER was not restricted, suggesting that increased HH signaling in the

mouse ectoderm does not produce shorter AERs in mice. In fact, ectodermal Fgf8 expression

was slightly expanded and Fgf4 expression was greatly expanded, suggesting that the AER

may be longer than the controls, not shorter. If so, this may suggest a biphasic response to

HH signaling in the AER, where both too much and too little signaling leads to longer AERs.

Overall, specific mesenchymal or ectodermal upregulation of HH signaling in the

mouse face and limb led to both shared and distinct craniofacial and limb defects. Most,

strikingly, upregulation of HH in both the limb ectoderm and mesenchyme led to

polydactyly, suggesting that HH signaling can induce formation of extra digits independent

of ectopic Shh expression in the anterior mesenchyme. Several signaling pathways were

altered in both the EctSMO and FaceSMO limbs. While further studies are necessary to

determine the exact function of these genes in promoting the HH mediated phenotypes, my

230 analysis identifies possible candidates. Additionally, these data suggest that the SHH morphogen gradient, partially responsible for assigning A-P limb patterning during normal development, may act by regulating other downstream genes and pathways. Finally, regulation of Shh expression in the mesoderm of appendages has been identified as a possible mechanism for evolution [380]. Given the phenotypic effects of increased HH signaling in the ectoderm, it is possible that response to HH signaling within the ectoderm could also function as a source of evolutionary change. Because HH signaling in the ectoderm alters autopod patterning, it is possible that the sensitivity to HH signaling in the ectoderm alters

digit number during evolution.

231 CHAPTER VII

OUTLOOK

My research has focused on congenital defects using the mouse as a model system.

As both FGF and HH signaling are associated with craniosynostosis, I investigated the specific role of FGF8 and HH signaling in ectodermal and mesenchymal tissue during growth and patterning of the head. Additionally, as craniosynostosis often presents as part of as syndrome that includes limb defects, I also investigated the specific role of FGF8 and HH signaling in ectodermal and mesenchymal tissue during growth and patterning of the limb.

One of the big biological questions is why we have two different ways to form bone – intramembranous and endochondral ossification – and what regulates that decision at both the molecular and evolutionary level. As the cranial vault and limb form through two different types of ossification, examining the role of FGF8 and HH in these locations allowed me to begin to answer that question. Alteration of the FGF8 and HH signaling pathways led to several unexpected results. As the conclusions are quite different from each other, I will mainly address them individually below, combining conclusions from both studies where appropriate.

Craniosynostosis as a Bone Growth Disorder

In Chapter 4 and 5, I investigated how Fgf8 dosage impacts craniofacial shape and ossification. The results were striking; while moderate overexpression of Fgf8 led to coronal craniosynostosis, high levels of Fgf8 overexpression resulted in ectopic cartilage formation throughout much of the cranial vault. While the premature suture fusion seen in craniosynostosis is the most commonly studied and treated aspect of craniosynostosis,

232 craniosynostosis also refers to the abnormal skull bone development that is associated with

dysmorphic skull shape. Compared to the cranial vault sutures, relatively little is known

about how mutations that are often associated with craniosynostosis (such as FGFRs) affect

cranial cells and tissues to cause facial shape aberrations, including midfacial retrusion,

hypertelorism, cranial base diminution, and palate deformities [381]. In addition to cranial

vault defects, both MR26F8 and MCAGF8 mice had a shortened snout and, unlike the

ossification defect, this facial shape malformation was not ameliorated by reduced Axin2 gene dosage. Additionally, the FaceCAGF8 mice also had a shortened snout, despite having decreased ossification, rather than craniosynostosis. Thus, in my mouse models overexpression of Fgf8 resulted in shortened snout regardless of craniosynostosis presentation. Historically, facial defects in craniosynostosis patients have been thought to be caused by the premature suture closure. However, my data and other studies [39] suggest that the facial defects in at least some craniosynostosis syndromes arise as a direct result of the mutation on patterning and development of the face, rather than as an indirect effect of premature suture closure. In support of this hypothesis, defects in facial and cranial base morphology preceded coronal suture fusion in Apert syndrome mice [382]. Thus, mouse models are useful in delineating the etiology of abnormal skull bone formation from the

premature suture fusion as: 1) abnormal skull bone shapes can be detected prior to the fusion

of the sutures and; 2) conditional misexpression of genes can be achieved with a variety of

Cre-recombinase transgenes, allowing for comparative analysis of the effect of the gene on

various aspects of skull and facial development. This comparative analysis can help

determine whether facial deformities arise solely in the presence of suture closure or

independent of suture closure, as is the case of the shortened snout in the FaceCAGF8 mice.

233 There is no evidence that potential future therapies that target and prevent premature suture

closure will also ameliorate the accompanying facial defects. Therefore, as new

craniosynostosis mutations are identified, both the causative role of the mutation in

premature suture closure and facial defects should be examined.

FGF Signaling and Endochondral Ossification

The presentation of craniosynostosis in the MR26F8 skulls, but cartilage formation in the MCAGF8 skulls, led to a model in which moderate overexpression of Fgf8 slows the differentiation of osteogenic progenitors leading to wider sutures at birth, while also inhibiting normal function of the sutures, ultimately causing craniosynostosis; however, as the dose of Fgf8 is increased, the skeletogenic progenitor cells are shifted from an osteogenic to a cartilaginous cell fate, at least partly due to decreased WNT signaling. While at first glance, the dosage of Fgf8 appears to lead to completely disparate outcomes, I postulate that these outcomes may not be as different as they appear on the surface.

Ectopic cartilage formation within the suture prior to premature suture closure has been observed in previous studies [343, 383, 384]. Additionally, ectopic expression of Fgf9 in the cranial vault led to cartilage formation that was later replaced by bone, suggesting bone formation occurred through endochondral ossification [334]. Thus, my data, in combination with the previous studies, suggests that craniosynostosis may result from a shift towards endochondral, rather than intramembranous ossification in the cranial bone. Therefore, impaired differentiation in the mesenchymal progenitor cells, resulting in a shift from osteogenic to a chondrogenic cell fate, may represent the first step in endochondral ossification and eventual craniosynostosis presentation.

234 Expression analysis of the MCAGF8 cranial tissue demonstrated that abnormal skull chondrogenesis was accompanied by changes in genes required for WNT signaling.

Moreover, the reduction in ossification could be partially rescued by manipulating Axin2 gene dosage, indicating a role for decreased WNT signaling in the cartilage pathology. In previous studies, decreased WNT signaling within the sutures led to premature suture closure through the process of endochondral ossification [343, 384]. Together, this suggests that increased FGF8 signaling leads to cartilage formation and craniosynostosis by decreasing

WNT signaling. Though I don’t observe ossification in the MCAGF8 skulls, this may be a result of the unusually high levels of FGF8 dysregulating other signals that make the cartilage abnormal and unable to form bone. Alternatively, the high dose of FGF8 may halt endochondral ossification at the cartilage phase while moderately increased FGF8 allows ossification to proceed. Additionally, it is unknown if the MR26F8 skulls have cartilage formation in the sutures prior to synostosis as the sutures were not closely examined prior to

P0. While ectopic cartilage does not form throughout the embryogenic cranial vault in the

MR26F8 skulls, more subtle cartilage formation may occur. More thorough analysis of the

MR26F8 sutures at multiple stages would be necessary to determine if cartilage formations

precedes craniosynostosis presentation. Additionally, it would be interesting to see if the

ectopic cartilage in the MCAGF8s could be driven to ossify, perhaps by driving Fgf8

overexpression in the ectoderm through an inducible Cre recombinase that could be turned

on long enough for cartilage to form, but off before the cartilage was so abnormal it could no

longer ossify.

Thus, it is still unclear how FGF8 mediated changes in WNT signaling affect suture

closure. In the MCAGF8, Axin2+/- mice, the expansion of bone development rescued the

235 lambdoid, but not the coronal suture. However, future studies are needed to determine how manipulating Axin2 gene dosage in the MR26F8 mice affects suture patency. Additionally,

manipulating Axin2 gene dosage may not be sufficient to normalize WNT signaling in the

sutures. Additional studies manipulating WNT signaling in the sutures through other means

are necessary to determine if FGF8 deactivation of WNT signaling is sufficient to cause

craniosynostosis. For example, the CAGF8 or MR26F8 mice could be used in conjunction

with a β−catenin LOF or GOF allele and Msx2-Cre or OC-Cre. This would allow me to

examine if the loss of β−catenin made the phenotypes worse – for example made the

MR26F8 skull appearance more like that or the MCAGF8 and if gain of β−catenin made the

phenotypes better – for example eliminated the craniosynostosis in the MR26F8s or made the

MCAGF8 skull appearance more like that of the MR26F8s. It is also possible that FGF8

mediated decrease in WNT signaling has differential effects depending on the suture. For

example, given the rescue of the lambdoid, but not the coronal suture in the MCAGF8,

Axin2+/- mice, the lambdoid suture may be sensitive to decreased WNT signaling while the coronal suture may be induced to prematurely fuse through other FGF mediated signals.

The presence of a stem cell population in the mouse suture was recently confirmed in both the paraxial mesoderm-derived coronal suture and neural crest-derived sagittal and interfrontal suture [127, 385]. This suture stem cell population is marked by Axin2

expression and persists until the suture closes . Interestingly, though Axin2 expression marks

calvarial suture mesenchymal stem cells, involved in intramembranous ossification, it does

not mark the bone marrow-derived skeletal stem cells, involved in endochondral ossification

[383, 385]. Thus, increased Fgf8 expression may be altering cell fate through modulation of

Axin2 expression levels in the suture stem cells. The phenotypic consequences of altering the

236 balance of FGF and WNT signaling in the calvaria are complex and, at times, contradictory.

For example, in mice, loss of Axin2 in combination with heterozygous deficiency of Fgfr1 in the sutural stem cells leads to cartilage formation within the sagittal suture followed by endochondral ossification resulting in craniosynostosis. This endochondral ossification may

be a consequence of altered BMP signaling as altering the balance of WNT and FGF

signaling in the sutural stem cells alters BMP signaling in the regions of chondrogenesis in

the suture [383]. While BMP signaling was not the primary focus of my investigations, the

RNAseq data indicated that that BMP signaling was altered in the MCAGF8 cranial vault.

However, in contrast to the above data, my studies suggest that increased FGF signaling

promotes cartilage formation through increased, rather than decreased Axin2 expression. In

support of this model, loss of β−catenin leads to cartilage formation in the calvaria [249]; this suggests that decreased WNT signaling, through either loss of β−catenin or gain of WNT

inhibitor Axin2, leads to cartilage formation. In combination with the above data, it therefore

seems that both a loss and gain of WNT signaling can result in cartilage formation suggesting

that the calvarial stem cells may have a biphasic response to WNT signaling. Thus, there

remains much to be elucidated about the complex interplay between WNT, FGF, and BMP

signaling in the sutural mesenchymal stem cells and how the manipulation of these signals

contributes to ossification and craniosynostosis.

Regardless of the importance of WNT signaling in FGF mediated craniosynostosis,

the idea that activation of FGF signaling leads to craniosynostosis through promoting

endochondral, rather than intramembranous ossification, is an intriguing notion as it hints at a possible explanation for the evolution of two types of ossification. Currently, there is no established hypothesis for why vertebrate skeletogenesis occurs through two different

237 mechanisms: endochondral and intramembranous ossification. One possible explanation for

this phenomenon is that intramembranous ossification evolved as a way to allow for skull

growth by preventing premature suture closure. Given that FGF activation may cause premature suture closure by promoting endochondral ossification, I would expect that FGF

signaling promotes chondrogenesis in the osteochondroprogenitor cells and thus loss of FGF

signaling in these cells would impair endochondral ossification. Indeed, deletion of Fgfr1 in

the osteochondroprogenitor cells leads to delayed osteoblast differentiation and deletion of

Fgfr2IIIc in mice leads to reduced growth of the appendicular skeleton and fewer

proliferating chondrocytes [163]. However, despite the large number of studies of the role of

FGF signaling in both endochondral and intramembranous bone, it is difficult to conclude if

loss of FGF signaling specifically effects endochondral bone formation more than

intramembranous bone formation due to functional redundancy between the FGF ligands and

the vast numbers of roles FGF signaling plays during development. Because FGF signaling

is important for so many aspects of development, the exact timing and location are critical for

interpretation. For example, limb ossification is most often used as the model for

endochondral ossification. However, deletion of FGFs in the limb AER result in severe limb

defects, but this is likely due to the importance of FGFs in limb outgrowth rather than

ossification. Additionally, since specific regulation of FGFs is critical in several of the steps

of both chondrogenesis and osteogenesis, it is difficult to tease apart the early effects from

later effects without carefully analyzing the development at each stage. For example, the

deletion of FGF2IIIc also caused delayed differentiation and mineralization of the skull vault,

but it is unclear if this is due to similar effects on the osteochondroprogenitor cells or later

effects on the osteoblast precursors [140]. Therefore, analysis of deletion, as well as

238 activation, of specific FGF pathway members in mesenchymal or osteochondroprogenitor

cells are needed to further the study of which signals promote endochondral vs

intramembranous ossification.

Spatiotemporal Dynamics

One initially puzzling aspect of the ectopic cartilageXII formation in the OR26F8,

OCAGF8, MCAGF8, and MCAGF8, Axin2+/- skulls was the location of the ectopic cartilage.

In the OR26F8 and OCAGF8 mice, bone formed in the lateral aspects of the skull, while cartilage formed in the dorsal region of the skull. This pattern was consistent with my expectations given that ossification begins in the lateral aspects of the skull and progresses towards the midline of the dorsal skull; thus, since OC-Cre is expressed in osteoblasts, I would expect ossification to occur in the lateral head, causing overexpression of Fgf8, that shifts cell fate in the dorsal head from osteogenic to chondrogenic. However, in the

MCAGF8 skulls, cartilage formed in the lateral skull and bone formed in the dorsal skull.

Additionally, the rescued ossification in the MCAGF8, Axin2+/- skulls extends the bone

formation from the dorsal skull towards the lateral region. However, the most lateral aspects

of the skull still form cartilage. This was puzzling to me as I expected that since ossification

begins in the lateral aspects of the skull, the mesenchymal progenitor cells first undergoing

ossification in this region would have been exposed to increased Fgf8 signaling for less time than the mesenchymal progenitor cells in the dorsal region of the skull, and thus would be more likely to have a less severe phenotype, not a more severe phenotype. However, upon

XII In the OR26F8 and OCAGF8 mice non-stained matrix similar to that found in the MCAGF8 mice was found in the cranium. Though I have not done the extensive analysis to confirm this non-stained matrix is cartilage in the OR26F8 and OCAGF8 skulls, for the sake of simplicity I will refer to this presumptive cartilage as cartilage in this section. 239 further reflection I have come up with the following model to explain the pattern of cartilage and bone formation in the MCAGF8 and MCAGF8, Axin2+/skulls, as depicted in Figure 86.

In the early embryo, both the ectoderm and mesenchyme are thin layers. Thus, when FGF8

Figure 86. Model of FGF8 and WNT signaling in the developing MCAGF8 mesenchyme over time. As the skull develops, the majority of the mesenchymal progenitor cells are exposed to less FGF8 and thus more WNT signaling due to the growth of the mesenchyme and diffusion of FGF8. For more details, see text.

240 diffuses from the ectoderm to the mesenchyme, the mesenchymal progenitor cells that normally differentiate into bone are exposed to high levels of FGF8, as they are all near the source of expression. As the embryo grows, the mesenchymal layer grows thicker. Thus, while the mesenchymal progenitor cells nearest the ectoderm will still be exposed to high levels of FGF8, the mesenchymal progenitor cells further from the ectoderm (the source of

FGF8) will be exposed to lower levels of FGF8 due to extra distance the FGF8 must travel to reach those cells. Thus, an FGF8 morphogen gradient is established in the mesenchyme resulting in a population of mesenchymal progenitor cells that receive differential amounts of

FGF8 signal (Figure 86, purple). As the thickness of the mesenchyme increases with age, the average amount of FGF8 signal received by the cells would decrease with age. The rescue of ossification in MCAGF8, Axin2+/- skulls suggests that increased Fgf8 expression decreases

WNT signaling; thus, I expect that the cells nearest the ectoderm, that receive the highest dose of FGF8, would have a more dramatic decrease in WNT signaling then the cells further from the ectoderm (Figure 86, yellow). Thus, the average cell in the thicker mesenchyme of the older embryos would have more WNT signaling (less of a decrease) than the cells in the earlier, thinner mesenchyme. As bone in the dorsal region of the head differentiates later than the bone in the lateral regions, bone in the dorsal region is derived from cells in the thicker mesenchyme, that are less exposed to the FGF8 overexpression. Thus, the dorsal bone can retain its bone identity, while the cells of the lateral head that differentiate from cells exposed to higher levels of FGF8 and lower levels of WNT signaling shift to a cartilage differentiation pathway.

