Disorders of sex development: Genetic analysis and development of a novel in vitro cell model

Ingrid May Knarston

B.Sc. (Hons.)

Submitted in total fulfilment of the requirements of the degree of Doctor of Philosophy

October 2018

Department of Paediatrics

The University of Melbourne

2 Abstract

Disorders/Differences of Sex Development (DSDs) are conditions where the chromosomal, anatomical or gonadal sex is atypical. DSDs are caused by a breakdown in the molecular pathways controlling development of the reproductive organs, such as ovarian/testicular differentiation. These conditions can carry a number of clinical complications such as an increased risk of gonadal cancer, infertility and psychosocial consequences. Importantly, the underlying genetic cause is still unknown in 60% of DSD patients, meaning clinical care is severely compromised.

In the first part of this thesis, I studied a cohort of 34 patients with 46,XX (ovo)testicular DSDs. In these individuals, the testicular differentiation pathway is activated in 46,XX genetic females, resulting in the formation of testes or ovotestes. The cohort was studied using massively parallel sequencing and PCR-based approaches. This identified diagnostic findings in nine patients in two known DSD genes (NR5A1 and SOX9), as well as variants in candidate DSD genes (EMX2, FOXL2, LGR5, RXFP2 and WNT9A). In vitro analysis of the NR5A1 variants showed how these variants repress ovarian signalling pathways and factors, sufficient to switch ovarian to testicular development. In vitro and in vivo analyses of three of the candidate genes (EMX2, LGR5 and RXFP2) indicated that they are likely benign variants that don’t contribute to the phenotype. Ongoing studies of two further candidate genes (FOXL2 and WNT9A) will establish their potential role in these DSD phenotypes.

In the second part of the thesis, I aimed to develop an improved in vitro model for functionally analysing DSD gene variants. Several recent studies have differentiated human induced pluripotent stem cells (iPSCs) into many different tissues, which can be used as human- and tissue-specific disease models. I developed a protocol to differentiate human iPSCs into testis- like lineages. In this step-wise protocol, cells are directed through the developmental stages that give rise to the embryonic testis. Gene expression profiling has shown that at day 10-12 of iPSC differentiation, cells reach a bipotential gonad-like stage and by day 15 testis-like lineages are induced. This protocol will continue to be optimised, yet already I have shown its promising utility to study novel DSD genes.

In summary, genetic analysis of a DSD cohort revealed diagnoses for a number of patients; these findings will likely improve their clinical management. It has also provided information on the most suitable genetic testing approach for 46,XX (ovo)testicular DSDs, a phenotypic group traditionally challenging to diagnose. Further, I showed functional insights into the molecular

pathogenesis underlying NR5A1-mediated 46,XX (ovo)testicular DSD. Finally, development of a stem cell-based model of the human testis will help us to establish how novel DSD genes and variants affect human gonad development.

ii Declaration

This is to certify that:

i. This thesis comprises only my original work towards the PhD except where indicated in the Preface, ii. Due acknowledgement has been made in the text to all other material used, iii. This thesis is less than 100,000 words in length, exclusive of tables, maps, bibliographies and appendices

Ingrid Knarston:

iii Preface

Work carried out in collaboration with others as part of this thesis:

Chapter 2

 DNA and clinical information from all patients used in the studies presented here were collected from national and international collaborating clinicians.  MLPA reactions and data analysis was performed in collaboration with Dr. Thomas Ohnesorg (Reproductive Development group, MCRI). Follow up CGH arrays were performed and analysed by Jocelyn van den Bergen (Reproductive Development group, MCRI).  Preparation of DNA libraries for MPS was performed by Gorjana Robevska and Jocelyn van den Bergen (Reproductive Development group, MCRI) or at sequencing facilities (Australian Genomics Research Facility and Garvan Institute).  The bioinformatic analysis of MPS data was undertaken in collaboration with Dr. Simon Sadedin (MCRI), Katrina Bell (MCRI) and Ben Lundie (Garvan Institute).  The in vitro assay testing mutant RXFP2 function was performed by Prof. Ross Bathgate’s group (Florey Institute).  The Emx2 and Wnt9a CRISPR mutant mice were generated by Dr. Liang Zhao (Institute of Molecular Bioscience, The University of Queensland) and characterised by Dr. Ella Thomson (Institute of Molecular Bioscience, The University of Queensland) and Dr. Anthony Bird (Hudson Institute for Medical Research, Monash University) respectively.

Chapter 3

 The in silico modelling of NR5A1 variant proteins and immunofluorescence staining for NR5A1 protein in COS-7 cells was performed by Gorjana Robevska (Reproductive Development group, MCRI).

I conducted all other work, comprising 90% of the thesis.

iv Publications

Ingrid Knarston*, Katie L. Ayers*, Andrew H. Sinclair. Molecular mechanisms associated with 46, XX disorders of sex development. Clinical Science 2016 March 01; 130 (6): 421-432.

Corresponding chapter in the thesis: Chapter 1

Contribution to manuscript: I was involved in all aspects of this manuscript. I conceptualised and designed the investigation with KA, conducted the literature analysis and co-wrote the paper with KA.

Brittany Croft*, Thomas Ohnesorg*, Josephine Bowles, Katie Ayers, Jacky Hewitt, Jacqueline Tan, Vincent Corbin, Emanuele Pelosi, Jocelyn van den Bergen, Alexander Quinn, Rajini Sreenivisan, Ingrid Knarston, Gorjana Robevska, Dung Vu Chi, John Hutson, Vincent Harley, Peter Koopman and Andrew Sinclair. Human sex reversal is caused by duplication or deletion of core enhancers upstream of SOX9. Nature Communications 2018 Dec 14; 9 (1): 5319.

Corresponding chapter in the thesis: Chapter 2

Contribution to manuscript: I was involved in the identification of duplications in the upstream region of SOX9 in two 46,XX DSD patients.

Ingrid M. Knarston*, Gorjana Robevska*, Jocelyn A. van den Bergen, Stefanie Eggers, Brittany Croft, Jason Yates, Remko Hersmus, Leendert H.J. Looijenga, Fergus J. Cameron, Klaus Monhike, Katie L. Ayers* and Andrew H. Sinclair*. NR5A1 gene variants repress the ovarian-specific WNT signalling pathway in 46,XX Disorders of Sex Development patients. Human Mutation 2019 Feb; 40 (2): 207-216.

Corresponding chapter in thesis: Chapter 3

Contribution to manuscript: I was involved in all aspects of this manuscript. I conceptualised and designed the study with GR and KLA. Experimental work was performed in collaboration with GR and I and co-wrote the paper with GR and KLA.

v Acknowledgements

I would like to thank each of my supervisors for their support, Dr. Katie Ayers, Prof. Andrew Sinclair, Prof. Melissa Little, Dr. Alex Combes and Dr. Stefanie Eggers. My principal supervisors Katie Ayers and Andrew Sinclair have shaped a PhD project that has been incredibly exciting and challenging. Katie has been so generous with her time in developing my skills as a researcher and providing a constant source of advice and guidance. I am also very grateful for her encouragement to reach outside my comfort zone in so many instances. I am incredibly grateful for the opportunities and guidance that Andrew has provided me, particularly in shaping a PhD project that allowed me to gain experience in both the genomics and iPSC fields. I would like to thank both Melissa Little and Alex Combes for their support in my iPSC project; our discussions have taught me so much about developmental biology and their questions have been really valuable in directing this work. Finally, I would like to thank Stefanie Eggers for sharing her knowledge on DSD and genetic analyses.

I would like to thank the entire Reproductive Development group for being such a welcoming and supportive group, Prof. Andrew Sinclair, Katie Griffin, Dr. Katie Ayers, Jocelyn van den Bergen, Gorjana Robevska, Dr. Elena Tucker, Dr. Rajini Sreenivasan, Brittany Croft, Dr. Aurore Bouty, Dr. Thomas Ohnesorg and Chloe Hanna. It has been a privilege to work alongside and learn from such a talented team of people. In particular I would like to thank Gorjana Robevska for helping me with so many aspects of this project and teaching me so many skills, as well as for her constant support and friendship over the years. I would also like to thank Jocelyn van den Bergen for being a constant source of helpful advice and such a big support.