In the MCAGF8, Axin2+/skulls, the loss of one allele of Axin2, a repressor of WNT signaling, leads to increased WNT signaling in the mesenchymal progenitor cells. Therefore,

241 I predict that this increase is enough to offset the decrease in WNT signaling caused by FGF8 overexpression in the mesenchymal progenitor cells that are exposed to small increases in

FGF8 (further from ectoderm), but not the cells closer to the ectoderm that are exposed to larger increases. This model explains why the rescue in bone formation occurs in the dorsal, but not lateral aspects of the skull. Additionally, this model explains why in certain regions of the MCAGF8 skull sections there was some bone mineralization (von Kossa) and differentiation (alkaline phosphatase staining) near the brain overlayed by cartilage cells. It also offers an explanation for why bone and cartilage overlapped in the transition zone of wholemount skulls.

This model also emphasizes the importance in considering how spatiotemporal differences in expression can lead to dramatically different phenotypes.

While this is not a new observation, its consideration grows increasingly important as the possibility of potential therapeutic interventions, via chemical inhibition, grows closer.

Variation in the application of treatments could lead to dramatically different effects if not applied in the correct location or time for that stage of development. Thus, the rational design of therapies will require an in-depth understanding of the spatiotemporal pattern of the genes regulating development of the applicable structure.

Increased Hedgehog Signaling from the Ectoderm and Mesenchyme Impairs

Craniofacial and Limb Development

In Chapter 6, I described the consequences of specific mesenchymal or ectodermal upregulation of HH signaling in the mouse face and limb using SmoM2, a constitutively active form of SMO. This allowed me to delineate between the effects of ectopic or increased HH ligand expression and overall increased signaling through the pathway. The

242 origin of increased HH signaling (ectoderm vs mesenchyme) in the head and limb was

important for some phenotypes, but not for others. In general, the origin affected patterning,

but not ossification. For example, both the EctSMOs and FaceSMOs had hypertelorism, but only the FaceSMOs had severely decreased ossification in the nasal bones. In contrast, only the EctSMOs had several other defects including secondary cleft palate, loss of the incisors, and ectopic bone formation in the palatal process of the maxilla. In the limbs, both EctSMOs and FaceSMOs had polydactyly, syndactyly, and shortened limbs, but FaceSMOs had thicker zeugopod bones whereas the EctSMOs had almost completely lost formation and ossification of the zeugopod.

While a previous study found that there was PtchLacZ expression, and thus HH signaling, in the limb ectoderm [363], PtchLacZ expression was not observed in the EctSMO control limb ectoderm. However, there was PtchLacZ expression in the EctSMO mutant ectoderm, suggesting that the ectodermal cells can respond to HH signaling. Additionally, while that study found that loss of SMO in the ectoderm at E10.5 using Msx2-Cre led to ectopic cartilage condensations, unpublished data by a former Williams lab member, Jian

Huang, did not find any difference in limb development when Smo or Shh was conditionally removed from the limb ectoderm at E8.5. Therefore, more analysis is necessary to determine if SHH has a role in the normal ectoderm and if that role varies between species or even mouse backgrounds.

Regardless of the role of HH signaling in the ectoderm during normal limb development, ectopic activation of HH signaling in the ectoderm led to severe defects in outgrowth and A-P patterning. While Shh expression in the mesenchyme may diffuse to the ectoderm during normal development, this was still unexpected as A-P patterning is thought

243 to be determined by a SHH morphogen gradient in the mesenchyme, with ectopic

mesenchymal expression of SHH in the anterior leading to pre-axial polydactyly. However,

increased HH signaling through constitutive activation of SMO (rather than SHH expression)

in both the FaceSMO and EctSMO limbs led to similar patterns of upregulation of several

genes in the mesenchyme, including Hoxd11 and Hoxd13, also associated with polydactyly.

Additionally, increased HH signaling in both the EctSMO and FaceSMO model similarly

allowed for the expression of Hand2 in the mesenchyme. Previously, both Hand2 and

Hoxd13 were believed to regulate HH signaling, but these data suggest a potential feedback

loop. Given the increasing number of potential feedback loops and HH target genes found to

be important in limb development, bioinformatic approaches may be necessary to further

elucidate the gene regulatory network in the limb. As there is wide variation in human

polydactyly presentation and relatively few causative genes have been identified [70], I postulate that some of these cases of polydactyly may be caused by mutations in regulators that lead to ectopic HH expression in the AER, similar to how mutations in the ZRS lead to ectopic HH signaling in the anterior mesenchyme.

Genetic Interactions Between FGF8 and HH

The genetic interaction between Shh and Fgf8 in the limb has been well described

[60, 386]. In the head, mutual antagonism between Fgf8 and Shh in the avian FNP

establishes the boundary of the FEZ and that same antagonism is believed to maintain the

FEZ in mice. Using Creface to increase both FGF8 ligand and HH signaling in the FNP, I

was struck by how similar the cranial phenotype was in the FaceCAGF8 and FaceSMO mice.

Both models display hypertelorism and decreased ossification of the nasal bones, but no

secondary cleft palate. Thus, the cross talk between these two signals could be better

244 elucidated by examining the phenotype of these mice when either both FGF8 and HH are

increased (CAGF8, SmoM2) or when one signal is increased and the other decreased

(CAGF8, Smoflox/Shhflox or SmoM2, Fgf8flox). This would help determine if one signal is upstream of the other or if the balance of these signals is more important during craniofacial morphogenesis.

Together, my findings with the various mouse models discussed in this dissertation have advanced our understanding of the role of FGF8 and HH signaling in head and limb morphogenesis. Additionally, some of these models had other defects that may be studied in the future, including the hair defect in MR26F8 and the skin, tooth, body wall closure, and

exencephaly defects seen in the EctSMO mice. Finally, described here are some of the first

studies using the R26F8 and CAGF8 to overexpress Fgf8. In conjunction with different Cre

recombinases, these alleles could be used to study the role of FGF8 signaling in many other

systems.

245 REFERENCES

1. Perrimon, Norbert, Chrysoula Pitsouli, and Ben-Zion Shilo. 2012. Signaling mechanisms controlling cell fate and embryonic patterning. Cold Spring Harbor perspectives in biology 4. Cold Spring Harbor Laboratory Press: a005975. doi:10.1101/cshperspect.a005975.

2. Moore, Keith L., T. V. N. Persaud, and Mark G. Torchia. 2016. Before we are born : essentials of embryology and birth defects. 9thed. Elsevier.

3. Schneider, Richard A., Diane Hu, and J. A. Helms. 1999. From head to toe: conservation of molecular signals regulating limb and craniofacial morphogenesis. Cell and Tissue Research 296. Springer-Verlag: 103–109. doi:10.1007/s004410051271.

4. Wilkie, Andrew O. M., Michael Oldridge, Tony Tang, and Robert E Maxson. 2017. Craniosynostosis and Related Limb Anomalies. In , 122–143. John Wiley & Sons, Ltd. doi:10.1002/0470846658.ch9.

5. Roosenboom, Jasmien, Greet Hens, Brooke C Mattern, Mark D Shriver, and Peter Claes. 2016. Exploring the Underlying Genetics of Craniofacial Morphology through Various Sources of Knowledge. BioMed research international 2016. Hindawi: 3054578. doi:10.1155/2016/3054578.

6. Kohn, L. A. P. 1991. The Role of Genetics in Craniofacial Morphology and Growth. Annual Review of Anthropology 20: 261–278. doi:10.1146/annurev.an.20.100191.001401.

7. A, Peter, Mossey, and Eduardo E Catilla. 2003. Global registry and database on craniofacial anomalies: report of a WHO Registry Meeting on Craniofacial Anomolies.

8. Vanderas, A P. 1987. Incidence of cleft lip, cleft palate, and cleft lip and palate among races: a review. The Cleft palate journal 24: 216–25.

9. Boulet, Sheree L, Scott D Grosse, Margaret A Honein, and Adolfo Correa-Villaseñor. 2009. Children with orofacial clefts: health-care use and costs among a privately insured population. Public health reports (Washington, D.C. : 1974) 124. SAGE Publications: 447–53. doi:10.1177/003335490912400315.

10. Centers for Disease Control and Prevention (CDC). 1995. Economic costs of birth defects and cerebral palsy--United States, 1992. MMWR. Morbidity and mortality weekly report 44: 694–9.

11. Sulik, Kathleen K. 2005. Genesis of alcohol-induced craniofacial dysmorphism. Experimental biology and medicine (Maywood, N.J.) 230: 366–75.

246 12. Correa, Adolfo, Suzanne M. Gilboa, Lilah M. Besser, Lorenzo D. Botto, Cynthia A. Moore, Charlotte A. Hobbs, Mario A. Cleves, Tiffany J. Riehle-Colarusso, D. Kim Waller, and E. Albert Reece. 2008. Diabetes mellitus and birth defects. American Journal of Obstetrics and Gynecology 199: 237.e1-237.e9. doi:10.1016/j.ajog.2008.06.028.

13. Honein, Margaret A., Sonja A. Rasmussen, Jennita Reefhuis, Paul A. Romitti, Edward J. Lammer, Lixian Sun, and Adolfo Correa. 2007. Maternal Smoking and Environmental Tobacco Smoke Exposure and the Risk of Orofacial Clefts. Epidemiology 18: 226–233. doi:10.1097/01.ede.0000254430.61294.c0.

14. Reefhuis, J., M. A. Honein, L. A. Schieve, S. A. Rasmussen, and National Birth Defects Prevention Study. 2011. Use of clomiphene citrate and birth defects, National Birth Defects Prevention Study, 1997-2005. Human Reproduction 26: 451–457. doi:10.1093/humrep/deq313.

15. Rasmussen, Sonja A., Mahsa M. Yazdy, Suzan L. Carmichael, Denise J. Jamieson, Mark A. Canfield, and Margaret A. Honein. 2007. Maternal Thyroid Disease as a Risk Factor for Craniosynostosis. Obstetrics & Gynecology 110: 369–377. doi:10.1097/01.AOG.0000270157.88896.76.

16. Dixon, Michael J., Mary L. Marazita, Terri H. Beaty, and Jeffrey C. Murray. 2011. Cleft lip and palate: understanding genetic and environmental influences. Nature Reviews Genetics 12: 167–178. doi:10.1038/nrg2933.

17. Christensen, K, and L E Mitchell. 1996. Familial recurrence-pattern analysis of nonsyndromic isolated cleft palate--a Danish Registry study. American journal of human genetics 58: 182–90.

18. Mossey, Peter A, Julian Little, Ron G Munger, Mike J Dixon, and William C Shaw. 2009. Cleft lip and palate. The Lancet 374: 1773–1785. doi:10.1016/S0140- 6736(09)60695-4.

19. Bister, D., P. Set, C. Cash, N. Coleman, and T. Fanshawe. 2011. Incidence of facial clefts in Cambridge, United Kingdom. The European Journal of Orthodontics 33: 372–376. doi:10.1093/ejo/cjq117.

20. Jones, M C. 1988. Etiology of facial clefts: prospective evaluation of 428 patients. The Cleft palate journal 25: 16–20.

21. FitzPatrick, D, and M Farrall. 1993. An estimation of the number of susceptibility loci for isolated cleft palate. Journal of craniofacial genetics and developmental biology 13: 230–5.

247 22. Marazita, Mary L., L. Leigh Field, Margaret E. Cooper, Rose Tobias, Brion S. Maher, Supakit Peanchitlertkajorn, and You-e Liu. 2002. Nonsyndromic Cleft Lip With or Without Cleft Palate in China: Assessment of Candidate Regions. The Cleft Palate- Craniofacial Journal 39: 149–156. doi:10.1597/1545- 1569(2002)039<0149:NCLWOW>2.0.CO;2.

23. Wilkie, Andrew O. M., and Gillian M. Morriss-Kay. 2001. GENETICS OF CRANIOFACIAL DEVELOPMENT AND MALFORMATION. Nature Reviews Genetics 2. Nature Publishing Group: 458–468. doi:10.1038/35076601.

24. Thomason, Helen A, Michael J Dixon, Helen A Thomason, and Michael J Dixon. 2009. Craniofacial Defects and Cleft Lip and Palate. In Encyclopedia of Life Sciences. Chichester, UK: John Wiley & Sons, Ltd. doi:10.1002/9780470015902.a0020915.

25. Kosowski, Tomasz R, William M Weathers, Erik M Wolfswinkel, and Emily B Ridgway. 2012. Cleft palate. Seminars in plastic surgery 26. Thieme Medical Publishers: 164–9. doi:10.1055/s-0033-1333883.

26. Melnick, Michael. 2003. Cleft Lip and Palate: From Origin to Treatment. American Journal of Human Genetics 72. Elsevier: 503.

27. Jugessur, Astanand, Min Shi, Håkon Kristian Gjessing, Rolv Terje Lie, Allen James Wilcox, Clarice Ring Weinberg, Kaare Christensen, et al. 2009. Genetic Determinants of Facial Clefting: Analysis of 357 Candidate Genes Using Two National Cleft Studies from Scandinavia. Edited by Syed A. Aziz. PLoS ONE 4: e5385. doi:10.1371/journal.pone.0005385.

28. Murray, J C. 2002. Gene/environment causes of cleft lip and/or palate. Clinical genetics 61: 248–56.

29. Mossey, Peter, and Julian Little. 2009. Addressing the challenges of cleft lip and palate research in India. Indian Journal of Plastic Surgery 42: 9. doi:10.4103/0970- 0358.57182.

30. Rahimov, Fedik, Mary L Marazita, Axel Visel, Margaret E Cooper, Michael J Hitchler, Michele Rubini, Frederick E Domann, et al. 2008. Disruption of an AP-2α binding site in an IRF6 enhancer is associated with cleft lip. Nature Genetics 40: 1341–1347. doi:10.1038/ng.242.

31. Kohli, Sarvraj Singh, and Virinder Singh Kohli. 2012. A comprehensive review of the genetic basis of cleft lip and palate. Journal of oral and maxillofacial pathology : JOMFP 16. Wolters Kluwer -- Medknow Publications: 64–72. doi:10.4103/0973- 029X.92976.

32. Jezewski, P A, A R Vieira, C Nishimura, B Ludwig, M Johnson, S E O’Brien, S Daack-Hirsch, et al. 2003. Complete sequencing shows a role for MSX1 in non- syndromic cleft lip and palate. Journal of medical genetics 40: 399–407.

248 33. Suzuki, Satoshi, Mary L. Marazita, Margaret E. Cooper, Nobutomo Miwa, Anne Hing, Astanand Jugessur, Nagato Natsume, et al. 2009. Mutations in BMP4 Are Associated with Subepithelial, Microform, and Overt Cleft Lip. The American Journal of Human Genetics 84: 406–411. doi:10.1016/j.ajhg.2009.02.002.

34. Osoegawa, K, G M Vessere, K H Utami, M A Mansilla, M K Johnson, B M Riley, J L’Heureux, et al. 2007. Identification of novel candidate genes associated with cleft lip and palate using array comparative genomic hybridisation. Journal of Medical Genetics 45: 81–86. doi:10.1136/jmg.2007.052191.

35. Riley, B. M., M. A. Mansilla, J. Ma, S. Daack-Hirsch, B. S. Maher, L. M. Raffensperger, E. T. Russo, et al. 2007. Impaired FGF signaling contributes to cleft lip and palate. Proceedings of the National Academy of Sciences 104: 4512–4517. doi:10.1073/pnas.0607956104.

36. Boulet, Sheree L, Sonja A Rasmussen, and Margaret A Honein. 2008. A population- based study of craniosynostosis in metropolitan Atlanta, 1989-2003. American journal of medical genetics. Part A 146A: 984–91. doi:10.1002/ajmg.a.32208.

37. Lajeunie, E, M Le Merrer, C Bonaïti-Pellie, D Marchac, and D Renier. 1995. Genetic study of nonsyndromic coronal craniosynostosis. American journal of medical genetics 55: 500–4. doi:10.1002/ajmg.1320550422.

38. Johnson, David, and Andrew O M Wilkie. 2011. Craniosynostosis. European journal of human genetics : EJHG 19. Nature Publishing Group: 369–76. doi:10.1038/ejhg.2010.235.

39. Flaherty, Kevin, Nandini Singh, and Joan T. Richtsmeier. 2016. Understanding craniosynostosis as a growth disorder 5: 429–459. doi:10.1002/wdev.227.

40. Heuzé, Yann, Gregory Holmes, Inga Peter, Joan T. Richtsmeier, and Ethylin Wang Jabs. 2014. Closing the Gap: Genetic and Genomic Continuum from Syndromic to Nonsyndromic Craniosynostoses. Current Genetic Medicine Reports 2. Springer US: 135–145. doi:10.1007/s40142-014-0042-x.