I would also like to thank the Kidney group at MCRI for allowing me to learn about iPSC and organoid modelling alongside them. I am very grateful to Irene Ghobrial and Pei Xuan Er for sharing so much of their knowledge and time training me in iPSC and organoid culture. I would also like to thank Dr. Santhosh Kumar for his valuable advice on organoid culture and Dr. Minoru Takasato for his guidance in the early stages of this project.

I would like to thank my advisory committee, including Assoc. Prof. Shireen Lamande and Dr. Tiong Tan, for their input on the development of my project. Also at MCRI, I would like to thank

vi the Animal house staff for their help with my mouse work and Matt Burton for sharing his knowledge on confocal imaging.

Outside of MCRI, I would like to thank all members of the Australian DSD genetics program for the annual retreats that provided a fantastic forum to share ideas and get feedback on my project. Particularly I would like to thank Dr. Liang Zhao, Dr. Ella Thomson and Dr. Anthony Bird for their work on the mutant mouse models. I would also like to thank our collaborators Prof. Ross Bathgate (Florey Institute), Dr. Simon Sadedin (MCRI), Katrina Bell (MCRI) and Ben Lundie (Garvan Institute), for each of their contributions to this work.

I would like to thank all of the patients and families who have been involved in this study, as well as all of the collaborating clinicians. This research would not have been possible without their involvement. This PhD research was supported by generous funding from the University of Melbourne (RTS scholarship), Murdoch Children’s Research Institute (PhD Top Up Scholarship, Internal grants scheme, Travel scholarship) and the National Health and Medical Research Council (Program grant number 546517).

Finally I would like to thank my family and friends for their support and encouragement over the last four years. In particular, I am hugely grateful to my parents, whose unwavering support has helped me each step of the way.

vii Table of Contents

Chapter 1: Introduction ...... 1 1.1 Mammalian sex differentiation ...... 2 1.1.1 Origins of the mammalian gonad ...... 2 1.1.2 Regulatory pathways controlling gonadal differentiation ...... 2 1.1.2.1 The bipotential gonad ...... 2 1.1.2.2 Ovarian differentiation ...... 5 1.1.2.3 Testis differentiation ...... 8 1.1.2.4 Mutual repression between ovarian and testicular pathways ...... 11 1.2 46,XX Disorders/Differences of Sex Development ...... 12 1.2.1 46,XX testicular DSD ...... 12 1.2.2 46,XX ovotesticular DSD ...... 15 1.2.3 46,XX ovarian dysgenesis ...... 17 1.2.4 Identifying genetic factors underlying Disorders of Sex Development ...... 21 1.3 In vitro differentiation of gonadal lineages ...... 22 1.4 Conclusion and project aims ...... 26

Chapter 2: Genetic analysis of 46,XX Disorders of Sex Development ...... 28 2.1 Introduction ...... 28 2.2 Methods ...... 29 2.2.1 Patient cohort...... 29 2.2.2 MLPA analysis ...... 30 2.2.3 Custom DSD CGH array ...... 31 2.2.4 Targeted DSD gene panel ...... 32 2.2.5 LGR5 in vitro assay ...... 32 2.2.6 RXFP2 in vitro assay ...... 33 2.2.7 WNT9A cloning ...... 33 2.2.8 CRISPR mutant mice ...... 34 2.2.8 Whole genome sequencing analysis ...... 34 2.2.9 Routine molecular techniques ...... 35 2.2.9.1 Polymerase chain reaction ...... 35 2.2.9.2 Visualisation of nucleic acids ...... 35 2.2.9.3 Sanger sequencing ...... 36 2.2.9.4 In silico protein structure analysis ...... 36 2.2.10 Transcription factor binding site analysis...... 36

viii 2.2.11 Whole Exome Sequencing ...... 36 2.3 Results ...... 37 2.3.1 DSD patient cohort ...... 37 2.3.2 MLPA analysis for copy number variants in DSD genes ...... 43 2.3.2.1 VAMP7 ...... 43 2.3.2.2 MLPA identifies duplications in the upstream enhancer region of SOX9...... 44 2.3.3 Targeted DSD gene panel ...... 52 2.3.3.1 A candidate variant in ovarian regulator FOXL2 ...... 54 2.3.4 Candidate variants from the targeted DSD panel – LGR genes...... 57 2.3.5 Candidate genes from the targeted DSD gene panel – CRISPR mouse models for Wnt9a and Emx2 ...... 65 2.3.5.1 WNT9A ...... 65 2.3.5.2 EMX2 ...... 69 2.3.6 Whole genome sequencing ...... 75 2.3.6.1 Whole genome sequencing: CNV analysis ...... 76 2.3.6.2 Whole genome sequencing: Single nucleotide variant analysis ...... 84 2.4 Discussion ...... 87

Chapter 3: Characterisation of variants in the NR5A1 gene: implications for 46,XX Disorders of Sex Development ...... 93 3.1 Introduction ...... 93 3.2 Methods ...... 94 3.2.1 Massively Parallel Sequencing ...... 94 3.2.2 Plasmid construction ...... 94 3.2.3 Protein immunofluorescence ...... 95 3.2.4 In silico protein modelling ...... 96 3.2.5 Luciferase assays ...... 96 3.3 Results ...... 97 3.3.1 Identification of NR5A1 variants via Massively Parallel Sequencing ...... 97 3.3.2 NR5A1 protein localisation ...... 101 3.3.3 Protein modelling ...... 102 3.3.4 Luciferase assays ...... 103 3.3.5 Additional genomic variants may contribute to oligogenic inheritance ...... 108 3.4 Discussion ...... 112

Chapter 4: Differentiation of human pluripotent stem cells to gonadal lineages .... 117

ix 4.1 Introduction ...... 117 4.2 Methods ...... 119 4.2.1 Mouse dissociation and reaggregation ...... 119 4.2.2 Mouse whole mount immunofluorescence ...... 120 4.2.3 Paraffin sectioning and staining ...... 120 4.2.4 Reagent preparation ...... 121 4.2.5 Maintenance of iPSCs ...... 124 4.2.5.1 iPSC lines ...... 124 4.2.5.2 Frozen stock of human iPSCs ...... 124 4.2.6 Differentiation of iPSCs into gonadal lineages ...... 125 4.2.6.1 Plating iPSCs for differentiation (Day -1) ...... 125 4.2.6.2 Inducing the embryonic mesoderm (Days 0-7) ...... 125 4.2.7 Organoid culture ...... 125 4.2.7.1 Air liquid interface organoid culture ...... 125 4.2.7.2 Swirler culture ...... 126 4.2.8 Immunofluorescence staining of monolayer differentiations ...... 126 4.2.9 Generation of reporter iPSCs ...... 127 4.2.10 RNA extraction and Reverse Transcription ...... 127 4.2.11 Real-time Quantitative Reverse Transcriptase PCR ...... 127 4.3 Results ...... 128 4.3.1 Testing the self-organising ability of embryonic testis in mouse ...... 128 4.3.2 Identification of markers specific for gonad lineages and cell types ...... 132 4.3.3 Immunofluorescence characterisation of gonad-specific markers in human fetal testis ...... 140 4.3.4 Patterning axes of the embryonic mesoderm in iPSCs ...... 143 4.3.4.1 Testing the anterior-posterior axis of the intermediate mesoderm for gonad marker induction ...... 144 4.3.4.2 Testing the lateral-medial axis of the mesoderm for gonad marker induction ...... 150 4.3.5 Investigating the role of Hedgehog and Retinoic acid signalling in early gonad differentiation ...... 151 4.3.6 A note on iPSC differentiation media ...... 156 4.3.7 Induction of bipotential gonad lineages from iPSCs ...... 156 4.3.8 Testis pathway activation in iPSC-derived cells...... 164 4.3.9 Differentiation in 3D ...... 183 4.3.10 Differentiation in a SOX9 reporter iPSC line ...... 191

x 4.4 Discussion ...... 193

Chapter 5: Conclusions and Future Directions ...... 199 5.1 Approaching diagnoses in a 46,XX DSD cohort ...... 199 5.2 Future directions for undiagnosed and new cases ...... 201 5.3 NR5A1 variants ...... 203 5.4 Developing a stem-cell based model for DSD ...... 204 5.5 Future directions in iPSC modelling of DSD ...... 206 5.6 Conclusion ...... 207