41. Srivastava, H C. 1992. Ossification of the membranous portion of the squamous part of the occipital bone in man. Journal of anatomy 180 ( Pt 2. Wiley-Blackwell: 219–24.

42. Cohen, M M. 1980. Perspectives on Craniosynostosis. The Western journal of medicine 132. BMJ Publishing Group: 507–13.

43. Haidar, Kabbani, and Talkad S. Raghuveer. 2004. Craniosynostosis. American Family Physician 69. American Academy of Family Physicians: 2863–2870.

44. Ciurea, Alexandru Vlad, and Corneliu Toader. 2009. Genetics of craniosynostosis: review of the literature. Journal of medicine and life 2. Carol Davila - University Press: 5–17.

249 45. Chumas, Paul D., Giuseppe Cinalli, Eric Arnaud, Daniel Marchac, and Dominique Renier. 1997. Classification of previously unclassified cases of craniosynostosis. Journal of Neurosurgery 86: 177–181. doi:10.3171/jns.1997.86.2.0177.

46. Wilkie, A O. 1997. Craniosynostosis: genes and mechanisms. Human molecular genetics 6. Oxford University Press: 1647–56. doi:10.1093/HMG/6.10.1647.

47. Ko, Jung Min. 2016. Genetic Syndromes Associated with Craniosynostosis. Journal of Korean Neurosurgical Society 59. The Korean Neurosurgical Society: 187–91. doi:10.3340/jkns.2016.59.3.187.

48. Chai, Yang, and Robert E. Maxson. 2006. Recent advances in craniofacial morphogenesis. Developmental Dynamics 235: 2353–2375. doi:10.1002/dvdy.20833.

49. Lohnes, D, M Mark, C Mendelsohn, P Dollé, A Dierich, P Gorry, A Gansmuller, and P Chambon. 1994. Function of the retinoic acid receptors (RARs) during development (I). Craniofacial and skeletal abnormalities in RAR double mutants. Development (Cambridge, England) 120: 2723–48.

50. Nottoli, T, S Hagopian-Donaldson, J Zhang, A Perkins, and T Williams. 1998. AP-2- null cells disrupt morphogenesis of the eye, face, and limbs in chimeric mice. Proceedings of the National Academy of Sciences of the United States of America 95: 13714–9.

51. Juriloff, Diana M., and Muriel J. Harris. 2008. Mouse genetic models of cleft lip with or without cleft palate. Birth Defects Research Part A: Clinical and Molecular Teratology 82: 63–77. doi:10.1002/bdra.20430.

52. Hu, Diane, and Ralph S. Marcucio. 2009. Unique organization of the frontonasal ectodermal zone in birds and mammals. Developmental Biology 325: 200–210. doi:10.1016/j.ydbio.2008.10.026.

53. Chang, Ching-Fang, Elizabeth N. Schock, David A. Billmire, and Samantha A. Brugmann. 2015. Craniofacial Syndromes: Etiology, Impact, and Treatment. In Principles of Developmental Genetics, 653–676. Elsevier. doi:10.1016/B978-0-12- 405945-0.00035-1.

54. Nelson, D.K K, and T Williams. 2004. Frontonasal process-specific disruption of AP- 2α results in postnatal midfacial hypoplasia, vascular anomalies, and nasal cavity defects. Developmental Biology 267: 72–92. doi:10.1016/j.ydbio.2003.10.033.

55. Grova, Monica, David D Lo, Daniel Montoro, Jeong S Hyun, Michael T Chung, Derrick C Wan, and Michael T Longaker. 2012. Models of cranial suture biology. The Journal of craniofacial surgery 23. NIH Public Access: 1954–8. doi:10.1097/SCS.0b013e318258ba53.

250 56. Zuniga, Aimée. 2015. Next generation limb development and evolution: old questions, new perspectives. Development (Cambridge, England) 142. Oxford University Press for The Company of Biologists Limited: 3810–20. doi:10.1242/dev.125757.

57. SAUNDERS, J W. 1948. The proximo-distal sequence of origin of the parts of the chick wing and the role of the ectoderm. The Journal of experimental zoology 108: 363–403.

58. Niswander, Lee. 2003. Pattern formation: old models out on a limb. Nature Reviews Genetics 4. Nature Publishing Group: 133–143. doi:10.1038/nrg1001.

59. Tabin, Cliff, and Lewis Wolpert. 2007. Rethinking the proximodistal axis of the vertebrate limb in the molecular era. Genes & development 21. Cold Spring Harbor Laboratory Press: 1433–42. doi:10.1101/gad.1547407.

60. Capdevila, Javier, and Juan Carlos Izpisúa Belmonte. 2001. Patterning Mechanisms Controlling Vertebrate Limb Development. Annual Review of Cell and Developmental Biology 17: 87–132. doi:10.1146/annurev.cellbio.17.1.87.

61. Vasluian, Ecaterina, Corry K van der Sluis, Anthonie J van Essen, Jorieke E H Bergman, Pieter U Dijkstra, Heleen A Reinders-Messelink, and Hermien E K de Walle. 2013. Birth prevalence for congenital limb defects in the northern Netherlands: a 30-year population-based study. BMC musculoskeletal disorders 14. BioMed Central: 323. doi:10.1186/1471-2474-14-323.

62. Jaruratanasirikul, Somchit, Boonsin Tangtrakulwanich, Pornruedee Rachatawiriyakul, Hutcha Sriplung, Wannee Limpitikul, Pathikan Dissaneevate, Nattasit Khunnarakpong, and Pongsak Tantichantakarun. 2016. Prevalence of congenital limb defects: Data from birth defects registries in three provinces in Southern Thailand. Congenital Anomalies 56: 203–208. doi:10.1111/cga.12154.

63. Wilkie, Andrew O M. 2003. Why study human limb malformations? Journal of anatomy 202. Wiley-Blackwell: 27–35. doi:10.1046/J.1469-7580.2003.00130.X.

64. Chitty, Lyn S, Louise Wilson, and David R. Griffen. 2009. Fetal skeletal abnormalities. In Fetal Medicine: Basic Science and Clinical Practice, ed. Charles Rodeck and Martin Whittle, 478–488. Elsevier.

65. Kher, A S, D R Gahankari, S R Tambwekar, A Doraiswamy, S Iyer, B A Bharucha, and R E Rana. 1996. Supernumerary limbs: a case report of a rare congenital anomaly. Annals of plastic surgery 37: 549–52.

66. McPherson, Fiona, Jaime L. Frias, Diane Spicer, John M. Opitz, and Enid F. Gilbert- Barness. 2003. Splenogonadal fusion-limb defect "syndrome" and associated malformations. American Journal of Medical Genetics 120A: 518–522. doi:10.1002/ajmg.a.10728.

251 67. Küster, Wolfgang, Widukind Lenz, Helena Kääriäinen, Frank Majewski, John M. Opitz, and James F. Reynolds. 1988. Congenital scalp defects with distal limb anomalies (Adams-Oliver syndrome): Report of ten cases and review of the literature. American Journal of Medical Genetics 31. Wiley Subscription Services, Inc., A Wiley Company: 99–115. doi:10.1002/ajmg.1320310112.

68. Goodman, Frances R. 2002. Limb malformations and the humanHOX genes. American Journal of Medical Genetics 112. Wiley Subscription Services, Inc., A Wiley Company: 256–265. doi:10.1002/ajmg.10776.

69. Swanson, Alfred B. 1976. A classification for congenital limb malformations. The Journal of Hand Surgery 1. W.B. Saunders: 8–22. doi:10.1016/S0363-5023(76)80021- 4.

70. Phadke, Shubha R., and V.H. Sankar. 2010. Polydactyly and genes. The Indian Journal of Pediatrics 77. Springer-Verlag: 277–281. doi:10.1007/s12098-010-0033-1.

71. Holmes, Lewis B. 2002. Teratogen-induced limb defects. American Journal of Medical Genetics 112: 297–303. doi:10.1002/ajmg.10781.

72. Therapontos, Christina, Lynda Erskine, Erin R Gardner, William D Figg, and Neil Vargesson. 2009. Thalidomide induces limb defects by preventing angiogenic outgrowth during early limb formation. Proceedings of the National Academy of Sciences of the United States of America 106. National Academy of Sciences: 8573–8. doi:10.1073/pnas.0901505106.

73. Taher, Leila, Nicole M. Collette, Deepa Murugesh, Evan Maxwell, Ivan Ovcharenko, and Gabriela G. Loots. 2011. Global Gene Expression Analysis of Murine Limb Development. Edited by Costanza Emanueli. PLoS ONE 6. Public Library of Science: e28358. doi:10.1371/journal.pone.0028358.

74. Al-Qattan, Mohammad M., and Scott H. Kozin. 2013. Update on Embryology of the Upper Limb. The Journal of Hand Surgery 38: 1835–1844. doi:10.1016/j.jhsa.2013.03.018.

75. Theiler, Karl. 1989. Earliest Signs of Fingers. In The House Mouse, 87–93. Berlin, Heidelberg: Springer Berlin Heidelberg. doi:10.1007/978-3-642-88418-4_23.

76. Kaufman, Matthew H., and Jonathan B. L. Bard. 1999. The anatomical basis of mouse development. Academic Press.

77. Niswander, Lee. 2002. Interplay between the molecular signals that control vertebrate limb development. The International journal of developmental biology 46: 877–81.

78. Laufer, E, C E Nelson, R L Johnson, B A Morgan, and C Tabin. 1994. Sonic hedgehog and Fgf-4 act through a signaling cascade and feedback loop to integrate growth and patterning of the developing limb bud. Cell 79: 993–1003.

252 79. Niswander, Lee, Susan Jeffrey, Gail R. Martin, and Cheryll Tickle. 1994. A positive feedback loop coordinates growth and patterning in the vertebrate limb. Nature 371: 609–612. doi:10.1038/371609a0.

80. Tickle, Cheryll. 2006. Making digit patterns in the vertebrate limb. Nature Reviews Molecular Cell Biology 7. Nature Publishing Group: 45–53. doi:10.1038/nrm1830.

81. Parr, Brian A., and Andrew P. McMahon. 1995. Dorsalizing signal Wnt-7a required for normal polarity of D–V and A–P axes of mouse limb. Nature 374: 350–353. doi:10.1038/374350a0.

82. Loomis, Cynthia A., Esther Harris, Jacques Michaud, Wolfgang Wurst, Mark Hanks, and Alexandra L. Joyner. 1996. The mouse Engrailed-1 gene and ventral limb patterning. Nature 382. Nature Publishing Group: 360–363. doi:10.1038/382360a0.

83. Ibrahim, Daniel. 2015. ChIP-seq reveals mutation-specific pathomechanisms of HOXD13 missense mutations. Humboldt University of Berlin.

84. Mariani, Francesca V, Christina P Ahn, and Gail R Martin. 2008. Genetic evidence that FGFs have an instructive role in limb proximal-distal patterning. Nature 453. NIH Public Access: 401–5. doi:10.1038/nature06876.

85. Zeller, Rolf, Javier López-Ríos, and Aimée Zuniga. 2009. Vertebrate limb bud development: moving towards integrative analysis of organogenesis. Nature Reviews Genetics 10. Nature Publishing Group: 845–858. doi:10.1038/nrg2681.

86. Mariani, Francesca V., and Gail R. Martin. 2003. Deciphering skeletal patterning: clues from the limb. Nature 423. Nature Publishing Group: 319–325. doi:10.1038/nature01655.

87. Summerbell, D. 1974. A quantitative analysis of the effect of excision of the AER from the chick limb-bud. Journal of embryology and experimental morphology 32: 651–60.

88. Niswander, L, C Tickle, A Vogel, I Booth, and G R Martin. 1993. FGF-4 replaces the apical ectodermal ridge and directs outgrowth and patterning of the limb. Cell 75: 579–87.

89. Martin, Gail R., Mark Lewandoski, and Xin Sun. 2000. Fgf8 signalling from the AER is essential for normal limb development. Nature Genetics 26. Nature Publishing Group: 460–463. doi:10.1038/82609.

90. Kraus, P, D Fraidenraich, and C A Loomis. 2001. Some distal limb structures develop in mice lacking Sonic hedgehog signaling. Mechanisms of development 100: 45–58.

253 91. Chiang, Chin, Ying Litingtung, Matthew P. Harris, B.Kay Simandl, Yina Li, Philip A. Beachy, and John F. Fallon. 2001. Manifestation of the Limb Prepattern: Limb Development in the Absence of Sonic Hedgehog Function. Developmental Biology 236: 421–435. doi:10.1006/dbio.2001.0346.

92. Chiang, Chin, Ying Litingtung, Eric Lee, Keith E. Young, Jeffrey L Corden, Heiner Westphal, and Philip A. Beachy. 1996. Cyclopia and defective axial patterning in mice lacking Sonic hedgehog gene function. Nature 383. Nature Publishing Group: 407– 413. doi:10.1038/383407a0.

93. Hill, Robert E. 2007. How to make a zone of polarizing activity: Insights into limb development via the abnormality preaxial polydactyly. Development, Growth & Differentiation 49: 439–448. doi:10.1111/j.1440-169X.2007.00943.x.

94. Fernandez-Teran, M, M E Piedra, I S Kathiriya, D Srivastava, J C Rodriguez-Rey, and M A Ros. 2000. Role of dHAND in the anterior-posterior polarization of the limb bud: implications for the Sonic hedgehog pathway. Development (Cambridge, England) 127: 2133–42.

95. Charité, J, D G McFadden, and E N Olson. 2000. The bHLH transcription factor dHAND controls Sonic hedgehog expression and establishment of the zone of polarizing activity during limb development. Development (Cambridge, England) 127: 2461–70.

96. Sagai, T., Masaki Hosoya, Youichi Mizushina, Masaru Tamura, and Toshihiko Shiroishi. 2005. Elimination of a long-range cis-regulatory module causes complete loss of limb-specific Shh expression and truncation of the mouse limb. Development 132: 797–803. doi:10.1242/dev.01613.

97. Lettice, L. A., Simon J.H. Heaney, Lorna A. Purdie, Li Li, Philippe de Beer, Ben A. Oostra, Debbie Goode, Greg Elgar, Robert E. Hill, and Esther de Graaff. 2003. A long-range Shh enhancer regulates expression in the developing limb and fin and is associated with preaxial polydactyly. Human Molecular Genetics 12. Oxford University Press: 1725–1735. doi:10.1093/hmg/ddg180.

98. Crossley, P.H., and G.R. Martin. 1995. The mouse Fgf8 gene encodes a family of polypeptides and is expressed in regions that direct outgrowth and patterning in the developing embryo. Development 121.

99. Khokha, Mustafa K, David Hsu, Lisa J Brunet, Marc S Dionne, and Richard M Harland. 2003. Gremlin is the BMP antagonist required for maintenance of Shh and Fgf signals during limb patterning. Nature Genetics 34: 303–307. doi:10.1038/ng1178.

100. Zeller, Rolf, Aimée Zúñiga, Anna-Pavlina G. Haramis, and Andrew P. McMahon. 1999. Signal relay by BMP antagonism controls the SHH/FGF4 feedback loop in vertebrate limb buds. Nature 401: 598–602. doi:10.1038/44157.

254 101. Pownall, Mary Elizabeth, and Harry V. Isaacs. 2010. FGF Signalling in Vertebrate Development. San Rafael, CA: Morgan & Claypool Life Sciences.

102. Chen, H, R L Johnson, Haixu Chen, and Randy L Johnson. 1999. Dorsoventral patterning of the vertebrate limb: a process governed by multiple events. Cell Tissue Res 296: 67–73.

103. Parr, B A, M J Shea, G Vassileva, and A P McMahon. 1993. Mouse Wnt genes exhibit discrete domains of expression in the early embryonic CNS and limb buds. Development (Cambridge, England) 119: 247–61.

104. Ornitz, David M, and Pierre J Marie. 2015. Fibroblast growth factor signaling in skeletal development and disease. Genes & development 29. Cold Spring Harbor Laboratory Press: 1463–86. doi:10.1101/gad.266551.115.

105. Long, Fanxin. 2012. Building strong bones: molecular regulation of the osteoblast lineage. Nature reviews. Molecular cell biology 13. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.: 27–38. doi:10.1038/nrm3254.

106. Brugmann, Samantha A., Minal D. Tapadia, and Jill A. Helms. 2006. The Molecular Origins of Species‐Specific Facial Pattern. In Current topics in developmental biology, 73:1–42. doi:10.1016/S0070-2153(05)73001-5.

107. Mori-Akiyama, Y., H. Akiyama, D. H. Rowitch, and B. de Crombrugghe. 2003. Sox9 is required for determination of the chondrogenic cell lineage in the cranial neural crest. Proceedings of the National Academy of Sciences 100: 9360–9365. doi:10.1073/pnas.1631288100.