References ...... 208

Appendices ...... 233

xi Abbreviations

Abbreviation Definition ACMG American College of Medical Genetics and Genomics BSA Bovine Serum Albumin CADD Combined Annotation Dependent Depletion CGH Comparative genomic hybridisation CHIP Chromatin Immunoprecipitation CNV Copy number variant CRISPR Clustered Regularly Interspaced Short Palindromic Repeats DAPI 4′,6-diamidino-2-phenylindole DMEM Dulbecco's Modified Eagle's medium DMSO Dimethyl sulfoxide DNA Deoxyribonucleic acid DSD Disorder of Sex Development EDTA Ethylenediaminetetraacetic acid ESC Embryonic stem cell FCS Fetal calf serum HEK Human embryonic kidney IF Immunofluorescence IGV Integrated Genome Viewer iPSC Induced pluripotent stem cell KI Knock-in KO Knock-out LOF Loss of function MEF Mouse embryonic fibroblast MPS Massively Parallel Sequencing MSC Mesenchymal stem cell OT-DSD Ovotesticular DSD PBS Phosphate-buffered saline PFA Paraformaldehyde PGCs Primordial germ cells POI Primary ovarian insufficiency PVA Polyvinyl alcohol qRT-PCR Quantitative real-time polymerase chain reaction RNA-Seq RNA MPS SEM Standard error of the mean SNP Single nucleotide polymorphism SNV Single nucleotide variant T-DSD Testicular DSD VCGS Victorian Clinical Genetic Service VUS Variant of Unknown Significance WES Whole Exome sequencing WGS Whole Genome sequencing WT Wild-type

xii List of Tables

Table 1.1. Genes associated with 46,XX Disorders of ovarian development in humans ...... 19 Table 2.1. MLPA probes targeting 14 diagnostic DSD genes ...... 30 Table 2.2. Clinical information in a 46,XX DSD cohort ...... 38 Table 2.3. Clinical presentation of patients with 46,XX (ovo)testicular DSD caused by SOX9 regulatory region duplication ...... 47 Table 2.4. Candidate variants from the targeted DSD gene panel ...... 53 Table 2.5. Candidate variants from whole genome sequencing ...... 78 Table 3.1. NR5A1 variant information ...... 97 Table 3.2. Additional genomic variants identified in 46,XX (ovo)testicular DSD cases ...... 110 Table 4.1 Primary antibodies used for immunofluorescence staining ...... 119 Table 4.2 Markers of the human fetal gonad for characterisation of gonad differentiations.. 133 Table 4.3 Growth factors involved in gonad development ...... 152

xiii List of Figures

Figure 1.1 Genes and pathways activated during the development and differentiation of the genital ridge and bipotential gonad...... 5 Figure 1.2 Molecular signalling pathways during differentiation of the gonads into testes or ovaries ...... 8 Figure 2.1 A maternally inherited VAMP7 duplication in DSD14 ...... 43 Figure 2.2 MLPA identifies SOX9 enhancer duplications in four 46,XX (ovo)testicular DSD patients ...... 46 Figure 2.3 Representation of duplications identified upstream of SOX9 in 46,XX individuals ... 51 Figure 2.4 The targeted DSD panel identifies a paternally inherited missense variant in FOXL2 in DSD34 ...... 55 Figure 2.5 A heterozygous missense variant in FOXL2 is located in the highly conserved DNA binding domain ...... 56 Figure 2.6 A heterozygous missense variant identified in LGR5 in DSD04 via the targeted DSD gene panel ...... 59 Figure 2.7 A missense variant in LGR5 is located in the highly conserved signal peptide region ...... 60 Figure 2.8 Protein localisation and expression of variant LGR5 is unaffected ...... 61 Figure 2.9 A heterozygous frameshift variant identified in RXFP2 in DSD32 via the targeted DSD gene panel ...... 63 Figure 2.10 A heterozygous frameshift variant in RXFP2 is located in the C-terminal intracellular domain ...... 64 Figure 2.11 Testing receptor signalling function of variant RXFP2 compared to wild type RXFP2 ...... 65 Figure 2.12 The targeted DSD panel identifies two missense variants in WNT9A in DSD10 ...... 67 Figure 2.13 Two heterozygous missense variants in WNT9A located in a conserved binding domain ...... 68 Figure 2.14 The targeted DSD panel identifies a missense variant in EMX2 in DSD01 ...... 70 Figure 2.15 A heterozygous missense variant in EMX2 is located in a highly conserved region of exon 1 ...... 71 Figure 2.16 Gene expression in Emx2-p.A94V+/+ mouse gonads ...... 73 Figure 2.17 Immunofluorescence staining in Emx2-p.A94V+/+ mouse gonads ...... 74 Figure 2.18 Status of the 46,XX (ovo)testicular DSD cohort (August 2018) ...... 75

xiv Figure 2.19 A heterozygous duplication at chr1: 22391601-22432700 includes a potential enhancer/repressor downstream of WNT4 ...... 82 Figure 2.20 Whole genome sequencing analysis pipeline ...... 85 Figure 3.1 Generation of a reporter vector for the human NR0B1 promoter ...... 95 Figure 3.2 IGV visualisation of the NM_004959.4(NR5A1):c.C274T;p.(Arg92Trp) variant in DSD07, DSD17 and DSD28 ...... 99 Figure 3.3 IGV visualisation of the NM_004959.4(NR5A1):c.C779T;p.(Ala260Val) variant in DSD21 ...... 100 Figure 3.4 Location of variants in NR5A1 protein and sequence conservation ...... 101 Figure 3.5 Immunofluorescence staining for NR5A1 protein in COS-7 cells ...... 102 Figure 3.6 Protein structure modelling of NR5A1 variants ...... 103 Figure 3.7 NR5A1 variants show altered function in sex differentiation pathways in vitro ..... 106 Figure 3.8 Variants in NR5A1 disrupt the ovary-promoting effects of the NR5A1/β-catenin complex...... 108 Figure 4.1 Shared developmental origins of gonad and kidney ...... 118 Figure 4.2 Embryonic mouse testis shows self-organisation potential ...... 129 Figure 4.3 Self-organisation ability of mouse testis declines after E14.5 ...... 131 Figure 4.4 Expression heatmaps of gonad markers in mouse and human fetal kidney ...... 135 Figure 4.5 Gonad marker expression profiles in the developing mouse testis and ovary ...... 137 Figure 4.6 Expression patterns of gonad markers in human fetal testis ...... 141 Figure 4.7 Patterning of the anterior-posterior and lateral-medial axes of the fetal mouse mesoderm ...... 144 Figure 4.8 Longer CHIR duration promotes gonad marker induction ...... 146 Figure 4.9 3 μM CHIR promotes induction of markers for bipotential gonad and testis ...... 148 Figure 4.10 Shifting differentiation towards the lateral plate mesoderm favours induction of gonadal lineages ...... 151 Figure 4.11 Hedgehog signalling is dispensable in induction of gonadal lineages...... 154 Figure 4.12 Retinoic acid regulates WT1 and NR0B1 during early gonadogenesis ...... 155 Figure 4.13 Differentiation of gonad-like lineages from iPSCs...... 158 Figure 4.14 Effects of FGF9 treatment on induction of testis lineages ...... 161 Figure 4.15 NT2D1 condition media does not support induction of testis lineages ...... 165 Figure 4.16 Prostaglandin D2 induces testis-specific markers ...... 168 Figure 4.17 Activation of Hedgehog signalling via SAG inhibits testis differentiation ...... 173 Figure 4.18 Prostaglandin D2 has minimal effect on testis pathway induction ...... 178

xv Figure 4.19 Air-liquid interface organoid culture does not improve induction of testis markers ...... 183 Figure 4.20 Swirler culture organoids show testicular characteristics after 17 days of differentiation ...... 188 Figure 4.21 Differentiation to gonad-like lineages in a SOX9 reporter iPSC line ...... 192 Figure 4.22 Schematic of the differentiation protocol for generating male gonad lineages from human iPSCs ...... 195

xvi

Chapter 1: Introduction

The mammalian gonad develops from an initial bipotential state, with the potential to develop into either ovaries or testes, depending on the sex chromosome constitution. While this initial decision of sexual fate is straightforward, the molecular networks controlling subsequent differentiation of the gonad are remarkably complex. The differentiated embryonic gonad is comprised of at least three essential cell types; supporting cells, steroidogenic cells and germ cells. These cells underlie the structure and function of the gonad and together allow it to serve its function as the primary reproductive organ.