108. Nakashima, Kazuhisa, Xin Zhou, Gary Kunkel, Zhaoping Zhang, Jian Min Deng, Richard R. Behringer, and Benoit de Crombrugghe. 2002. The Novel Zinc Finger- Containing Transcription Factor Osterix Is Required for Osteoblast Differentiation and Bone Formation. Cell 108: 17–29. doi:10.1016/S0092-8674(01)00622-5.

109. Rodda, Stephen J., and Andrew P McMahon. 2006. Distinct roles for Hedgehog and canonical Wnt signaling in specification, differentiation and maintenance of osteoblast progenitors. Development (Cambridge, England) 133: 3231–44. doi:10.1242/dev.02480.

110. Day, Timothy F., Xizhi Guo, Lisa Garrett-Beal, and Yingzi Yang. 2005. Wnt/β- Catenin Signaling in Mesenchymal Progenitors Controls Osteoblast and Chondrocyte Differentiation during Vertebrate Skeletogenesis. Developmental Cell 8: 739–750. doi:10.1016/j.devcel.2005.03.016.

111. Hill, Theo P., Daniela Später, Makoto M. Taketo, Walter Birchmeier, and Christine Hartmann. 2005. Canonical Wnt/β-Catenin Signaling Prevents Osteoblasts from Differentiating into Chondrocytes. Developmental Cell 8: 727–738. doi:10.1016/j.devcel.2005.02.013.

255 112. Peck, William A., and Leonard Rifas. 1982. Regulation of Osteoblast Activity and the Osteoblast-Osteocyte Transformation. In , 393–400. Springer, Boston, MA. doi:10.1007/978-1-4684-4259-5_45.

113. Franz-Odendaal, Tamara A., Brian K. Hall, and P. Eckhard Witten. 2006. Buried alive: How osteoblasts become osteocytes. Developmental Dynamics 235. Wiley‐Liss, Inc.: 176–190. doi:10.1002/dvdy.20603.

114. Bellido, Teresita, Lilian I. Plotkin, and Angela Bruzzaniti. 2014. Bone Cells. In Basic and Applied Bone Biology, 27–45. Elsevier. doi:10.1016/B978-0-12-416015-6.00002- 2.

115. Mackie, E.J., Y.A. Ahmed, L. Tatarczuch, K.-S. Chen, and M. Mirams. 2008. Endochondral ossification: How cartilage is converted into bone in the developing skeleton. The International Journal of Biochemistry & Cell Biology 40. Pergamon: 46–62. doi:10.1016/J.BIOCEL.2007.06.009.

116. Zhou, Xin, Klaus von der Mark, Stephen Henry, William Norton, Henry Adams, and Benoit de Crombrugghe. 2014. Chondrocytes Transdifferentiate into Osteoblasts in Endochondral Bone during Development, Postnatal Growth and Fracture Healing in Mice. Edited by Matthew L. Warman. PLoS Genetics 10: e1004820. doi:10.1371/journal.pgen.1004820.

117. Yang, L., K. Y. Tsang, H. C. Tang, D. Chan, and K. S. E. Cheah. 2014. Hypertrophic chondrocytes can become osteoblasts and osteocytes in endochondral bone formation. Proceedings of the National Academy of Sciences 111: 12097–12102. doi:10.1073/pnas.1302703111.

118. Couly, G.F., P.M. Coltey, and N.M. Le Douarin. 1993. The triple origin of skull in higher vertebrates: a study in quail-chick chimeras. Development 117.

119. Jiang, Xiaobing, Sachiko Iseki, Robert E. Maxson, Henry M. Sucov, and Gillian M. Morriss-Kay. 2002. Tissue Origins and Interactions in the Mammalian Skull Vault. Developmental Biology 241: 106–116. doi:10.1006/dbio.2001.0487.

120. Noden, Drew M. 1992. Vertebrate craniofacial development: novel approaches and new dilemmas. Current Opinion in Genetics & Development 2: 576–581. doi:10.1016/S0959-437X(05)80175-3.

121. Opperman, L A. 2000. Cranial sutures as intramembranous bone growth sites. Developmental dynamics : an official publication of the American Association of Anatomists 219. John Wiley & Sons, Inc.: 472–85. doi:10.1002/1097- 0177(2000)9999:9999<::AID-DVDY1073>3.0.CO;2-F.

122. Evans, Darrell J.R., and Drew M. Noden. 2006. Spatial relations between avian craniofacial neural crest and paraxial mesoderm cells. Developmental Dynamics 235: 1310–1325. doi:10.1002/dvdy.20663.

256 123. Noden, Drew M, and Paul A Trainor. 2005. Relations and interactions between cranial mesoderm and neural crest populations. Journal of anatomy 207. Wiley-Blackwell: 575–601. doi:10.1111/j.1469-7580.2005.00473.x.

124. Hall, Brian K., and T. Miyake. 2000. All for one and one for all: condensations and the initiation of skeletal development. BioEssays 22: 138–147. doi:10.1002/(SICI)1521- 1878(200002)22:2<138::AID-BIES5>3.0.CO;2-4.

125. Nie, X, K Luukko, and P Kettunen. 2006. FGF signalling in craniofacial development and developmental disorders. Oral diseases 12. Blackwell Publishing Ltd: 102–11. doi:10.1111/j.1601-0825.2005.01176.x.

126. Mishina, Yuji, and Taylor Nicholas Snider. 2014. Neural crest cell signaling pathways critical to cranial bone development and pathology. Experimental Cell Research 325. NIH Public Access: 138–147. doi:10.1016/j.yexcr.2014.01.019.

127. Quarto, Natalina, Bjorn Behr, Shuli Li, and Michael T Longaker. 2009. Differential FGF ligands and FGF receptors expression pattern in frontal and parietal calvarial bones. Cells, tissues, organs 190. Karger Publishers: 158–69. doi:10.1159/000202789.

128. Ornitz, David M, and Pierre J Marie. 2002. FGF signaling pathways in endochondral and intramembranous bone development and human genetic disease. Genes & development 16. Cold Spring Harbor Laboratory Press: 1446–65. doi:10.1101/gad.990702.

129. Cohen, M M. 2000. Merging the old skeletal biology with the new. II. Molecular aspects of bone formation and bone growth. Journal of craniofacial genetics and developmental biology 20: 94–106.

130. Opperman, Lynne A., Thomas M. Sweeney, Julie Redmon, John A. Persing, and Roy C. Ogle. 1993. Tissue interactions with underlying dura mater inhibit osseous obliteration of developing cranial sutures. Developmental Dynamics 198: 312–322. doi:10.1002/aja.1001980408.

131. Levine, J P, J P Bradley, D A Roth, J G McCarthy, and M T Longaker. 1998. Studies in cranial suture biology: regional dura mater determines overlying suture biology. Plastic and reconstructive surgery 101: 1441–7.

132. Cooper, Gregory M, Emily L Durham, James J Cray, Michael I Siegel, Joseph E Losee, Mark P Mooney, and Mark P. Mooney. 2012. Tissue interactions between craniosynostotic dura mater and bone. The Journal of craniofacial surgery 23. NIH Public Access: 919–24. doi:10.1097/SCS.0b013e31824e645f.

133. Katsianou, Maria A, Christos Adamopoulos, Heleni Vastardis, and Efthimia K Basdra. 2016. Signaling mechanisms implicated in cranial sutures pathophysiology: Craniosynostosis. BBA clinical 6. Elsevier: 165–176. doi:10.1016/j.bbacli.2016.04.006.

257 134. King, Thomas C., and Thomas C. King. 2007. Tissue Homeostasis, Damage, and Repair. In Elsevier’s Integrated Pathology, 59–88. Elsevier. doi:10.1016/B978-0-323- 04328-1.50009-7.

135. Nubuyuki, Itoh, and David Ornitz. 2004. Evolution of the Fgf and Fgfr gene families. Trends in Genetics 20. Elsevier Current Trends: 563–569. doi:10.1016/J.TIG.2004.08.007.

136. Olsen, Shaun K., Meirav Garbi, Niccolo Zampieri, Anna V. Eliseenkova, David M. Ornitz, Mitchell Goldfarb, and Moosa Mohammadi. 2003. Fibroblast Growth Factor (FGF) Homologous Factors Share Structural but Not Functional Homology with FGFs. Journal of Biological Chemistry 278: 34226–34236. doi:10.1074/jbc.M303183200.

137. Smallwood, P M, I Munoz-Sanjuan, P Tong, J P Macke, S H Hendry, D J Gilbert, N G Copeland, N A Jenkins, and J Nathans. 1996. Fibroblast growth factor (FGF) homologous factors: new members of the FGF family implicated in nervous system development. Proceedings of the National Academy of Sciences of the United States of America 93. National Academy of Sciences: 9850–7.

138. Johnson, D E, and L T Williams. 1993. Structural and functional diversity in the FGF receptor multigene family. Advances in cancer research 60: 1–41.

139. Brooks, A. Nigel, Elaine Kilgour, and Paul D. Smith. 2012. Molecular Pathways: Fibroblast Growth Factor Signaling: A New Therapeutic Opportunity in Cancer. Clinical Cancer Research 18.

140. Xu, X, M Weinstein, C Li, M Naski, R I Cohen, D M Ornitz, P Leder, and C Deng. 1998. Fibroblast growth factor receptor 2 (FGFR2)-mediated reciprocal regulation loop between FGF8 and FGF10 is essential for limb induction. Development (Cambridge, England) 125: 753–65.

141. Min, H, D M Danilenko, S A Scully, B Bolon, B D Ring, J E Tarpley, M DeRose, and W S Simonet. 1998. Fgf-10 is required for both limb and lung development and exhibits striking functional similarity to Drosophila branchless. Genes & development 12. Cold Spring Harbor Laboratory Press: 3156–61.

142. MacArthur, C A, A Lawshé, J Xu, S Santos-Ocampo, M Heikinheimo, A T Chellaiah, and D M Ornitz. 1995. FGF-8 isoforms activate receptor splice forms that are expressed in mesenchymal regions of mouse development. Development (Cambridge, England) 121: 3603–13.

143. Miki, T, D P Bottaro, T P Fleming, C L Smith, W H Burgess, A M Chan, and S A Aaronson. 1992. Determination of ligand-binding specificity by alternative splicing: two distinct growth factor receptors encoded by a single gene. Proceedings of the National Academy of Sciences of the United States of America 89: 246–50.

258 144. Gemel, Joanna, Michael Gorry, Garth D. Ehrlich, and Craig A. MacArthur. 1996. Structure and Sequence of HumanFGF8. Genomics 35: 253–257. doi:10.1006/geno.1996.0349.

145. Guo, Qiuxia, James Y H Li, and H Nakamura. 2007. Distinct functions of the major Fgf8 spliceform, Fgf8b, before and during mouse gastrulation. Development (Cambridge, England) 134. The Company of Biologists Ltd: 2251–60. doi:10.1242/dev.004929.

146. Fantl, Wendy J., Daniel E. Johnson, and Lewis T. Williams. 1993. Signalling by Receptor Tyrosine Kinases. Annual Review of Biochemistry 62: 453–481. doi:10.1146/annurev.bi.62.070193.002321.

147. Brewer, J Richard, Pierre Mazot, and Philippe Soriano. 2016. Genetic insights into the mechanisms of Fgf signaling. Genes & development 30. Cold Spring Harbor Laboratory Press: 751–71. doi:10.1101/gad.277137.115.

148. Monsoro-Burq, Anne-Hélène, Russell B Fletcher, and Richard M Harland. 2003. Neural crest induction by paraxial mesoderm in Xenopus embryos requires FGF signals. Development (Cambridge, England) 130: 3111–24.

149. Villanueva, Sandra, Alvaro Glavic, Pablo Ruiz, and Roberto Mayor. 2002. Posteriorization by FGF, Wnt, and Retinoic Acid Is Required for Neural Crest Induction. Developmental Biology 241: 289–301. doi:10.1006/dbio.2001.0485.

150. Baker, C V, and M Bronner-Fraser. 1997. The origins of the neural crest. Part I: embryonic induction. Mechanisms of development 69: 3–11.

151. Wilke, Todd A., Sharon Gubbels, Jacquie Schwartz, and Joy M. Richman. 1997. Expression of fibroblast growth factor receptors (FGFR1, FGFR2, FGFR3) in the developing head and face. Developmental Dynamics 210: 41–52. doi:10.1002/(SICI)1097-0177(199709)210:1<41::AID-AJA5>3.0.CO;2-1.

152. Bachler, M, and A Neubüser. 2001. Expression of members of the Fgf family and their receptors during midfacial development. Mechanisms of development 100: 313–6.

153. Sun, Xin, Mark Lewandoski, Erik N. Meyers, Yi-Hsin H Liu, Robert E. Maxson, and Gail R. Martin. 2000. Conditional inactivation of Fgf4 reveals complexity of signalling during limb bud development. Nature genetics 25. Nature Publishing Group: 83–6. doi:10.1038/75644.

154. Meyers, Erik N., Mark Lewandoski1, Gail R. Martin, Mark Lewandoski, and Gail R. Martin. 1998. An Fgf8 mutant allelic series generated by Cre- and Flp-mediated recombination. Nature Genetics 18: 136–141. doi:10.1038/ng0298-136.

259 155. Kawauchi, S., Jianyong Shou, Rosaysela Santos, Jean M Hébert, Susan K McConnell, Ivor Mason, and Anne L Calof. 2005. Fgf8 expression defines a morphogenetic center required for olfactory neurogenesis and nasal cavity development in the mouse. Development 132: 5211–5223. doi:10.1242/dev.02143.

156. Cobourne, Martyn T, and Paul T Sharpe. 2003. Tooth and jaw: molecular mechanisms of patterning in the first branchial arch. Archives of oral biology 48: 1–14.

157. Griffin, John N, Claudia Compagnucci, Diane Hu, Jennifer Fish, Ophir Klein, Ralph Marcucio, and Michael J Depew. 2013. Fgf8 dosage determines midfacial integration and polarity within the nasal and optic capsules. Developmental biology 374. NIH Public Access: 185–97. doi:10.1016/j.ydbio.2012.11.014.

158. Newbern, J., J. Zhong, R. S. Wickramasinghe, X. Li, Y. Wu, I. Samuels, N. Cherosky, et al. 2008. Mouse and human phenotypes indicate a critical conserved role for ERK2 signaling in neural crest development. Proceedings of the National Academy of Sciences 105: 17115–17120. doi:10.1073/pnas.0805239105.

159. Trumpp, A, M J Depew, J L Rubenstein, J M Bishop, and G R Martin. 1999. Cre- mediated gene inactivation demonstrates that FGF8 is required for cell survival and patterning of the first branchial arch. Genes & development 13. Cold Spring Harbor Laboratory Press: 3136–48.

160. Ferguson, C A, A S Tucker, and P T Sharpe. 2000. Temporospatial cell interactions regulating mandibular and maxillary arch patterning. Development (Cambridge, England) 127: 403–12.

161. Tucker, A S, G Yamada, M Grigoriou, V Pachnis, and P T Sharpe. 1999. Fgf-8 determines rostral-caudal polarity in the first branchial arch. Development (Cambridge, England) 126: 51–61.

162. Abzhanov, Arhat, and Clifford J Tabin. 2004. Shh and Fgf8 act synergistically to drive cartilage outgrowth during cranial development. Developmental Biology 273: 134– 148. doi:10.1016/j.ydbio.2004.05.028.

163. Su, Nan, Min Jin, and Lin Chen. 2014. Role of FGF/FGFR signaling in skeletal development and homeostasis: learning from mouse models. Bone research 2. Nature Publishing Group: 14003. doi:10.1038/boneres.2014.3.

164. Tholpady, Sunil S, Mohamed M Abdelaal, Craig R Dufresne, Thomas J Gampper, Kant Y Lin, John A Jane, Raymond F Morgan, and Roy C Ogle. 2004. Aberrant bony vasculature associated with activating fibroblast growth factor receptor mutations accompanying Crouzon syndrome. The Journal of craniofacial surgery 15: 431-5-8.

165. Spector, Jason A, Joshua A Greenwald, Stephen M Warren, Pierre J Bouletreau, Robert C Detch, Peter J Fagenholz, Francesca E Crisera, and Michael T Longaker. 2002. Dura mater biology: autocrine and paracrine effects of fibroblast growth factor 2. Plastic and reconstructive surgery 109: 645–54.

260 166. Li, S., N. Quarto, and M. T. Longaker. 2007. Dura mater-derived FGF-2 mediates mitogenic signaling in calvarial osteoblasts. AJP: Cell Physiology 293: C1834–C1842. doi:10.1152/ajpcell.00135.2007.

167. Spicer, Douglas. 2009. FGF9 on the move. Nature Genetics 41. Nature Publishing Group: 272–273. doi:10.1038/ng0309-272.

168. Wu, Xiao-lin, Ming-min Gu, Lei Huang, Xue-song Liu, Hong-xin Zhang, Xiao-yi Ding, Jian-qiang Xu, et al. 2009. Multiple Synostoses Syndrome Is Due to a Missense Mutation in Exon 2 of FGF9 Gene. The American Journal of Human Genetics 85: 53– 63. doi:10.1016/j.ajhg.2009.06.007.