Molecular regulation of the testicular and ovarian pathways is essential for normal gonadal development and disruption to these pathways can result in Disorders/Differences of Sex Development (DSD) in humans. These are conditions in which the gonadal, anatomical or hormonal sex is atypical, affecting approximately 1.7% of live births (1). Phenotypes included in DSD are wide-ranging, from hypospadias (incorrect placement of the urethra on the penis) (1 in 250 boys), to cases of complete sex reversal (1 in 100,000 babies). The appropriate care and clinical management of these individuals often requires accurate diagnosis of the underlying molecular genetic cause, yet at present this is not obtained in ~60% DSD cases (2). This suggests that many other causative genes remain to be identified. Despite major advances in gene sequencing technologies, many potential causative gene variants cannot be associated with DSD as we lack appropriate functional assays to establish their pathogenicity.

The overarching aim of this study was to use new genetic technologies to identify molecular diagnoses and novel disease mechanisms in a cohort of DSD patients. Additionally, I aimed to develop a protocol to generate human embryonic gonad cells in culture, representing a novel in vitro tool in which to functionally analyse DSD genes. This review begins by outlining the development of the embryonic gonad and the molecular pathways controlling its differentiation into ovaries or testes. Our current understanding of one group of DSD, 46,XX DSD, will be explored with regard to the underlying genetic aetiology. Finally, future avenues of research in DSD and gonadal development will be discussed.

1 1.1 Mammalian sex differentiation

1.1.1 Origins of the mammalian gonad The gonad arises from the urogenital ridge, which consists of three parts: the pronephros, mesonephros (arises at day 24 in human, embryonic day (E) 9 in mice) and metanephros (which will form the kidney) (3). The gonads initially develop as genital ridges, multi-layered epithelial structures (4) on the ventromedial surface of the mesonephros. These develop at E10.3-10.4 in mice (5, 6), and 4.5th - 5th week of gestation in humans (7). During gonad morphogenesis, the genital ridge acts as the source for different gonadal cell types. Genital ridge cells delaminate, entering the underlying mesenchyme to differentiate into the supporting cells; the Sertoli cells in XY or granulosa cells in XX gonads (8). The genital ridge also acts as a source of interstitial gonadal cells, including the male and female steroidogenic cells (Leydig and theca respectively). Additional cell lineages are recruited from the mesonephros, e.g. Peritubular myoid cells in XY gonads (9). By contrast, the primordial germ cells arise at an independent site, the endodermal epithelium of the yolk sack (10), during gastrulation at week three in humans or E6.5 in mice (11). Approximately 50 germ cells actively migrate via the hindgut towards the genital ridge, during which time they also undergo significant proliferation. Germ cells colonise the genital ridge by E10 in mice (12) and between weeks 4-8 in humans (11). The presence of germ cells is not required for testis organogenesis; in their absence the formation of the mouse testis cords is delayed but normal (13). By contrast, in the ovary, germ cells assist in organisation and maintenance of the follicles (14). At E12.5, the division of germ cells in the XY gonad is arrested until after birth. In the XX gonad, germ cells undergo final mitosis at E13.5 then enter meiosis (15).

1.1.2 Regulatory pathways controlling gonadal differentiation

1.1.2.1 The bipotential gonad The development of the bipotential gonad requires a number of transcription factors, and disruptions in these can lead to a partial or complete loss of gonads in either sex. An example of this is “streak gonads”; structures consisting of non-functional fibrous tissue that fail to develop beyond the bipotential state. Below, key genes involved in the formation of the bipotential gonad are discussed.

2 Steroidogenic factor 1 (SF1), encoded by the Nuclear Receptor family 5 subfamily A member 1 (Nr5a1) gene, plays an essential role in the initiation and proliferation of the genital ridge. SF1 is expressed in the coelomic epithelium of the developing genital ridge and SF1-positive cells proliferate to give rise to supporting and steroidogenic lineages in the gonads (5, 16). Nr5a1-null mice do not have adrenal glands or gonads, and both sexes develop female external genitalia due to the lack of testosterone (17-19). In humans, heterozygous mutations in NR5A1 can cause DSDs such as 46,XY gonadal dysgenesis (20, 21). A number of transcription factors regulate the formation of the bipotential gonad, often by regulating Nr5a1 expression (Fig. 1.1). These are detailed below.

GATA Binding protein 4 (GATA4) is a central modulator of Nr5a1 activity. In mice, loss of Gata4 in the genital ridge causes reduced coelomic epithelium thickening and loss of ridge markers such as SF1. Instead, an undifferentiated monolayer remains and gonads do not form (5). GATA4 activity is in turn modulated by Friend of GATA 2 (FOG2), also known as Zinc Finger Protein, FOG Family Member 2 (ZFPM2) (22, 23). In the mouse embryo, Fog2 expression generally parallels that of Gata4 (23, 24) and loss of the GATA4–FOG2 interaction leads to a block in gonadal development (25). Also acting upstream of Nr5a1 is the transcription factor Lim Homeobox 9 (LHX9). Lhx9 mutant mice show failure of gonad formation, as the epithelial cells of the genital ridge do not proliferate into the mesenchyme (26). An alternate LIM gene, Lim Homeobox 1 (Lhx1), is required for the formation of intermediate mesoderm, and thus the genital ridge (27). Lhx1-null mice lack gonads (28). The transcription factors Six Homeobox 1 (Six1) and Six Homeobox 4 (Six4) bind upstream of Fog2 and mice null for both Six1 and Six4 have reduced gonadal precursors (29), and decreased Nr5a1 expression. Chromobox homolog 2 (Cbx2) is a polycomb gene that when mutated in mice results in disrupted genital ridge formation and consequently hypoplastic gonads (30, 31). Cbx2 regulates Nr5a1 and possibly Lhx9 and Gata4, given their reduced expression in Cbx2-null mice (30, 31). Wilms Tumor 1 (WT1) is a zinc finger transcription factor that can act as both a transcriptional activator and repressor. In mice, WT1 regulates Nr5a1 in parallel to LHX9/GATA4 during genital ridge development (5, 32). Wt1-null mice embryos have apoptosis of the genital ridge and complete gonadal dysgenesis (33-35). Finally, Empty Spiracles Homeobox 2 (Emx2) is required to maintain genital ridge epithelial cell polarity and proliferation, as well as regulate the epithelial to mesenchymal transition and migration (6). Emx2-null mice have absent gonads (36), disrupted cell division in the coelomic epithelium and decreased migration of gonadal epithelial cells to the mesenchymal

3 compartment (6). EMX2 appears to work in a separate pathway to both GATA4/LHX9 and WT1 as Emx2-KO gonads lose Nr5a1 expression yet Gata4, Wt1 and Lhx9 are not downregulated (5, 6).

WNT signalling also features in the developing gonad. Wingless-type MMTV integration site family member 4 (WNT4) and R-spondin family member 1 (RSPO1) are members of the WNT signalling pathway. Rspo1; Wnt4-null mice have reduced coelomic cell proliferation, resulting in hypoplastic testes (37). In the gonads of these double knock-out (KO) mice, Nr5a1 is not downregulated, which implies that Rspo1 and Wnt4 are acting either downstream of Nr5a1 or independently, to regulate growth of the bipotential gonad.