169. Robin, Nathaniel H, Marni J Falk, and Chad R Haldeman-Englert. 2011. FGFR- Related Craniosynostosis Syndromes. GeneReviews(®). University of Washington, Seattle.

170. Rutland, Paul, Louise J. Pulleyn, William Reardon, Michael Baraitser, Richard Hayward, Barry Jones, Sue Malcolm, et al. 1995. Identical mutations in the FGFR2 gene cause both Pfeiffer and Crouzon syndrome phenotypes. Nature Genetics 9: 173– 176. doi:10.1038/ng0295-173.

171. Meyers, G A, D Day, R Goldberg, D L Daentl, K A Przylepa, L J Abrams, J M Graham, et al. 1996. FGFR2 exon IIIa and IIIc mutations in Crouzon, Jackson-Weiss, and Pfeiffer syndromes: evidence for missense changes, insertions, and a deletion due to alternative RNA splicing. American journal of human genetics 58: 491–8.

172. Neilson, K M, and R Friesel. 1996. Ligand-independent activation of fibroblast growth factor receptors by point mutations in the extracellular, transmembrane, and kinase domains. The Journal of biological chemistry 271. American Society for Biochemistry and Molecular Biology: 25049–57. doi:10.1074/JBC.271.40.25049.

173. Monsonego-Ornan, E, R Adar, T Feferman, O Segev, and A Yayon. 2000. The transmembrane mutation G380R in fibroblast growth factor receptor 3 uncouples ligand-mediated receptor activation from down-regulation. Molecular and cellular biology 20. American Society for Microbiology (ASM): 516–22.

174. Sarabipour, Sarvenaz, and Kalina Hristova. 2016. Mechanism of FGF receptor dimerization and activation. Nature communications 7. Nature Publishing Group: 10262. doi:10.1038/ncomms10262.

175. Burke, David, David Wilkes, Tom L. Blundell, and Sue Malcolm. 1998. Fibroblast growth factor receptors: lessons from the genes. Trends in Biochemical Sciences 23: 59–62. doi:10.1016/S0968-0004(97)01170-5.

176. Anderson, J, H D Burns, P Enriquez-Harris, A O Wilkie, and J K Heath. 1998. Apert syndrome mutations in fibroblast growth factor receptor 2 exhibit increased affinity for FGF ligand. Human molecular genetics 7. Oxford University Press: 1475–83. doi:10.1093/HMG/7.9.1475.

261 177. Ibrahimi, O. A., A. V. Eliseenkova, A. N. Plotnikov, K. Yu, D. M. Ornitz, and M. Mohammadi. 2001. Structural basis for fibroblast growth factor receptor 2 activation in Apert syndrome. Proceedings of the National Academy of Sciences 98: 7182–7187. doi:10.1073/pnas.121183798.

178. Ibrahimi, O. A., Fuming Zhang, Anna V Eliseenkova, Nobuyuki Itoh, Robert J Linhardt, and Moosa Mohammadi. 2004. Biochemical analysis of pathogenic ligand- dependent FGFR2 mutations suggests distinct pathophysiological mechanisms for craniofacial and limb abnormalities. Human Molecular Genetics 13: 2313–2324. doi:10.1093/hmg/ddh235.

179. Yu, K., A. B. Herr, G. Waksman, and D. M. Ornitz. 2000. Loss of fibroblast growth factor receptor 2 ligand-binding specificity in Apert syndrome. Proceedings of the National Academy of Sciences 97: 14536–14541. doi:10.1073/pnas.97.26.14536.

180. Ibrahimi, O. A., Fuming Zhang, Anna V Eliseenkova, Robert J Linhardt, and Moosa Mohammadi. 2003. Proline to arginine mutations in FGF receptors 1 and 3 result in Pfeiffer and Muenke craniosynostosis syndromes through enhancement of FGF binding affinity. Human Molecular Genetics 13: 69–78. doi:10.1093/hmg/ddh011.

181. De Moerlooze, L, B Spencer-Dene, J M Revest, M Hajihosseini, I Rosewell, and C Dickson. 2000. An important role for the IIIb isoform of fibroblast growth factor receptor 2 (FGFR2) in mesenchymal-epithelial signalling during mouse organogenesis. Development (Cambridge, England) 127: 483–92.

182. Barrow, Jeffery R, Kirk R Thomas, Oreda Boussadia-Zahui, Robert Moore, Rolf Kemler, Mario R Capecchi, and Andrew P McMahon. 2003. Ectodermal Wnt3/beta- catenin signaling is required for the establishment and maintenance of the apical ectodermal ridge. Genes & development 17. Cold Spring Harbor Laboratory Press: 394–409. doi:10.1101/gad.1044903.

183. Kawakami, Y, J Capdevila, D Büscher, T Itoh, C Rodríguez Esteban, and J C Izpisúa Belmonte. 2001. WNT signals control FGF-dependent limb initiation and AER induction in the chick embryo. Cell 104: 891–900.

184. Yaylaoglu, Murat Burak, Andrew Titmus, Axel Visel, Gonzalo Alvarez-Bolado, Christina Thaller, and Gregor Eichele. 2005. Comprehensive expression atlas of fibroblast growth factors and their receptors generated by a novel robotic in situ hybridization platform. Developmental Dynamics 234: 371–386. doi:10.1002/dvdy.20441.

185. Moon, Anne M., and Mario R. Capecchi. 2000. Fgf8 is required for outgrowth and patterning of the limbs. Nature Genetics 26. Nature Publishing Group: 455–459. doi:10.1038/82601.

262 186. Yu, Kai, and David M Ornitz. 2008. FGF signaling regulates mesenchymal differentiation and skeletal patterning along the limb bud proximodistal axis. Development (Cambridge, England) 135. The Company of Biologists Ltd: 483–91. doi:10.1242/dev.013268.

187. Corson, Laura Beth, Yojiro Yamanaka, Ka-Man Venus Lai, and Janet Rossant. 2003. Spatial and temporal patterns of ERK signaling during mouse embryogenesis. Development (Cambridge, England) 130. Oxford University Press for The Company of Biologists Limited: 4527–37. doi:10.1242/dev.00669.

188. Bouldin, Cortney M., and Brian D. Harfe. 2009. Aberrant FGF signaling, independent of ectopic hedgehog signaling, initiates preaxial polydactyly in Dorking chickens. Developmental Biology 334. Academic Press: 133–141. doi:10.1016/j.ydbio.2009.07.009.

189. Ornitz, David M. 2005. FGF signaling in the developing endochondral skeleton. Cytokine & growth factor reviews 16. NIH Public Access: 205–13. doi:10.1016/j.cytogfr.2005.02.003.

190. Ohbayashi, Norihiko, Masaki Shibayama, Yoko Kurotaki, Mayumi Imanishi, Toshihiko Fujimori, Nobuyuki Itoh, and Shinji Takada. 2002. FGF18 is required for normal cell proliferation and differentiation during osteogenesis and chondrogenesis. Genes & development 16. Cold Spring Harbor Laboratory Press: 870–9. doi:10.1101/gad.965702.

191. Liu, Z., Jingsong Xu, Jennifer S Colvin, and David M Ornitz. 2002. Coordination of chondrogenesis and osteogenesis by fibroblast growth factor 18. Genes & Development 16: 859–869. doi:10.1101/gad.965602.

192. Xu, J, Z Liu, and D M Ornitz. 2000. Temporal and spatial gradients of Fgf8 and Fgf17 regulate proliferation and differentiation of midline cerebellar structures. Development (Cambridge, England) 127: 1833–43.

193. Guo, L, L Degenstein, and E Fuchs. 1996. Keratinocyte growth factor is required for hair development but not for wound healing. Genes & development 10: 165–75.

194. Fietz, M J, J P Concordet, R Barbosa, R Johnson, S Krauss, A P McMahon, C Tabin, and P W Ingham. 1994. The hedgehog gene family in Drosophila and vertebrate development. Development (Cambridge, England). Supplement: 43–51.

195. Kawai, Yasuhiro, Junko Noguchi, Kouyou Akiyama, Yuriko Takeno, Yasuhiro Fujiwara, Shimpei Kajita, Takehito Tsuji, Kazuhiro Kikuchi, Hiroyuki Kaneko, and Tetsuo Kunieda. 2011. A missense mutation of the Dhh gene is associated with male pseudohermaphroditic rats showing impaired Leydig cell development. Reproduction (Cambridge, England) 141. Society for Reproduction and Fertility: 217–25. doi:10.1530/REP-10-0006.

263 196. Yao, Humphrey Hung-Chang, Wendy Whoriskey, and Blanche Capel. 2002. Desert Hedgehog/Patched 1 signaling specifies fetal Leydig cell fate in testis organogenesis. Genes & development 16. Cold Spring Harbor Laboratory Press: 1433–40. doi:10.1101/gad.981202.

197. Pan, Angel, Le Chang, Alan Nguyen, and Aaron W. James. 2013. A review of hedgehog signaling in cranial bone development. Frontiers in physiology 4. Frontiers: 61. doi:10.3389/fphys.2013.00061.

198. Simpson, Fiona, Markus C Kerr, and Carol Wicking. 2009. Trafficking, development and hedgehog. Mechanisms of Development 126: 279–288. doi:10.1016/j.mod.2009.01.007.

199. Briscoe, James, and Pascal P. Thérond. 2013. The mechanisms of Hedgehog signalling and its roles in development and disease. Nature Reviews Molecular Cell Biology 14: 418–431. doi:10.1038/nrm3598.

200. Belloni, E., M. Muenke, E. Roessler, G. Traverse, J. Siegel-Bartelt, A. Frumkin, H.F. Mitchell, et al. 1996. Identification of Sonic hedgehog as a candidate gene responsible for holoprosencephaly. Nature Genetics 14: 353–356. doi:10.1038/ng1196-353.

201. Slaney, Sarah F., Frances R. Goodman, Betty L.C. Eilers-Walsman, Bryan D. Hall, Denise K. Williams, Ian D. Young, Richard D. Hayward, Barry M. Jones, Arnold L. Christianson, and Robin M. Winter. 1999. Acromelic Frontonasal Dysostosis. American Journal of Medical Genetics 83. Wiley Subscription Services, Inc., A Wiley Company: 109–116. doi:10.1002/(SICI)1096-8628(19990312)83:2<109::AID- AJMG6>3.0.CO;2-8.

202. Dworkin, Sebastian, Yeliz Boglev, Harley Owens, and Stephen Goldie. 2016. The Role of Sonic Hedgehog in Craniofacial Patterning, Morphogenesis and Cranial Neural Crest Survival. Journal of Developmental Biology 4. Multidisciplinary Digital Publishing Institute: 24. doi:10.3390/jdb4030024.

203. Hu, D, and J A Helms. 1999. The role of sonic hedgehog in normal and abnormal craniofacial morphogenesis. Development (Cambridge, England) 126: 4873–84.

204. Jeong, Juhee, Junhao Mao, Toyoaki Tenzen, Andreas H Kottmann, and Andrew P McMahon. 2004. Hedgehog signaling in the neural crest cells regulates the patterning and growth of facial primordia. Genes & development 18. Cold Spring Harbor Laboratory Press: 937–51. doi:10.1101/gad.1190304.

205. Coventry, S, R P Kapur, and J R Siebert. 1998. Cyclopamine-induced holoprosencephaly and associated craniofacial malformations in the golden hamster: anatomic and molecular events. Pediatric and developmental pathology : the official journal of the Society for Pediatric Pathology and the Paediatric Pathology Society 1: 29–41.

264 206. Ahlgren, Sara C, Vijaya Thakur, and Marianne Bronner-Fraser. 2002. Sonic hedgehog rescues cranial neural crest from cell death induced by ethanol exposure. Proceedings of the National Academy of Sciences of the United States of America 99. National Academy of Sciences: 10476–81. doi:10.1073/pnas.162356199.

207. O’Neil, Erica. 2010. Role of Sonic Hedgehog (Shh) in Alcohol-Induced Craniofacial Abnormalities. Arizona State University. School of Life Sciences. Center for Biology and Society. Embryo Project Encyclopedia.

208. Kim, HJ J, DP P Rice, PJ J Kettunen, and I Thesleff. 1998. FGF-, BMP- and Shh- mediated signalling pathways in the regulation of cranial suture morphogenesis and calvarial bone development. Development 125: 1241–1251.

209. Lenton, Kelly, Aaron W. James, Alina Manu, Samantha A. Brugmann, Daniel Birker, Emily R. Nelson, Philipp Leucht, Jill A. Helms, and Michael T. Longaker. 2011. Indian hedgehog positively regulates calvarial ossification and modulates bone morphogenetic protein signaling. Genesis (New York, N.Y. : 2000) 49: 784–96. doi:10.1002/dvg.20768.

210. Jacob, Shushan, Changshan Wu, Theresa A. Freeman, Eiki Koyama, and Richard E. Kirschner. 2007. Expression of Indian Hedgehog, BMP-4 and Noggin in Craniosynostosis Induced by Fetal Constraint. Annals of Plastic Surgery 58: 215–221. doi:10.1097/01.sap.0000232833.41739.a5.

211. Levi, Benjamin, Aaron W. James, Emily R. Nelson, Samantha A. Brugmann, Michael Sorkin, Alina Manu, and Michael T. Longaker. 2011. Role of Indian Hedgehog Signaling in Palatal Osteogenesis. Plastic and Reconstructive Surgery 127: 1182– 1190. doi:10.1097/PRS.0b013e3182043a07.

212. Nott, Rhoda L, Eric J Stelnicki, Judith A Mack, Yixin Ben, Ronal Mitchell, and Mark P Mooney. 2002. Changes in the protein expression of hedgehog and patched-1 in perisutural tissues induced by cranial distraction. Plastic and reconstructive surgery 110: 523–32.

213. Klopocki, Eva, Silke Lohan, Francesco Brancati, Randi Koll, Anja Brehm, Petra Seemann, Katarina Dathe, et al. 2011. Copy-Number Variations Involving the IHH Locus Are Associated with Syndactyly and Craniosynostosis. The American Journal of Human Genetics 88: 70–75. doi:10.1016/j.ajhg.2010.11.006.

214. Feng, Weiguo, Irene Choi, David E Clouthier, Lee Niswander, and Trevor Williams. 2013. The Ptch1(DL) mouse: a new model to study lambdoid craniosynostosis and basal cell nevus syndrome-associated skeletal defects. Genesis (New York, N.Y. : 2000) 51: 677–89. doi:10.1002/dvg.22416.

265 215. Jenkins, Dagan, Dominik Seelow, Fernanda S Jehee, Chad A Perlyn, Luis G Alonso, Daniela F Bueno, Dian Donnai, et al. 2007. RAB23 mutations in Carpenter syndrome imply an unexpected role for hedgehog signaling in cranial-suture development and obesity. American journal of human genetics 80. Elsevier: 1162–70. doi:10.1086/518047.

216. Abzhanov, Arhat, Stephen J Rodda, Andrew P McMahon, and Clifford J Tabin. 2007. Regulation of skeletogenic differentiation in cranial dermal bone. Development (Cambridge, England) 134: 3133–44. doi:10.1242/dev.002709.

217. Tian, Ye, Ying Xu, Qin Fu, and Yufeng Dong. 2012. Osterix is Required for Sonic Hedgehog-Induced Osteoblastic MC3T3-E1 Cell Differentiation. Cell Biochemistry and Biophysics 64: 169–176. doi:10.1007/s12013-012-9369-7.

218. James, Aaron W., Shen Pang, Asal Askarinam, Mirko Corselli, Janette N. Zara, Raghav Goyal, Le Chang, et al. 2012. Additive Effects of Sonic Hedgehog and Nell-1 Signaling in Osteogenic Versus Adipogenic Differentiation of Human Adipose- Derived Stromal Cells. Stem Cells and Development 21: 2170–2178. doi:10.1089/scd.2011.0461.

219. James, Aaron W, Philipp Leucht, Benjamin Levi, Antoine L Carre, Yue Xu, Jill A Helms, and Michael T Longaker. 2010. Sonic Hedgehog influences the balance of osteogenesis and adipogenesis in mouse adipose-derived stromal cells. Tissue engineering. Part A 16: 2605–16. doi:10.1089/ten.TEA.2010.0048.

220. van der Horst, Geertje, Hetty Farih-Sips, Clemens W G M Löwik, and Marcel Karperien. 2003. Hedgehog stimulates only osteoblastic differentiation of undifferentiated KS483 cells. Bone 33: 899–910.