The initiation of the genital ridge from the coelomic epithelium, followed by the proliferation and growth of the bipotential gonad are primarily controlled by SF1 and numerous factors that interact with this central transcription factor (Fig. 1.1). These set the stage for sexual differentiation, a process that reuses many of the same signalling factors essential in the bipotential gonad.

4

Figure 1.1. Genes and pathways activated during the development and differentiation of the genital ridge and bipotential gonad. The activation of NR5A1/SF1 is a central feature of genital ridge development; this is regulated by a number of transcription factors including CBX2, GATA4/FOG2 (directly and indirectly via LHX9) and SIX1/SIX4 (directly and indirectly via FOG2). WNT4/RSPO1 act either downstream or independently of SF1 and function in the proliferation of the genital ridge. Similarly, the activation of SF1 remains important during the development of the bipotential gonad, a structure arising from the genital ridge. SF1 is activated in the developing bipotential gonad by WT1 (-KTS isoform) and EMX2. EMX2 also controls a number of cellular processes in the bipotential gonad including the maintenance of epithelial cell polarity (via EGFR), proliferation and epithelial to mesenchymal cell transition.

1.1.2.2 Ovarian differentiation The process of sexual differentiation occurs from weeks 6-10 in humans and E11.5-E12.0 in mice. We know of several essential regulatory pathways controlling ovarian development, although as a whole there is a lot less is known about ovarian differentiation than its male counterpart. This review focuses mainly on signalling in the female gonadal supporting cells (granulosa), as these are the first cells to differentiate and control the rest of the ovarian cell’s fates.

5 Canonical WNT signalling is a key feature of the developing ovary, with secreted factors WNT4 and RSPO1 acting as the positive effectors that stabilise β-catenin (encoded by the Catenin Beta 1 (Ctnnb1) gene) (38-42) (Fig. 1.2). Rspo1 and Wnt4 are some of the earliest ovarian markers, with expression increasing from E12.5 in the mouse ovary (39, 40). WNT4 and RSPO1 proteins are secreted by the somatic cells, though RSPO1 is also found at the membrane of germ cells (43), suggesting a role in multiple ovarian lineages. Both Rspo1- and Wnt4-null XX mice show partial sex reversal and ovotestes (40, 44, 45). Rspo1-null XX gonads lose Wnt4 expression (46), while Rspo1 expression remains intact in Wnt4-null XX gonads, indicating that RSPO1/WNT4 act in a linear pathway during ovarian differentiation. In line with this, Rspo1; Wnt4-null XX mice show the same ovarian phenotypes as their single XX mutants (47), however earlier defects in coelomic cell proliferation are observed, highlighting their additional role in the bipotential gonad.

Once the canonical WNT signalling pathway has been activated, β-catenin becomes stabilised and can affect its downstream targets (Fig. 1.2). Indeed, forced β-catenin stabilisation in XY mouse gonads is sufficient to induce ovarian tissue (42). Interestingly conditional Ctnnb1 KO in fetal XX somatic cells does not disturb ovarian differentiation (41, 48), implying that Ctnnb1 loss must occur in additional ovarian cells to cause sex reversal. Alternatively, backup signalling pathways may be activated by RSPO1 and WNT4 to maintain the female fate.

Several feedback loops actively maintain canonical WNT signalling during ovarian differentiation (Fig. 1.2), one of which is the insulin signalling pathway. The insulin receptor tyrosine kinase family consists of insulin receptor (INSR/IR), IGF type I receptor (IGF1R) and insulin receptor- related protein (INSRR/IRR). Insr; Igf1r; Insrr-null and Insr; Igf1r-null XY mice show complete male-to-female sex reversal (49, 50). Insr; Igf1r-null XX mice have reduced levels of Wnt4, while Rspo; Wnt4-null mice show significantly reduced Igf1r expression (47), suggesting a positive feedback loop exists between these two pathways. Likewise, the WNT4 target gene Runt- Related Transcription Factor 1 (Runx1) also forms a positive feedback loop to maintain Wnt4 expression (51).

Another signalling factor contributing to ovarian cell fate is the Forkhead box L2 (Foxl2) gene. Foxl2 is a member of the Forkhead box gene family, a group of conserved transcription factors. In some species FOXL2 appears critical for ovarian fate; FOXL2-null goats and several fish species show female-to-male sex reversal (52, 53, 54). However, in mice, despite Foxl2 showing early ovary-specific upregulation, Foxl2-null mice do not show sex reversal in the embryonic or

6 perinatal period (55). Instead Foxl2 loss causes disorganised ovaries with absent secondary follicles, as granulosa cell differentiation is arrested at the cuboidal transition phase. Similarly, FOXL2 mutations in human patients can cause blepharophimosis ptosis epicanthus inversus syndrome (BPES), an autosomal dominant condition characterised by eyelid abnormalities and primary ovarian insufficiency (POI) (56, 57). Thus, in humans and mice FOXL2 has been shown to control the organisation and maintenance of gonadal fate (58), yet its role during ovarian differentiation is still uncertain. In other species, the importance of FOXL2 may be in its regulation of the Cytochrome P450, Family 19, Subfamily A, Polypeptide 1 (Cyp19a1) gene, which encodes the aromatase enzyme necessary for oestrogen production (59-61). In the developing mouse ovary FOXL2 antagonises several testis pathway genes, such as Wt1, and consequently Sf1 (62). Further study in additional mammalian species may help pinpoint the role of FOXL2 during gonadal differentiation.

Following their differentiation, the granulosa cells initiate the Hedgehog signalling pathway. Morphogens such as Desert Hedgehog (DHH) and Indian Hedgehog (IHH) are secreted from the granulosa cells, kick-starting differentiation of the steroidogenic (theca) cells. DHH signals to theca cells via binding to Hedgehog Interacting Protein 1 (HHIP) and membrane-bound receptors Patched (PTCH1 and PTCH2). This results in activation of GLI transcription factors, which will activate Hedgehog target genes.

Theca cell progenitors arise from two locations, the genital ridge and the mesonephros. These two theca cell-types show distinct transcriptional profiles (63); mesonephros-derived cells show upregulation of genes associated with steroidogenesis, such as Steroidogenic acute regulatory protein (StAR), Cytochrome P450 (Cyp11a1 and Cyp17a1), and Luteinizing Hormone/Choriogonadotropin Receptor (Lhcgr). By contrast, ridge-derived theca cells show high expression of Oestrogen receptor 1 (Esr1), Wt1 and genes implicated in cell growth and proliferation (63).

Theca cells function primarily in the production of androgens, subsequently converted to oestrogens by granulosa cells. Disrupted theca cell differentiation or function can result in ovarian conditions including polycystic ovary syndrome, POI and ovarian cancer. The differentiation of theca and granulosa cells is therefore tightly linked and a greater understanding of the multicellular interactions in this process is required.

7

Figure 1.2. Molecular signalling pathways during differentiation of the gonads into testes or ovaries. (A) Ovarian differentiation. The upregulation of WNT4 and RSPO1 activate the canonical WNT signalling pathway, resulting in the stabilisation of β-catenin. WNT4/RSPO1 signalling is also upregulated by positive feedback loops with the insulin signalling pathway members INSR/IGF1R and RUNX1. The WNT/β-catenin pathway activates numerous downstream effectors such as FST, which directs differentiation of the granulosa cells. The ovarian pathway is maintained through embryonic development by mutually antagonistic signals, with the RSPO1/WNT4/β-catenin pathway and the SRY/SOX9/FGF9 pathways repressing one another. Postnatally and through adulthood these interactions remain essential to maintain gonadal identity and prevent trans-differentiation between male and female cell types, for example FOXL2 supresses testicular factors DMRT1 and SOX9. (B) Testicular differentiation. SRY and SOX9 upregulation is central to testis differentiation. SRY and SF1 initiate the expression of SOX9, leading to differentiation of the pre-Sertoli cells. SOX9 directs the differentiation of pre-Sertoli to Sertoli cells and its expression is maintained through autoregulation, positive feedback loops with FGF9 and PGD2 and by upregulation from CBX2, DAX1, WT1 and SF1. Postnatally, the sexual fate of the testis is maintained by DMRT1 suppressing feminising factors FOXL2 and retinoic acid signalling. Asterisks indicate genes/proteins implicated in 46,XX disorders of ovarian development, all other proteins and interactions are inferred from other human DSDs or animal studies. Figure taken from (64).