221. Yuasa, Takahito, Hiroko Kataoka, Naoki Kinto, Masahiro Iwamoto, Motomi Enomoto-Iwamoto, Shun-ichiro Iemura, Naoto Ueno, Yasuaki Shibata, Hisashi Kurosawa, and Akira Yamaguchi. 2002. Sonic hedgehog is involved in osteoblast differentiation by cooperating with BMP-2. Journal of Cellular Physiology 193: 225– 232. doi:10.1002/jcp.10166.

222. Biesecker, Leslie G. 2011. Polydactyly: how many disorders and how many genes? 2010 update. Developmental dynamics : an official publication of the American Association of Anatomists 240. NIH Public Access: 931–42. doi:10.1002/dvdy.22609.

223. Hill, Robert E, and Laura A Lettice. 2013. Alterations to the remote control of Shh gene expression cause congenital abnormalities. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 368. The Royal Society: 20120357. doi:10.1098/rstb.2012.0357.

224. Kalff-Suske, M, A Wild, J Topp, M Wessling, E M Jacobsen, D Bornholdt, H Engel, et al. 1999. Point mutations throughout the GLI3 gene cause Greig cephalopolysyndactyly syndrome. Human molecular genetics 8: 1769–77.

266 225. Hill, P., B. Wang, and U. Ruther. 2007. The molecular basis of Pallister Hall associated polydactyly. Human Molecular Genetics 16: 2089–2096. doi:10.1093/hmg/ddm156.

226. Nelson, Craig E, Bruce A Morgan, Ann C Burke, Ed Laufer, Enrico Dimambro, L Charles Murtaugh, Ellen Gonzales, Lino Tessarollo, Luis F Parada, and Cliff Tabin. 1996. Analysis of Hox gene expression in the chick limb bud. Development 122: 1449–1466.

227. Tickle, Cheryll, and Matthew Towers. 2017. Sonic Hedgehog Signaling in Limb Development. Frontiers in cell and developmental biology 5. Frontiers Media SA: 14. doi:10.3389/fcell.2017.00014.

228. Cohen, M. Michael. 2003. The hedgehog signaling network. American Journal of Medical Genetics 123A: 5–28. doi:10.1002/ajmg.a.20495.

229. St-Jacques, B., M. Hammerschmidt, and A. P. McMahon. 1999. Indian hedgehog signaling regulates proliferation and differentiation of chondrocytes and is essential for bone formation. Genes & Development 13: 2072–2086. doi:10.1101/gad.13.16.2072.

230. Long, F., and D. M. Ornitz. 2013. Development of the Endochondral Skeleton. Cold Spring Harbor Perspectives in Biology 5: a008334–a008334. doi:10.1101/cshperspect.a008334.

231. Hilton, Matthew J, Xiaolin Tu, Julie Cook, Hongliang Hu, and Fanxin Long. 2005. Ihh controls cartilage development by antagonizing Gli3, but requires additional effectors to regulate osteoblast and vascular development. Development (Cambridge, England) 132. The Company of Biologists Ltd: 4339–51. doi:10.1242/dev.02025.

232. Litingtung, Ying, Randall D. Dahn, Yina Li, John F. Fallon, and Chin Chiang. 2002. Shh and Gli3 are dispensable for limb skeleton formation but regulate digit number and identity. Nature 418: 979–983. doi:10.1038/nature01033.

233. Yang, Jing, Philipp Andre, Ling Ye, and Ying-Zi Yang. 2015. The Hedgehog signalling pathway in bone formation. International Journal of Oral Science 7. Nature Publishing Group: 73–79. doi:10.1038/ijos.2015.14.

234. Peifer, M, and P Polakis. 2000. Wnt signaling in oncogenesis and embryogenesis--a look outside the nucleus. Science (New York, N.Y.) 287. American Association for the Advancement of Science: 1606–9. doi:10.1126/SCIENCE.287.5458.1606.

235. Hobmayer, Bert, Fabian Rentzsch, Kerstin Kuhn, Christoph M. Happel, Christoph Cramer von Laue, Petra Snyder, Ute Rothbächer, and Thomas W. Holstein. 2000. WNT signalling molecules act in axis formation in the diploblastic metazoan Hydra. Nature 407. Nature Publishing Group: 186–189. doi:10.1038/35025063.

267 236. Cadigan, K M, and R Nusse. 1997. Wnt signaling: a common theme in animal development. Genes & development 11. Cold Spring Harbor Laboratory Press: 3286– 305. doi:10.1101/GAD.11.24.3286.

237. Eisenmann, David M. 2005. Wnt signaling. Edited by Iva Greenwald. WormBook The C. elegans Research Community. doi:10.1895/wormbook.1.7.1.

238. Nusse, Roel, and Hans Clevers. 2017. Wnt/β-Catenin Signaling, Disease, and Emerging Therapeutic Modalities. Cell 169. Cell Press: 985–999. doi:10.1016/J.CELL.2017.05.016.

239. BARROW, J. 2006. Wnt/PCP signaling: A veritable polar star in establishing patterns of polarity in embryonic tissues. Seminars in Cell & Developmental Biology 17: 185– 193. doi:10.1016/j.semcdb.2006.04.002.

240. Kühl, M, L C Sheldahl, M Park, J R Miller, and R T Moon. 2000. The Wnt/Ca2+ pathway: a new vertebrate Wnt signaling pathway takes shape. Trends in genetics : TIG 16: 279–83.

241. Eastman, Q, and R Grosschedl. 1999. Regulation of LEF-1/TCF transcription factors by Wnt and other signals. Current opinion in cell biology 11: 233–40.

242. Geetha-Loganathan, Poongodi, Suresh Nimmagadda, and Martin Scaal. 2008. Wnt signaling in limb organogenesis. Organogenesis 4. Taylor & Francis: 109–15.

243. Jho, Eek-hoon, Tong Zhang, Claire Domon, Choun-Ki Joo, Jean-Noel Freund, and Frank Costantini. 2002. Wnt/beta-catenin/Tcf signaling induces the transcription of Axin2, a negative regulator of the signaling pathway. Molecular and cellular biology 22. American Society for Microbiology: 1172–83. doi:10.1128/MCB.22.4.1172- 1183.2002.

244. Haegel, H, L Larue, M Ohsugi, L Fedorov, K Herrenknecht, and R Kemler. 1995. Lack of beta-catenin affects mouse development at gastrulation. Development (Cambridge, England) 121: 3529–37.

245. Brault, V, R Moore, S Kutsch, M Ishibashi, D H Rowitch, A P McMahon, L Sommer, O Boussadia, and R Kemler. 2001. Inactivation of the beta-catenin gene by Wnt1-Cre- mediated deletion results in dramatic brain malformation and failure of craniofacial development. Development (Cambridge, England) 128: 1253–64.

246. Day, Timothy F., Xizhi Guo, Lisa Garrett-Beal, and Yingzi Yang. 2005. Wnt/β- Catenin Signaling in Mesenchymal Progenitors Controls Osteoblast and Chondrocyte Differentiation during Vertebrate Skeletogenesis. Developmental Cell 8: 739–750. doi:10.1016/j.devcel.2005.03.016.

268 247. Tufan, A C, and R S Tuan. 2001. Wnt regulation of limb mesenchymal chondrogenesis is accompanied by altered N-cadherin-related functions. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 15: 1436–8.

248. Rudnicki, Julie A., and Anthony M.C. Brown. 1997. Inhibition of Chondrogenesis by Wnt Gene Expression in Vivo and in Vitro. Developmental Biology 185: 104–118. doi:10.1006/dbio.1997.8536.

249. Tran, Thu H., Andrew Jarrell, Gabriel E. Zentner, Adrienne Welsh, Isaac Brownell, Peter C. Scacheri, and Radhika Atit. 2010. Role of canonical Wnt signaling/β-catenin via Dermo1 in cranial dermal cell development. Development 137.

250. Kishigami, Satoshi, and Yuji Mishina. 2005. BMP signaling and early embryonic patterning. Cytokine & Growth Factor Reviews 16: 265–278. doi:10.1016/j.cytogfr.2005.04.002.

251. Nie, Xuguang, Keijo Luukko, and Paivi Kettunen. 2006. BMP signalling in craniofacial development. The International Journal of Developmental Biology 50: 511–21. doi:10.1387/ijdb.052101xn.

252. Bandyopadhyay, Amitabha, Prem Swaroop Yadav, and Paritosh Prashar. 2013. BMP signaling in development and diseases: A pharmacological perspective. Biochemical Pharmacology 85: 857–864. doi:10.1016/j.bcp.2013.01.004.

253. Massague, J., and D Wotton. 2000. NEW EMBO MEMBERS REVIEW: Transcriptional control by the TGF-beta/Smad signaling system. The EMBO Journal 19: 1745–1754. doi:10.1093/emboj/19.8.1745.

254. Graf, Daniel, Zeba Malik, Satoru Hayano, and Yuji Mishina. 2016. Common mechanisms in development and disease: BMP signaling in craniofacial development. Cytokine & Growth Factor Reviews 27: 129–139. doi:10.1016/j.cytogfr.2015.11.004.

255. Warren, Stephen M., Lisa J. Brunet, Richard M. Harland, Aris N. Economides, and Michael T. Longaker. 2003. The BMP antagonist noggin regulates cranial suture fusion. Nature 422: 625–629. doi:10.1038/nature01545.

256. Francis-West, P H, T Tatla, and P M Brickell. 1994. Expression patterns of the bone morphogenetic protein genes Bmp-4 and Bmp-2 in the developing chick face suggest a role in outgrowth of the primordia. Developmental dynamics : an official publication of the American Association of Anatomists 201: 168–78. doi:10.1002/aja.1002010207.

257. Francis-West, P, R Ladher, A Barlow, and A Graveson. 1998. Signalling interactions during facial development. Mechanisms of development 75: 3–28.

258. Bennett, J.H., P Hunt, and P. Thorogood. 1995. Bone morphogenetic protein-2 and -4 expression during murine orofacial development. Archives of Oral Biology 40. Pergamon: 847–854. doi:10.1016/0003-9969(95)00047-S.

269 259. Bandyopadhyay, Amitabha, Kunikazu Tsuji, Karen Cox, Brian D. Harfe, Vicki Rosen, and Clifford J. Tabin. 2006. Genetic Analysis of the Roles of BMP2, BMP4, and BMP7 in Limb Patterning and Skeletogenesis. PLoS Genetics 2: e216. doi:10.1371/journal.pgen.0020216.

260. Benazet, J.-D., M. Bischofberger, E. Tiecke, A. Goncalves, J. F. Martin, A. Zuniga, F. Naef, and R. Zeller. 2009. A Self-Regulatory System of Interlinked Signaling Feedback Loops Controls Mouse Limb Patterning. Science 323: 1050–1053. doi:10.1126/science.1168755.

261. Selever, Jennifer, Wei Liu, Mei-Fang Lu, Richard R. Behringer, and James F. Martin. 2004. Bmp4 in limb bud mesoderm regulates digit pattern by controlling AER development. Developmental Biology 276: 268–279. doi:10.1016/j.ydbio.2004.08.024.

262. Norrie, Jacqueline L, Jordan P Lewandowski, Cortney M Bouldin, Smita Amarnath, Qiang Li, Martha S Vokes, Lauren I R Ehrlich, Brian D Harfe, and Steven A Vokes. 2014. Dynamics of BMP signaling in limb bud mesenchyme and polydactyly. Developmental biology 393. NIH Public Access: 270–281. doi:10.1016/j.ydbio.2014.07.003.

263. Benazet, J.-D., E. Pignatti, A. Nugent, E. Unal, F. Laurent, and R. Zeller. 2012. Smad4 is required to induce digit ray primordia and to initiate the aggregation and differentiation of chondrogenic progenitors in mouse limb buds. Development 139: 4250–4260. doi:10.1242/dev.084822.

264. Badugu, Amarendra, Conradin Kraemer, Philipp Germann, Denis Menshykau, and Dagmar Iber. 2012. Digit patterning during limb development as a result of the BMP- receptor interaction. Scientific Reports 2: 991. doi:10.1038/srep00991.

265. Sears, K E. 2008. Molecular determinants of bat wing development. Cells, tissues, organs 187. Karger Publishers: 6–12. doi:10.1159/000109959.

266. Barna, Maria, and Lee Niswander. 2007. Visualization of Cartilage Formation: Insight into Cellular Properties of Skeletal Progenitors and Chondrodysplasia Syndromes. Developmental Cell 12: 931–941. doi:10.1016/j.devcel.2007.04.016.

267. Duprez, D, E J Bell, M K Richardson, C W Archer, L Wolpert, P M Brickell, and P H Francis-West. 1996. Overexpression of BMP-2 and BMP-4 alters the size and shape of developing skeletal elements in the chick limb. Mechanisms of development 57: 145– 57.

268. Hsu, D R, A N Economides, X Wang, P M Eimon, and R M Harland. 1998. The Xenopus dorsalizing factor Gremlin identifies a novel family of secreted proteins that antagonize BMP activities. Molecular cell 1: 673–83.

270 269. Eimon, Peter M., and Richard M. Harland. 1999. In Xenopus Embryos, BMP Heterodimers Are Not Required for Mesoderm Induction, but BMP Activity Is Necessary for Dorsal/Ventral Patterning. Developmental Biology 216: 29–40. doi:10.1006/dbio.1999.9496.

270. Qin, Jane Yuxia, Li Zhang, Kayla L. Clift, Imge Hulur, Andy Peng Xiang, Bing- Zhong Ren, and Bruce T. Lahn. 2010. Systematic Comparison of Constitutive Promoters and the Doxycycline-Inducible Promoter. Edited by Immo A. Hansen. PLoS ONE 5. Public Library of Science: e10611. doi:10.1371/journal.pone.0010611.

271. Schock, Elizabeth N., Jaime N. Struve, Ching-Fang Chang, Trevor J. Williams, John Snedeker, Aria C. Attia, Rolf W. Stottmann, and Samantha A. Brugmann. 2017. A tissue-specific role for intraflagellar transport genes during craniofacial development. Edited by Knut Stieger. PLOS ONE 12. Public Library of Science: e0174206. doi:10.1371/journal.pone.0174206.

272. Reid, Bethany S., Hui Yang, Vida Senkus Melvin, Makoto M. Taketo, and Trevor Williams. 2011. Ectodermal WNT/β-catenin signaling shapes the mouse face. Developmental Biology 349: 261–269. doi:10.1016/j.ydbio.2010.11.012.

273. Sun, X, M Lewandoski, E N Meyers, Y H Liu, R E Maxson, and G R Martin. 2000. Conditional inactivation of Fgf4 reveals complexity of signalling during limb bud development. Nature genetics 25: 83–6. doi:10.1038/75644.

274. McLeod, M. Jean. 1980. Differential staining of cartilage and bone in whole mouse fetuses by alcian blue and alizarin red S. Teratology 22. Wiley Subscription Services, Inc., A Wiley Company: 299–301. doi:10.1002/tera.1420220306.

275. Jegalian, Beatrice G., and Eddy M. De Robertis. 1992. Homeotic transformations in the mouse induced by overexpression of a human Hox3.3 transgene. Cell 71: 901–910. doi:10.1016/0092-8674(92)90387-R.

276. Gruber, H E. 1992. Adaptations of Goldner’s Masson trichrome stain for the study of undecalcified plastic embedded bone. Biotechnic & histochemistry : official publication of the Biological Stain Commission 67: 30–4.

277. George Clark. 1981. Staining procedures. 4th ed. Baltimore, MD: Williams & Wilkins.

278. Mowry, R.W. 1956. Alcian blue technics for the histochemical study of acidic carbohydrates. Journal of Histochemistry & Cytochemistry 4: 407.

279. Sridharan, Gokul, and Akhil A Shankar. 2012. Toluidine blue: A review of its chemistry and clinical utility. Journal of oral and maxillofacial pathology : JOMFP 16. Medknow Publications: 251–5. doi:10.4103/0973-029X.99081.

280. Matt Lewis. 2017. In situ hybridisation to alpha satellite sequences (chromosome specific). http://www.methodbook.net/probes/insitu.html. Accessed May 14.

271 281. Chai, Y, X Jiang, Y Ito, P Bringas, J Han, DH H Rowitch, P Soriano, AP P McMahon, and HM M Sucov. 2000. Fate of the mammalian cranial neural crest during tooth and mandibular morphogenesis. Development 127. The Company of Biologists Ltd: 1671– 1679.

282. Bradford, Andrew P., Kenneth Jones, Katerina Kechris, Justin Chosich, Michael Montague, Wesley C. Warren, Margaret C. May, et al. 2015. Joint MiRNA/mRNA Expression Profiling Reveals Changes Consistent with Development of Dysfunctional Corpus Luteum after Weight Gain. Edited by Meijia Zhang. PLOS ONE 10. Public Library of Science: e0135163. doi:10.1371/journal.pone.0135163.

283. Huang, D. W., B. T. Sherman, and R. A. Lempicki. 2009. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Research 37: 1–13. doi:10.1093/nar/gkn923.

284. Huang, Da Wei, Brad T Sherman, and Richard A Lempicki. 2008. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature Protocols 4: 44–57. doi:10.1038/nprot.2008.211.

285. Sun, PP, and JA Persing. 1999. Craniosynostosis. In Principles and practice of pediatric neurosurgery, ed. A.Leland. Albright, P.David. Adelson, and Ian F. Pollack, 2nd ed., 219–242. Thieme.

286. Panchal, Jayesh, and Venus Uttchin. 2003. Management of Craniosynostosis. Plastic and Reconstructive Surgery 111: 2032–2049. doi:10.1097/01.PRS.0000056839.94034.47.

287. Berry-Candelario, John, Emily B. Ridgway, Ronald T. Grondin, Gary F. Rogers, and Mark R. Proctor. 2011. Endoscope-assisted strip craniectomy and postoperative helmet therapy for treatment of craniosynostosis. Neurosurgical Focus 31: E5. doi:10.3171/2011.6.FOCUS1198.

288. Shillito, J, and D D Matson. 1968. Craniosynostosis: a review of 519 surgical patients. Pediatrics 41: 829–53.

289. Fearon, Jeffrey A., Emily B. McLaughlin, and John C. Kolar. 2006. Sagittal Craniosynostosis: Surgical Outcomes and Long-Term Growth. Plastic and Reconstructive Surgery 117: 532–541. doi:10.1097/01.prs.0000200774.31311.09.

290. Thomas, Gregory P L, Andrew O M Wilkie, Peter G Richards, and Steven A Wall. 2005. FGFR3 P250R mutation increases the risk of reoperation in apparent “nonsyndromic” coronal craniosynostosis. The Journal of craniofacial surgery 16: 347-52–4.

291. Arnaud, Eric, Pedro Meneses, Elizabeth Lajeunie, John A Thorne, Daniel Marchac, and Dominique Renier. 2002. Postoperative mental and morphological outcome for nonsyndromic brachycephaly. Plastic and reconstructive surgery 110: 6–12; discussion 13.

272 292. Woods, Roger H, Ehtesham Ul-Haq, Andrew O M Wilkie, Jayaratnam Jayamohan, Peter G Richards, David Johnson, Tracy Lester, and Steven A Wall. 2009. Reoperation for intracranial hypertension in TWIST1-confirmed Saethre-Chotzen syndrome: a 15-year review. Plastic and reconstructive surgery 123. Europe PMC Funders: 1801–10. doi:10.1097/PRS.0b013e3181a3f391.

293. Johnson, D, S W Horsley, D M Moloney, M Oldridge, S R Twigg, S Walsh, M Barrow, et al. 1998. A comprehensive screen for TWIST mutations in patients with craniosynostosis identifies a new microdeletion syndrome of chromosome band 7p21.1. American journal of human genetics 63. Elsevier: 1282–93. doi:10.1086/302122.

294. Kress, Wolfram, Christian Schropp, Gabriele Lieb, Birgit Petersen, Maria Büsse- Ratzka, Jürgen Kunz, Edeltraut Reinhart, et al. 2005. Saethre–Chotzen syndrome caused by TWIST 1 gene mutations: functional differentiation from Muenke coronal synostosis syndrome. European Journal of Human Genetics 14: 39–48. doi:10.1038/sj.ejhg.5201507.

295. Holmes, Greg. 2012. The role of vertebrate models in understanding craniosynostosis. Child’s Nervous System 28. Springer-Verlag: 1471–1481. doi:10.1007/s00381-012- 1844-3.

296. Holmes, Greg. 2012. Mouse models of Apert syndrome. Child’s Nervous System 28. Springer-Verlag: 1505–1510. doi:10.1007/s00381-012-1872-z.

297. Johnson, D, S Iseki, A O Wilkie, and G M Morriss-Kay. 2000. Expression patterns of Twist and Fgfr1, -2 and -3 in the developing mouse coronal suture suggest a key role for twist in suture initiation and biogenesis. Mechanisms of development 91: 341–5.

298. Rice, David P. C., Elaine C. Connor, Jacqueline M. Veltmaat, Eva Lana-Elola, Lotta Veistinen, Yukiho Tanimoto, Saverio Bellusci, and Ritva Rice. 2010. Gli3Xt-J/Xt-J mice exhibit lambdoid suture craniosynostosis which results from altered osteoprogenitor proliferation and differentiation. Human molecular genetics 19: 3457– 67. doi:10.1093/hmg/ddq258.

299. Perlyn, Chad A, Gillian Morriss-Kay, Tron Darvann, Marissa Tenenbaum, and David M Ornitz. 2006. A model for the pharmacological treatment of crouzon syndrome. Neurosurgery 59. NIH Public Access: 210-5-5. doi:10.1227/01.NEU.0000224323.53866.1E.

300. Shukla, Vivek, Xavier Coumoul, Rui-Hong Wang, Hyun-Seok Kim, and Chu-Xia Deng. 2007. RNA interference and inhibition of MEK-ERK signaling prevent abnormal skeletal phenotypes in a mouse model of craniosynostosis. Nature Genetics 39: 1145–1150. doi:10.1038/ng2096.

273 301. Yin, Liangjun, Xiaolan Du, Cuiling Li, Xiaoling Xu, Zhi Chen, Nan Su, Ling Zhao, et al. 2008. A Pro253Arg mutation in fibroblast growth factor receptor 2 (Fgfr2) causes skeleton malformation mimicking human Apert syndrome by affecting both chondrogenesis and osteogenesis. Bone 42: 631–643. doi:10.1016/j.bone.2007.11.019.

302. Wang, Yingli, Xueyan Zhou, Kurun Oberoi, Robert Phelps, Ross Couwenhoven, Miao Sun, Amélie Rezza, et al. 2012. p38 Inhibition ameliorates skin and skull abnormalities in Fgfr2 Beare-Stevenson mice. The Journal of clinical investigation 122. American Society for Clinical Investigation: 2153–64. doi:10.1172/JCI62644.

303. Motch Perrine, Susan M, Theodore M Cole, Neus Martínez-Abadías, Kristina Aldridge, Ethylin Jabs, and Joan T Richtsmeier. 2014. Craniofacial divergence by distinct prenatal growth patterns in Fgfr2 mutant mice. BMC Developmental Biology 14: 8. doi:10.1186/1471-213X-14-8.

304. Percival, Christopher J., Yuan Huang, Ethylin Wang Jabs, Runze Li, and Joan T. Richtsmeier. 2014. Embryonic craniofacial bone volume and bone mineral density in Fgfr2 +/P253R and nonmutant mice. Developmental Dynamics 243: 541–551. doi:10.1002/dvdy.24095.

305. De Moerlooze, Laurence, and Clive Dickson. 1997. Skeletal disorders associated with fibroblast growth factor receptor mutatios. Current Opinion in Genetics & Development 7: 378–385. doi:10.1016/S0959-437X(97)80152-9.

306. Ratisoontorn, Chootima, Gao-Feng Fan, Kerry McEntee, and Hyun-Duck Nah. 2003. Activating (P253R, C278F) and dominant negative mutations of FGFR2: differential effects on calvarial bone cell proliferation, differentiation, and mineralization. Connective tissue research 44 Suppl 1: 292–7.

307. Boyadjiev, SA, and International Craniosynostosis Consortium. 2007. Genetic analysis of non-syndromic craniosynostosis. Orthodontics & Craniofacial Research 10: 129– 137. doi:10.1111/j.1601-6343.2007.00393.x.

308. Jabs, E W. 1998. Toward understanding the pathogenesis of craniosynostosis through clinical and molecular correlates. Clinical genetics 53: 79–86.

309. Mai, S, K Wei, A Flenniken, S L Adamson, J Rossant, J E Aubin, and S-G Gong. 2010. The missense mutation W290R in Fgfr2 causes developmental defects from aberrant IIIb and IIIc signaling. Developmental dynamics : an official publication of the American Association of Anatomists 239: 1888–900. doi:10.1002/dvdy.22314.

310. Snyder-Warwick, A. K., C. A. Perlyn, J. Pan, K. Yu, L. Zhang, and D. M. Ornitz. 2010. Analysis of a gain-of-function FGFR2 Crouzon mutation provides evidence of loss of function activity in the etiology of cleft palate. Proceedings of the National Academy of Sciences 107: 2515–2520. doi:10.1073/pnas.0913985107.

274 311. Hung, Irene H., Gary C. Schoenwolf, Mark Lewandoski, and David M. Ornitz. 2016. A combined series of Fgf9 and Fgf18 mutant alleles identifies unique and redundant roles in skeletal development. Developmental Biology 411: 72–84. doi:10.1016/j.ydbio.2016.01.008.

312. Naganawa, T., L. Xiao, E. Abogunde, T. Sobue, I. Kalajzic, M. Sabbieti, D. Agas, and M.M. Hurley. 2006. In vivo and in vitro comparison of the effects of FGF-2 null and haplo-insufficiency on bone formation in mice. Biochemical and Biophysical Research Communications 339: 490–498. doi:10.1016/j.bbrc.2005.10.215.

313. Montero, Aldemar, Yosuke Okada, Masato Tomita, Masako Ito, Hiroshi Tsurukami, Toshitaka Nakamura, Thomas Doetschman, J. Douglas Coffin, and Marja M. Hurley. 2000. Disruption of the fibroblast growth factor-2 gene results in decreased bone mass and bone formation. Journal of Clinical Investigation 105: 1085–1093. doi:10.1172/JCI8641.

314. Coffin, J D, R Z Florkiewicz, J Neumann, T Mort-Hopkins, G W Dorn, P Lightfoot, R German, P N Howles, A Kier, and B A O’Toole. 1995. Abnormal bone growth and selective translational regulation in basic fibroblast growth factor (FGF-2) transgenic mice. Molecular biology of the cell 6: 1861–73.

315. Liu, Zhonghao, Kory J. Lavine, Irene H. Hung, and David M. Ornitz. 2007. FGF18 is required for early chondrocyte proliferation, hypertrophy and vascular invasion of the growth plate. Developmental Biology 302: 80–91. doi:10.1016/j.ydbio.2006.08.071.

316. Ornitz, David M, and Nobuyuki Itoh. 2015. The Fibroblast Growth Factor signaling pathway. Wiley interdisciplinary reviews. Developmental biology 4. Wiley-Blackwell: 215–66. doi:10.1002/wdev.176.

317. Zhang, Xiuqin, Omar A Ibrahimi, Shaun K Olsen, Hisashi Umemori, Moosa Mohammadi, and David M Ornitz. 2006. Receptor specificity of the fibroblast growth factor family. The complete mammalian FGF family. The Journal of biological chemistry 281. American Society for Biochemistry and Molecular Biology: 15694– 700. doi:10.1074/jbc.M601252200.

318. Xu, J, M Nakahara, J W Crabb, E Shi, Y Matuo, M Fraser, M Kan, J Hou, and W L McKeehan. 1992. Expression and immunochemical analysis of rat and human fibroblast growth factor receptor (flg) isoforms. The Journal of biological chemistry 267: 17792–803.

319. Orr-Urtreger, Avi, Mark T. Bedford, Tatjana Burakova, Esther Arman, Yitzhak Zimmer, Avner Yayon, David Givol, and Peter Lonai. 1993. Developmental Localization of the Splicing Alternatives of Fibroblast Growth Factor Receptor-2 (FGFR2). Developmental Biology 158: 475–486. doi:10.1006/dbio.1993.1205.

275 320. Olsen, Shaun K, James Y H Li, Carrie Bromleigh, Anna V Eliseenkova, Omar A Ibrahimi, Zhimin Lao, Fuming Zhang, Robert J Linhardt, Alexandra L Joyner, and Moosa Mohammadi. 2006. Structural basis by which alternative splicing modulates the organizer activity of FGF8 in the brain. Genes & development 20. Cold Spring Harbor Laboratory Press: 185–98. doi:10.1101/gad.1365406.

321. Choe, Youngshik, Julie A. Siegenthaler, and Samuel J. Pleasure. 2012. A Cascade of Morphogenic Signaling Initiated by the Meninges Controls Corpus Callosum Formation. Neuron 73: 698–712. doi:10.1016/j.neuron.2011.11.036.

322. Mayne, Richard. 1989. Cartilage collagens. What Is Their Function, and Are They Involved in Articular Disease? Arthritis & Rheumatism 32. John Wiley & Sons, Inc.: 241–246. doi:10.1002/anr.1780320302.

323. Koch, M., F. Laub, P. Zhou, R. A. Hahn, S. Tanaka, R. E. Burgeson, D. R. Gerecke, F. Ramirez, and M. K. Gordon. 2003. Collagen XXIV, a Vertebrate Fibrillar Collagen with Structural Features of Invertebrate Collagens: SELECTIVE EXPRESSION IN DEVELOPING CORNEA AND BONE. Journal of Biological Chemistry 278: 43236– 43244. doi:10.1074/jbc.M302112200.

324. Gu, J, Y Lu, F Li, L Qiao, Q Wang, N Li, J A Borgia, Y Deng, G Lei, and Q Zheng. 2014. Identification and characterization of the novel Col10a1 regulatory mechanism during chondrocyte hypertrophic differentiation. Cell Death and Disease 5: e1469. doi:10.1038/cddis.2014.444.

325. Lefebvre, V, and B de Crombrugghe. 1998. Toward understanding SOX9 function in chondrocyte differentiation. Matrix biology : journal of the International Society for Matrix Biology 16: 529–40.

326. Akiyama, Haruhiko, Marie-Christine Chaboissier, James F Martin, Andreas Schedl, and Benoit de Crombrugghe. 2002. The transcription factor Sox9 has essential roles in successive steps of the chondrocyte differentiation pathway and is required for expression of Sox5 and Sox6. Genes & development 16. Cold Spring Harbor Laboratory Press: 2813–28. doi:10.1101/gad.1017802.

327. Lefebvre, V, W Huang, V R Harley, P N Goodfellow, and B de Crombrugghe. 1997. SOX9 is a potent activator of the chondrocyte-specific enhancer of the pro alpha1(II) collagen gene. Molecular and cellular biology 17: 2336–46.

328. Lee, M S, G N Lowe, D D Strong, J E Wergedal, and C A Glackin. 1999. TWIST, a basic helix-loop-helix transcription factor, can regulate the human osteogenic lineage. Journal of cellular biochemistry 75: 566–77.

329. Lai, Wen-Tzu, Veena Krishnappa, and Donald G Phinney. 2011. Fibroblast growth factor 2 (Fgf2) inhibits differentiation of mesenchymal stem cells by inducing Twist2 and Spry4, blocking extracellular regulated kinase activation, and altering Fgf receptor expression levels. Stem cells (Dayton, Ohio) 29. NIH Public Access: 1102–11. doi:10.1002/stem.661.

276 330. Bialek, Peter, Britt Kern, Xiangli Yang, Marijke Schrock, Drazen Sosic, Nancy Hong, Hua Wu, et al. 2004. A twist code determines the onset of osteoblast differentiation. Developmental cell 6: 423–35. doi:10.1016/S1534-5807(04)00058-9.

331. Goodnough, L. Henry, Andrew T. Chang, Charles Treloar, Jing Yang, Peter C. Scacheri, and Radhika P. Atit. 2012. Twist1 mediates repression of chondrogenesis by β-catenin to promote cranial bone progenitor specification. Development 139.

332. Lustig, Barbara, Boris Jerchow, Martin Sachs, Sigrid Weiler, Torsten Pietsch, Uwe Karsten, Marc van de Wetering, et al. 2002. Negative feedback loop of Wnt signaling through upregulation of conductin/axin2 in colorectal and liver tumors. Molecular and cellular biology 22: 1184–93.

333. Eswarakumar, Vereragavan P., Efrat Monsonego-Ornan, Mark Pines, Ileana Antonopoulou, Gillian M. Morriss-Kay, and Peter Lonai. 2002. The IIIc alternative of Fgfr2 is a positive regulator of bone formation. Development 129.

334. Govindarajan, Venkatesh, and Paul A Overbeek. 2006. FGF9 can induce endochondral ossification in cranial mesenchyme. BMC developmental biology 6: 7. doi:10.1186/1471-213X-6-7.

335. Wang, Y., Ran Xiao, Fan Yang, Baktiar O Karim, Anthony J Iacovelli, Juanliang Cai, Charles P Lerner, et al. 2005. Abnormalities in cartilage and bone development in the Apert syndrome FGFR2+/S252W mouse. Development 132: 3537–3548. doi:10.1242/dev.01914.

336. Day, Timothy F., Xizhi Guo, Lisa Garrett-Beal, and Yingzi Yang. 2005. Wnt/β- Catenin Signaling in Mesenchymal Progenitors Controls Osteoblast and Chondrocyte Differentiation during Vertebrate Skeletogenesis. Developmental Cell 8: 739–750.