1.1.2.3 Testis differentiation The Y-linked Sex determining region Y (SRY) gene (65-67) is the key factor directing differentiation of the mammalian bipotential gonad into a testis. The removal of SRY/Sry causes male-to-female sex reversal in humans (65, 68) and mice (66, 69). Similarly, ectopic expression

8 of SRY/Sry initiates testis development in both humans (70, 71) and mice (72). Sry is expressed specifically in the XY mouse gonad from E10.5, regulated by factors including SF1, GATA4, WT1 and INSR (25, 34, 49, 73) (reviewed in (74)).

The expression of SRY in the bipotential gonad triggers differentiation of coelomic cells into male supporting (Sertoli) cells through activation of SRY-Box 9 (Sox9). Sertoli cells play a central role in testis development; directing the differentiation of Leydig cells, corralling germ cells into the testis cords, and organising the formation of seminiferous tubules. SOX9 is necessary for testis differentiation across all vertebrates studied, its expression during Sertoli cell differentiation is maintained by several factors (Fig. 1.2). In the mouse, SF1 initiates low level Sox9 expression in the genital ridges of each sex, however once Sry reaches a critical expression threshold SRY and SF1 form a complex that strongly enhances Sox9 transcription. SOX9 expression is then maintained via auto-regulation (75, 76) and positive feedback loops with Fibroblast growth factor 9 (FGF9) and Prostaglandin D2 (PGD2) (77, 78)(Fig. 1.2). Similar to SRY, loss of SOX9 results in male-to-female sex reversal in humans and mice (79, 80), while ectopic expression of SOX9 causes testis development in XX females (81, 82). In addition to its feedback loop partners, FGF9 and PGD2, targets of SOX9 (often in conjunction with SF1) include Anti-Müllerian hormone (Amh) (83), Vanin 1 (Vnn1) (84), Cytochrome P450, Family 26, Subfamily B, Polypeptide 1 (Cyp26b1) (85) and Precerebellin 4 (Cbln4) (86). The exact roles of VNN1 and CBLN4 in testis formation are still undetermined, however AMH is responsible for regression of the Müllerian ducts. Studies in E13.0 rat testis identified 109 gene promoters that are potential SOX9 binding targets (87), highlighting that further targets remain to be identified.

SOX9 does not appear to be required for testis development beyond E14.5 in mice, possibly due to functional redundancy with other genes such as SRY-Box 8 (Sox8). However, during this functional window (E11.5-14.5) SOX9 is tightly regulated by a number of known modifiers. One modifier is the nuclear receptor DSS-AHC critical region on the X chromosome 1 (DAX1), encoded by Nuclear Receptor Subfamily 0 Group B Member 1 (Nr0b1). This protein has complex roles in the gonadal development of both sexes; consequently, testis development is disrupted by both Nr0b1 mutations and overexpression. In 46,XY individuals, NR0B1 deletions have been associated with X-linked adrenal insufficiency and hypogonadotropic hypogonadism (88), suggesting that DAX1 promotes testis development. However, an anti-testis function has also been shown, where duplications encompassing NR0B1 cause male-to-female sex reversal in

9 humans (89) and mice (90, 91). Other factors regulating Sox9 either directly or indirectly include CBX2 (30), WT1 (33), GATA4 and its modulator FOG2 (92) (Fig. 1.2).

Thus, in the differentiating Sertoli cells numerous factors downstream of SRY initiate and contribute to the male regulatory pathway. These systems help to recruit as many cells as possible to the Sertoli cell fate, increasing the likelihood of successful differentiation of the gonad into a testis, while simultaneously repressing the female fate (see below). Following Sertoli cell differentiation, additional interstitial cell types differentiate. One of these key cell types is the male steroidogenic (Leydig) lineage, which arises from both the coelomic epithelium and gonad–mesonephros border cells around E12.5 (93). Fetal Leydig cells (FLCs) reinforce the male-specific differentiation of the testis by producing steroid hormones such as androgens (reviewed in (94)). SF1 is a master regulator in these cells, initiating the expression of the majority of factors involved in cholesterol mobilisation and steroid hormone synthesis (reviewed in (95)). This includes StAR (96) and genes encoding the cytochrome P450 steroid hydroxylases and 3β steroid dehydrogenase (3βHSD), which act together in the conversion of cholesterol to steroid hormones (97).

Another key regulator of FLC differentiation is the Hedgehog signalling pathway. Differentiated Sertoli cells secrete the DHH (98), which Leydig and peritubular myoid cells respond to via PTCH1, eliciting a transcriptional response by GLI transcription factors. Dhh-null mice have testis dysgenesis, reduced Leydig cells and feminised external genitalia due to a lack of testosterone (98). Several 46, XY DSD patients have been shown to carry DHH mutations, these patients have a range of phenotypes from mixed, partial to pure gonadal dysgenesis (99, 100).

An additional feature of the embryonic gonads that diverges between the sexes is the vasculature. From E11.5 in the mouse testis, vascular endothelial cells migrate from the mesonephros to form the coelomic vessel, which branches through the interstitium between testis cords (9, 101). The remodelling of the XY vasculature is an important step in both the patterning and physiological function of the developing testis and is regulated by a number of factors, including EphrinB2 (Efnb2), Notch signalling pathway members Notch1, Jagged1, Jagged2 (102), and Inhibin beta B (Inhbb) (103). Fetal macrophages also act as important regulators of vascular remodelling (104).

10 1.1.2.4 Mutual repression between ovarian and testicular pathways While addressed separately above, a complex interplay between the male and female molecular pathways underlies the initiation, propagation and maintenance of gonad differentiation. The testicular and ovarian pathways each act to suppress each other during gonadal differentiation and throughout adulthood (58, 105, 106) (Fig. 1.2). In the embryonic gonads, the key antagonistic relationship exists between the SRY/SOX9/FGF9 and WNT4/β-catenin signalling pathways. In XY gonads, SRY antagonises the WNT transcription factor β-catenin, targeting it to nuclear bodies to trigger its degradation and inhibit transcriptional activity (107) (Fig. 1.2). In addition, FGF9 maintains transcription of Sox9 while also down-regulating ovarian genes (e.g. Wnt4) (108, 109). During ovarian development, WNT4 suppresses Sox9 expression, proposed to be via the activation of β-catenin (77) (Fig. 1.2). The WNT4 signalling pathway also up-regulates the Follistatin (Fst) gene (via β-catenin), which in turn antagonises Activin B, preventing the formation of testis-specific vasculature (110). Thus, the WNT4/β-catenin and SRY/SOX9/FGF9 pathways are mutually exclusive, and upsetting balance during early differentiation can result in gonadal sex reversal. For example, ectopic expression of β-catenin in XY mouse somatic cells is powerful enough to result in male-to-female sex reversal (42). Similarly, SOX9 or FGF9 overexpression can cause female-to-male sex reversal (77, 82).

Once established, the sex identity of the somatic cells must be maintained throughout life to avoid transdifferentiation. This is achieved, at least in mice, by the mutual repression of Doublesex and mab-3 related transcription factor 1 (DMRT1) and FOXL2. Dmrt1-null XY gonads show evidence of transdifferentiation, with FOXL2+ granulosa cells appearing 4 weeks postnatally (105). DMRT1 activates testis genes Sox9, Sox8 and Prostaglandin D2 Synthase (Ptgds) and represses ovarian genes Foxl2, Wnt4 and Rspo1 (111). It is also proposed that DMRT1 antagonises retinoic acid (RA) signalling in the testis, a known activator of the ovarian pathway (112). Similarly, in the female, FOXL2 is central to the maintenance of granulosa cell fate. It binds to the upstream enhancer region (TESCO) of Sox9 to prevent transcription in the adult mouse ovary (58). Foxl2 overexpression in XY mice results in disorganised tubules and ovotestis-like gonads (113), while conditional deletion results in transdifferentiation of adult granulosa and theca cells into Sertoli-like and Leydig-like cells respectively (58). It is not yet known whether transdifferentiation occurs in human gonads and its contribution to DSD needs to be addressed.

11 1.2 46,XX Disorders/Differences of Sex Development So far I have discussed genes involved in human/mouse sex determination and differentiation. This process is tightly controlled and we are slowly annotating the specific pathways involved in differentiation of the testis and ovary. Disruptions to these pathways can result in human DSDs. Around a quarter of DSD conditions occur in 46,XX individuals (114-116). The Chicago consensus statement defined three categories of 46,XX DSD (117). The first group includes disorders relating to androgen excess, such as congenital adrenal hyperplasia (CAH), aromatase deficiency and P450 oxidoreductase deficiency. The second group includes rare syndromes and disorders, such as uterine abnormalities, vaginal atresia and MURCS (Müllerian duct aplasia, Renal dysplasia, Cervical Somite anomalies) syndrome. The third group, which is the focus of this review, is disorders of ovarian development. This is made up of three phenotypes, 46,XX testicular DSD (T-DSD), 46,XX ovotesticular DSD (OT-DSD) and 46,XX ovarian (gonadal) dysgenesis. Categorising these DSD subgroups has been instrumental in understanding the genetic basis of each condition. Broadly, 46,XX (ovo)testicular DSDs are caused by gain-of- function in male pathway genes or loss-of-function in female pathway genes, while 46,XX ovarian dysgenesis is commonly caused by disruptions to hormone signalling or folliculogenesis genes. As a whole, the molecular cause of many 46,XX DSDs is unknown, highlighting gaps in our knowledge of ovarian development.

1.2.1 46,XX testicular DSD 46,XX T-DSD has a prevalence of 1 in 20,000-25,000 males (118) and is characterised by 46,XX karyotype with male external genitalia (ranging from normal to ambiguous), testicles and absent Müllerian structures (119). Only about 15% of cases are diagnosed at birth due to ambiguous genitalia, however this figure is likely to increase with the growing use of prenatal screening (including karyotyping). Many individuals are diagnosed following puberty, presenting with small testes, gynecomastia and azoospermia (119). Some 46,XX T-DSD cases present with hypospadias (10-15%) due to incomplete virilisation (120) and the majority will be infertile due to the absence of Y-chromosome linked azoospermia factors.

Around 90% of 46,XX T-DSD cases are caused by translocation of the SRY gene onto the X chromosome (70, 71, 121). Short stature is frequently observed in these cases (122) due to altered expression of growth-related genes, as well as gynecomastia (119), likely caused by

12 abnormal hormonal signalling. In the remaining 10% of 46,XX T-DSDs, many of these also result from gain-of-function in testicular genes. One example is the ectopic expression of SOX9, resulting from duplication of the gene (82) or its upstream enhancer region (81, 123) (Table 1.1). So far SOX9 duplications have been reported in over 20 cases of 46,XX T-DSD, making it the most common genetic cause after SRY translocation. Deletions in the SOX9 upstream enhancer region have also been reported in 46,XY sex reversal. Identifying the minimal regions of overlap for these copy number variants (CNVs) has delineated two human testis-specific regulatory regions, termed RevSex and XYSR (81, 124, 125). In addition to duplications, reciprocal translocations including the SOX9 enhancer region have been found in cases of 46,XX T-DSD (125), where it was proposed that regulatory elements for adjacent genes (LOC204010 or DEYNAR) cause ectopic SOX9 expression. However our understanding of these cases is complicated by the fact that breakpoints or duplications in the same genomic region have been identified in 46,XX patients with no DSD phenotype but either campomelic dysplasia (126-128) or brachydactyly- anonychia (129) respectively. This suggests that SOX9 has a dosage-dependent effect on sex development and variable expressivity.

Other SOX genes have been implicated in 46,XX T-DSD, such as SRY-box 3 (SOX3) duplications and a deletion in four 46,XX T-DSD individuals (130) (Table 1.1). Duplications including SRY-box 10 (SOX10) have also been found in two cases of 46,XX T-DSD (131, 132) and single cases of both 46,XX OT-DSD (133) and 46,XX ovarian dysgenesis (134). Although neither of these SOX genes have a known role in gonadal development, their overexpression in the gonads is able to ectopically activate the testicular pathway via the conserved HMG domain they share with SRY.

A duplication in a regulator of SOX9, FGF9 (135), has also been implicated in a case of 46,XX T- DSD. Although given that FGF9 expression in the patient was still significantly lower than a normal male, it is likely that other genomic modifiers also played a part in the phenotype.

Finally, WT1, a regulator of bipotential gonad and testis development (Fig. 1.1 and 1.2), was recently implicated in both isolated and syndromic 46,XX T-DSD (136, 137). These were frameshift and missense mutations in the fourth zinc finger of this gene. The frameshift variant is predicted to have a gain-of-function effect (137), increasing target activation of WT1, yet in vitro studies confirming this testis-activating role have not been performed.

As shown here, the genetic causes of many SRY-negative 46,XX T-DSDs converge on regulation of SOX9, specifically, being caused by ectopic SOX9 upregulation. As there are still a significant

13 number of undiagnosed 46,XX T-DSD cases, it may be worthwhile to look in SOX9 target genes or regulators for novel genetic causes. Interestingly, a cohort of SRY-negative 46,XX T-DSD patients showed significantly higher (1.9-fold) SOX9 expression in gonads compared to control XY testes (138), suggesting that dysregulation of SOX9 may underlie these cases. Backup pathways for testis differentiation in the absence of SRY may also contribute to undiagnosed 46,XX T-DSDs. CBX2 is a key candidate, known to regulate testis differentiation through simultaneous activation of testis genes and suppression of ovarian genes (139). Furthermore, the absence of interaction between SRY and CBX2 (139) suggests that CBX2 may drive a male regulatory pathway independent of SRY. While variants in this gene have not yet been associated with DSD, a species of rat with XO/XO sex chromosomes (no Y chromosome or Sry) shows testis development, thought to be driven by extra copies of Cbx2 (140).

Loss-of-function in female pathway genes also underlies a small portion of SRY-negative 46,XX T-DSD cases. Homozygous truncating or missense mutations and deletions in RSPO1 have been found in six individuals (three families) presenting with 46,XX T-DSD as well as palmoplantaer hyperkeratosis and predisposition to squamous cell carcinoma of the skin (39, 141) (Table 1.1). Another ovarian gene, WNT4 has been mutated in a syndromic case of 46,XX T-DSD (142), discussed below. Identifying other regulators of ovarian development, working alongside or in parallel with WNT4/RSPO1 signalling may assist in the diagnosis of more 46,XX T-DSDs. Indeed, a recent exome sequencing study on a large cohort of SRY-negative 46,XX T-/OT-DSD cases identified three individuals with heterozygous frameshift mutations in the Nuclear Receptor Subfamily 2 Group F Member 2 (NR2F2) gene (143). These individuals each presented with congenital heart defects and additional anomalies including BPES and congenital diaphragmatic hernia. NR2F2 encodes COUP-TF2, a transcription factor expressed specifically in FOXL2- negative stromal cells of the ovary. Its role during ovarian development is not yet understood, but in non-gonadal tissues COUP-TF2 upregulates Wnt4 and suppresses Sox9 (144, 145).