337. Akiyama, Haruhiko, Jon P Lyons, Yuko Mori-Akiyama, Xiaohong Yang, Ren Zhang, Zhaoping Zhang, Jian Min Deng, et al. 2004. Interactions between Sox9 and beta- catenin control chondrocyte differentiation. Genes & development 18. Cold Spring Harbor Laboratory Press: 1072–87. doi:10.1101/gad.1171104.

338. Yano, Fumiko, Fumitaka Kugimiya, Shinsuke Ohba, Toshiyuki Ikeda, Hirotaka Chikuda, Toru Ogasawara, Naoshi Ogata, et al. 2005. The canonical Wnt signaling pathway promotes chondrocyte differentiation in a Sox9-dependent manner. Biochemical and Biophysical Research Communications 333: 1300–1308. doi:10.1016/j.bbrc.2005.06.041.

339. Usami, Yu, Aruni T Gunawardena, Masahiro Iwamoto, and Motomi Enomoto- Iwamoto. 2016. Wnt signaling in cartilage development and diseases: lessons from animal studies. Laboratory Investigation 96. Nature Publishing Group: 186–196. doi:10.1038/labinvest.2015.142.

277 340. Yates, Karen E., Sonya Shortkroff, and Richard G. Reish. 2005. Wnt Influence on Chondrocyte Differentiation and Cartilage Function. DNA and Cell Biology 24: 446– 457. doi:10.1089/dna.2005.24.446.

341. Regard, Jean B, Zhendong Zhong, Bart O Williams, and Yingzi Yang. 2012. Wnt signaling in bone development and disease: making stronger bone with Wnts. Cold Spring Harbor perspectives in biology 4. Cold Spring Harbor Laboratory Press: a007997. doi:10.1101/cshperspect.a007997.

342. Ambrosetti, D., G. Holmes, A. Mansukhani, and C. Basilico. 2008. Fibroblast Growth Factor Signaling Uses Multiple Mechanisms To Inhibit Wnt-Induced Transcription in Osteoblasts. Molecular and Cellular Biology 28: 4759–4771. doi:10.1128/MCB.01849-07.

343. Behr, Björn, Michael T. Longaker, and Natalina Quarto. 2010. Differential activation of canonical Wnt signaling determines cranial sutures fate: A novel mechanism for sagittal suture craniosynostosis. Developmental Biology 344: 922–940. doi:10.1016/j.ydbio.2010.06.009.

344. Ornitz, David M. 2000. FGFs, heparan sulfate and FGFRs: complex interactions essential for development. BioEssays 22: 108–112. doi:10.1002/(SICI)1521- 1878(200002)22:2<108::AID-BIES2>3.0.CO;2-M.

345. Lin, Congxing, Yan Yin, Sheila M. Bell, G. Michael Veith, Hong Chen, Sung-Ho Huh, David M. Ornitz, and Liang Ma. 2013. Delineating a Conserved Genetic Cassette Promoting Outgrowth of Body Appendages. Edited by David R. Beier. PLoS Genetics 9. Public Library of Science: e1003231. doi:10.1371/journal.pgen.1003231.

346. Lu, P., George Minowada, and Gail R Martin. 2006. Increasing Fgf4 expression in the mouse limb bud causes polysyndactyly and rescues the skeletal defects that result from loss of Fgf8 function. Development 133: 33–42. doi:10.1242/dev.02172.

347. Capecchi, Mario R., and Anne M. Moon. 2000. Fgf8 is required for outgrowth and patterning of the limbs. Nature Genetics 26. Nature Publishing Group: 455–459. doi:10.1038/82601.

348. Arnold, Michael A., Yuri Kim, Michael P. Czubryt, Dillon Phan, John McAnally, Xiaoxia Qi, John M. Shelton, James A. Richardson, Rhonda Bassel-Duby, and Eric N. Olson. 2007. MEF2C Transcription Factor Controls Chondrocyte Hypertrophy and Bone Development. Developmental Cell 12: 377–389. doi:10.1016/j.devcel.2007.02.004.

349. Senarath-Yapa, Kshemendra, Michael T Chung, Adrian McArdle, Victor W Wong, Natalina Quarto, Michael T Longaker, and Derrick C Wan. 2012. Craniosynostosis: molecular pathways and future pharmacologic therapy. Organogenesis 8. Taylor & Francis: 103–13. doi:10.4161/org.23307.

278 350. Eames, B. Frank, and Jill A. Helms. 2004. Conserved molecular program regulating cranial and appendicular skeletogenesis. Developmental Dynamics 231. Wiley Subscription Services, Inc., A Wiley Company: 4–13. doi:10.1002/dvdy.20134.

351. Eames, B. Frank, Paul T. Sharpe, and Jill A. Helms. 2004. Hierarchy revealed in the specification of three skeletal fates by Sox9 and Runx2. Developmental Biology 274: 188–200. doi:10.1016/j.ydbio.2004.07.006.

352. Hu, Hongliang, Matthew J Hilton, Xiaolin Tu, Kai Yu, David M Ornitz, and Fanxin Long. 2005. Sequential roles of Hedgehog and Wnt signaling in osteoblast development. Development (Cambridge, England) 132: 49–60. doi:10.1242/dev.01564.

353. Mansukhani, Alka, Davide Ambrosetti, Greg Holmes, Lizbeth Cornivelli, and Claudio Basilico. 2005. Sox2 induction by FGF and FGFR2 activating mutations inhibits Wnt signaling and osteoblast differentiation. The Journal of cell biology 168. The Rockefeller University Press: 1065–76. doi:10.1083/jcb.200409182.

354. Zhang, M., S. Xuan, M. L. Bouxsein, D. von Stechow, N. Akeno, M. C. Faugere, H. Malluche, et al. 2002. Osteoblast-specific Knockout of the Insulin-like Growth Factor (IGF) Receptor Gene Reveals an Essential Role of IGF Signaling in Bone Matrix Mineralization. Journal of Biological Chemistry 277: 44005–44012. doi:10.1074/jbc.M208265200.

355. Ovchinnikov, D A, J M Deng, G Ogunrinu, and R R Behringer. 2000. Col2a1-directed expression of Cre recombinase in differentiating chondrocytes in transgenic mice. Genesis (New York, N.Y. : 2000) 26: 145–6.

356. Brinkley, James F, Shannon Fisher, Matthew P Harris, Greg Holmes, Joan E Hooper, Ethylin Wang Jabs, Kenneth L Jones, et al. 2016. The FaceBase Consortium: a comprehensive resource for craniofacial researchers. Development 143: 2677–2688. doi:10.1242/dev.135434.

357. Teven, Chad M., Evan M. Farina, Jane Rivas, and Russell R. Reid. 2014. Fibroblast growth factor (FGF) signaling in development and skeletal diseases. Genes & Diseases 1: 199–213. doi:10.1016/j.gendis.2014.09.005.

358. Hu, Diane, Ralph S Marcucio, and Jill A Helms. 2003. A zone of frontonasal ectoderm regulates patterning and growth in the face. Development (Cambridge, England) 130: 1749–58.

359. Szabo-Rogers, Heather L., Poongodi Geetha-Loganathan, Cheryl J. Whiting, Suresh Nimmagadda, Katherine Fu, and Joy M. Richman. 2008. Novel skeletogenic patterning roles for the olfactory pit. Development 136.

360. Marcucio, Ralph S., Dwight R. Cordero, Diane Hu, and Jill A. Helms. 2005. Molecular interactions coordinating the development of the forebrain and face. Developmental Biology 284: 48–61. doi:10.1016/j.ydbio.2005.04.030.

279 361. Hu, Diane, and Ralph S. Marcucio. 2008. A SHH-responsive signaling center in the forebrain regulates craniofacial morphogenesis via the facial ectoderm. Development 136.

362. Ming, Jeffrey E, Erich Roessler, and Maximilian Muenke. 1998. Human developmental disorders and the Sonic hedgehog pathway. Molecular Medicine Today 4. Elsevier Current Trends: 343–349. doi:10.1016/S1357-4310(98)01299-4.

363. Bouldin, Cortney M, Amel Gritli-Linde, Sohyun Ahn, and Brian D Harfe. 2010. Shh pathway activation is present and required within the vertebrate limb bud apical ectodermal ridge for normal autopod patterning. Proceedings of the National Academy of Sciences of the United States of America 107. National Academy of Sciences: 5489–94. doi:10.1073/pnas.0912818107.

364. Xie, Jingwu, Maximilien Murone, Shiuh-Ming Luoh, Anne Ryan, Qimin Gu, Chaohui Zhang, Jeannette M Bonifas, et al. 1998. Activating Smoothened mutations in sporadic basal-cell carcinoma. Nature 391: 90–92.

365. Mao, Junhao, Keith L Ligon, Elena Y Rakhlin, Sarah P Thayer, Roderick T Bronson, David Rowitch, and Andrew P McMahon. 2006. A novel somatic mouse model to survey tumorigenic potential applied to the Hedgehog pathway. Cancer research 66. NIH Public Access: 10171–8. doi:10.1158/0008-5472.CAN-06-0657.

366. te Welscher, Pascal, Marian Fernandez-Teran, Marian A Ros, and Rolf Zeller. 2002. Mutual genetic antagonism involving GLI3 and dHAND prepatterns the vertebrate limb bud mesenchyme prior to SHH signaling. Genes & development 16. Cold Spring Harbor Laboratory Press: 421–6. doi:10.1101/gad.219202.

367. Matsubara, Haruka, Daisuke Saito, Gembu Abe, Hitoshi Yokoyama, Takayuki Suzuki, and Koji Tamura. 2017. Upstream regulation for initiation of restricted Shh expression in the chick limb bud. Developmental Dynamics 246: 417–430. doi:10.1002/dvdy.24493.

368. Zakany, Jozsef, and Denis Duboule. 2007. The role of Hox genes during vertebrate limb development. Current Opinion in Genetics & Development 17: 359–366. doi:10.1016/j.gde.2007.05.011.

369. Candi, E, I Amelio, M Agostini, and G Melino. 2015. MicroRNAs and p63 in epithelial stemness. Cell Death and Differentiation 22. Nature Publishing Group: 12– 21. doi:10.1038/cdd.2014.113.

370. Cobourne, Martyn T, Guilherme M Xavier, Michael Depew, Louise Hagan, Jane Sealby, Zoe Webster, and Paul T Sharpe. 2009. Sonic hedgehog signalling inhibits palatogenesis and arrests tooth development in a mouse model of the nevoid basal cell carcinoma syndrome. Developmental biology 331. Elsevier: 38–49. doi:10.1016/j.ydbio.2009.04.021.

280 371. Sharma, Seema, Vipin Sharma, and Meenakshi Bothra. 2012. Frontonasal dysplasia (Median cleft face syndrome). Journal of neurosciences in rural practice 3. Medknow Publications and Media Pvt. Ltd.: 65–7. doi:10.4103/0976-3147.91947.

372. Sharma, Reena, Poojan Dogra, Kapil Malhotra, and Vivek Kaushal. 2017. Frontonasal dysplasia-a rare case report. International Journal of Research in Medical Sciences International Journal of Research in Medical Sciences Sharma R Int J Res Med Sci 55: 4640–4642. doi:10.18203/2320-6012.ijrms20174612.

373. Chen, Guiqian, Chuxia Deng, and Yi-Ping Li. 2012. TGF-β and BMP signaling in osteoblast differentiation and bone formation. International journal of biological sciences 8. Ivyspring International Publisher: 272–88. doi:10.7150/ijbs.2929.

374. Kim, Jeong Hwan, Xing Liu, Jinhua Wang, Xiang Chen, Hongyu Zhang, Stephanie H Kim, Jing Cui, et al. 2013. Wnt signaling in bone formation and its therapeutic potential for bone diseases. Therapeutic advances in musculoskeletal disease 5. SAGE Publications: 13–31. doi:10.1177/1759720X12466608.

375. Zakany, J., Marie Kmita, and Denis Duboule. 2004. A Dual Role for Hox Genes in Limb Anterior-Posterior Asymmetry. Science 304: 1669–1672. doi:10.1126/science.1096049.

376. Zhao, Xiuli, Miao Sun, Jin Zhao, J Alfonso Leyva, Hongwen Zhu, Wei Yang, Xuan Zeng, et al. 2007. Mutations in HOXD13 underlie syndactyly type V and a novel brachydactyly-syndactyly syndrome. American journal of human genetics 80. Elsevier: 361–71. doi:10.1086/511387.

377. Kuss, Pia, Pablo Villavicencio-Lorini, Florian Witte, Joachim Klose, Andrea N. Albrecht, Petra Seemann, Jochen Hecht, and Stefan Mundlos. 2008. Mutant Hoxd13 induces extra digits in a mouse model of synpolydactyly directly and by decreasing retinoic acid synthesis. Journal of Clinical Investigation 119: 146–56. doi:10.1172/JCI36851.

378. Boulet, A. M., and Mario R Capecchi. 2003. Multiple roles of Hoxa11 and Hoxd11 in the formation of the mammalian forelimb zeugopod. Development 131: 299–309. doi:10.1242/dev.00936.

379. Salsi, Valentina, Maria Alessandra Vigano, Fabienne Cocchiarella, Roberto Mantovani, and Vincenzo Zappavigna. 2008. Hoxd13 binds in vivo and regulates the expression of genes acting in key pathways for early limb and skeletal patterning. Developmental Biology 317: 497–507. doi:10.1016/j.ydbio.2008.02.048.

380. Dahn, Randall D., Marcus C. Davis, William N. Pappano, and Neil H. Shubin. 2007. Sonic hedgehog function in chondrichthyan fins and the evolution of appendage patterning. Nature 445. Nature Publishing Group: 311–314. doi:10.1038/nature05436.

281 381. Heuzé, Yann, Neus Martínez-Abadías, Jennifer M Stella, Eric Arnaud, Corinne Collet, Gemma García Fructuoso, Mariana Alamar, et al. 2014. Quantification of facial skeletal shape variation in fibroblast growth factor receptor-related craniosynostosis syndromes. Birth defects research. Part A, Clinical and molecular teratology 100. NIH Public Access: 250–9. doi:10.1002/bdra.23228.

382. Martínez-Abadías, Neus, Christopher Percival, Kristina Aldridge, Cheryl A. Hill, Timothy Ryan, Satama Sirivunnabood, Yingli Wang, Ethylin Wang Jabs, and Joan T. Richtsmeier. 2010. Beyond the closed suture in apert syndrome mouse models: Evidence of primary effects of FGFR2 signaling on facial shape at birth. Developmental Dynamics 239: 3058–3071. doi:10.1002/dvdy.22414.

383. Maruyama, Takamitsu, Anthony J Mirando, Chu-Xia Deng, and Wei Hsu. 2010. The balance of WNT and FGF signaling influences mesenchymal stem cell fate during skeletal development. Science signaling 3: ra40. doi:10.1126/scisignal.2000727.

384. Behr, Björn, Michael T Longaker, and Natalina Quarto. 2011. Craniosynostosis of coronal suture in twist1 mice occurs through endochondral ossification recapitulating the physiological closure of posterior frontal suture. Frontiers in physiology 2. Frontiers Media SA: 37. doi:10.3389/fphys.2011.00037.

385. Maruyama, Takamitsu, Jaeim Jeong, Tzong-Jen Sheu, and Wei Hsu. 2016. Stem cells of the suture mesenchyme in craniofacial bone development, repair and regeneration. Nature Communications 7. Nature Publishing Group: 10526. doi:10.1038/ncomms10526.

386. Towers, Matthew, and Cheryll Tickle. 2009. Growing models of vertebrate limb development. Development (Cambridge, England) 136: 179–190. doi:10.1242/dev.024158.

282 APPENDIX A

Recipes and Abbreviations

KOH - Potassium Hydroxide

PBS - Phosphate Buffered Saline

PBT - Phosphate Buffered Saline with 1% Tween-20

SSC - Saline-Sodium Citrate

SDS – Sodium Dodecyl Sulfate

Hybridization Mixture for Whole Mount ISH 50% Formamide, 1.3x SSC (pH 4.5), 50 µg/ml Yeast tRNA, 100 µg/ml Heparin, 0.2% Tween 20, 0.5% CHAPS, 5mM EDTA

Hybridization Buffer for Section ISH 50% formamide, 5x SSC (pH4.5), 50 µg/ml Yeast tRNA, 1% SDS, 50 µg/ml heparin

MABT (5x) 58 g Maleic Acid, 43.5g NaCl, 55g Tween-20, water to 1 liter. pH to 7.5 with NaOH

Alkaline Phosphatase Buffer

100mM Tris (pH 9.5), 50mM MgCl2, 100mM NaCl, 0.1% Tween 20

Typical PCR reaction using Qiagen Taq

Q buffer 5µl 10x buffer 2.5 µl Mg2+ 2 µl dNTPs (10 mM) 0.5 µl Primer 1 (10µΜ) 1 µl Primer2 (10µΜ) 1 µl Τaq 0.1 µl DNA 0.5-2µl Water to 25 µl total volume

283