Until recently, single nucleotide variants (SNVs) had not been associated with isolated 46,XX T- DSD. In 2016, three reports emerged with the same heterozygous missense mutation in the NR5A1 gene (c.C274T, p.Arg92Trp) causing 46,XX T/OT-DSD in unrelated cases (146-148). This variant has since been associated with over ten 46,XX OT/T-DSD cases and interestingly is also found in unaffected 46,XX and 46,XY family members. Nr5a1 has known roles in the promotion of both the ovarian (via Wnt4/Foxl2 (149)) and testicular (via Sox9 (150)) pathways. The

14 mechanism by which this variant causes sex reversal has not yet been established, but given the appearance of this variant in wide ranging DSD and unaffected phenotypes, it is likely that other genomic or environmental factors play a role.

1.2.2 46,XX ovotesticular DSD OT-DSD is characterised by the presence of testicular and ovarian tissue, occurring most commonly in the same gonad but sometimes in different gonads. Gonadal biopsy is therefore required to establish the diagnosis in these cases. In the ovarian region of ovotestes, follicular growth and oestradiol production are often normal, with 50% of ovotestes showing evidence of ovulation. Spermatogonia development is inhibited by the presence of oestradiol, meaning that the testicular portion often regresses with time due to Leydig cell hyperplasia. A number of chromosome make-ups can underlie OT-DSD, approximately 10% are 46,XY, 25% have sex chromosome mosaicism and the remaining have 46,XX karyotype (151). 46,XX OT-DSD is rare, occurring in an estimated 1 in 100,000 live births (1). Babies are usually born with ambiguous genitalia; many are reared as male due to the size of the phallus. However as there is functioning ovarian tissue, both breast development and puberty can occur, and around two thirds of 46,XX OT-DSDs will menstruate (152).

From a molecular standpoint, fewer cases of 46,XX OT-DSD are understood than 46,XX T-DSD, SRY translocations account for just 10% of cases (151). Eight cases have been reported to have SOX9 enhancer duplications ((124, 125, 153, 154) and Croft et al., (2018), Nature Communications 9:5319) and five reported cases of 46,XX OT-DSD have the NR5A1 p.Arg92Trp variant (146-148, 255, 256), highlighting that routine genetic screening should include these genes. Loss-of-function mutations have been found in ovarian genes WNT4 and RSPO1. Specifically, both homozygous splice site and missense RSPO1 mutations have been found in two cases of 46,XX OT-DSD with palmoplantar keratoderma (141, 155) (Table 1.1). While WNT4 mutations have been found individuals with abnormal development of the reproductive tract and variable degrees of sex reversal. For example, two 46,XX fetus’ in a single pedigree carried the same homozygous missense mutation in WNT4 yet presented with either complete (T-DSD) or partial (OT-DSD) sex reversal (142). This finding defined a rare condition now known as SERKAL syndrome, characterised by gonads ranging from ovotestis to normal testis, ambiguous genitalia, and renal, adrenal and lung dysgenesis. Interestingly, heterozygous missense WNT4

15 mutations have also been identified in 46,XX females with normal ovarian tissue but absent Müllerian structures, unilateral renal agenesis and androgen excess (156, 157), similar features to Mayer-Rokitansky-Kuster-Hauser syndrome. WNT4 mutations are thus associated with a diverse spectrum of DSD phenotypes, suggesting that this gene acts at multiple points in female reproductive development and has incomplete penetrance.

Loss-of-function mutations in ovarian pathway genes can thus result in either complete or partial XX sex reversal. Relatively few genes have been implicated in ovarian development compared to testis; these unknown genes likely underlie some of the undiagnosed cases of 46,XX OT-DSD. Co-factors, regulators and downstream targets of WNT4 and RSPO1 are excellent candidates for these cases. A study on CNVs in four SRY-negative 46,XX T-DSD cases identified a candidate in Glypican 5 (GPC5) (158), a known mediator of WNT signalling. Mouse KO studies can also identify novel candidates, such as the Leucine-Rich Repeat-Containing G Protein- Coupled Receptor 4 (LGR4) gene. LGR4 acts as a receptor to R-spondin ligands during WNT4/β- catenin signalling in mouse gonads (159) and causes female-to-male sex reversal when knocked out (160). Finally, comparison of embryonic mouse gonads from wild-type and Wnt4 mutant mice has enabled the identification of potential targets of WNT signalling, such as Notum Pectinacetylesterase Homolog (Notum), Msh Homeobox 1 (Msx1), and Runx1 (51).

In addition to looking for novel gonad development genes, some 46,XX DSDs may be caused by known DSD genes exerting novel function. Indeed, this was the case for NR5A1, which until recently was only implicated in 46,XY gonadal dysgenesis (161, 162) and POI (163). This demonstrated how genes in gonad development can act as promoters of both the ovarian and testicular pathways. This is also the case for the NR0B1 gene, where an 80 kb microdeletion was recently identified in a single case of 46,XX OT-DSD (164). These examples suggest that all known DSD genes should be thoroughly examined during genetic screening of undiagnosed cases.

16 1.2.3 46,XX ovarian dysgenesis Individuals with 46,XX ovarian dysgenesis present as phenotypically normal females at birth, yet later show delayed or absent puberty, resulting in primary and sometimes secondary amenorrhea. This is due to the ovaries being underdeveloped, with early alterations to ovarian function or structure causing reduced reproductive development later (165). Development of external genitalia is usually normal, yet the presence of streak gonads can mean they later develop uterine hypoplasia and hypergonadotropic hypogonadism (166). This condition is etiologically heterogeneous and can be associated with syndromes such as Perrault syndrome, these syndromes are not discussed in this review. These conditions show genetic heterogeneity and usually have an autosomal recessive inheritance pattern (167). Variable expressivity is also frequently seen, for example two siblings held the same mutation yet one presented with streak gonads and the other primary amenorrhea and ovarian hypoplasia (167).

46,XX ovarian dysgenesis is considered a major cause of POI, in fact the distinction between these two conditions is blurred, with many POI genes also implicated in 46,XX ovarian dysgenesis. POI is increasingly used as a term that covers a spectrum of conditions where ovaries are dysfunctional and 46,XX ovarian dysgenesis is a point on this spectrum. As well as POI, 46,XX ovarian dysgenesis is thought to also contribute to reduced fertility, polycystic ovary syndrome, and reproductive cancers (165). The rise in such ovarian conditions in the last 100 years suggests that both genetic and environmental factors play a role in this condition. One suggested contributor is endocrine disrupting chemicals, found in diethylstilbestrol, a drug commonly prescribed to pregnant women from the 1940s to 1970s (168).

Disruptions to hormonal signalling and folliculogenesis underlie many 46,XX ovarian dysgenesis cases. The Follicle stimulating hormone receptor (FSHR) gene is essential in ovarian folliculogenesis (169) and numerous mutations in this gene have been reported in 46,XX ovarian dysgenesis (170) (Table 1.1). One example is a familial case with recessive inheritance of a missense mutation (p.Ala189Val) which disrupts cell-surface trafficking of FSHR, blocking the development of follicles (171). Interestingly, the majority of these reported FSHR mutations are in individuals of Finnish descent, suggesting that phenotypic modifiers found more frequently in the Finnish genome, likely due to a population bottleneck, may exacerbate the effects of FSHR mutations. Regulators of FSHR have also been implicated in 46,XX ovarian dysgenesis, such as

17 Bone Morphogenetic protein 15 (BMP15). Heterozygous mutations in BMP15 disrupt BMP15 processing (172), reducing the proliferative ability of granulosa cells. Impaired estrogen action has also been implicated in a recent case of 46,XX ovarian dysgenesis, where a woman presenting with streak gonads, absent primary amenorrhea and osteoporosis was found to have a heterozygous missense mutation found in the Estrogen receptor β (ESR2) gene (173).

In addition to FSH and its regulators, genes involved in other ovarian processes are also implicated in 46,XX ovarian dysgenesis. Proteasome 26S subunit, ATPase, 3-Interacting Protein (PSMC3IP) regulates chromosome pairing during meiosis and stimulates transcription in response to nuclear receptors (oestrogen, thyroid hormone, androgen, glucocorticoid and progesterone receptors) (174). A 3 bp deletion in PSMC3IP was reported in a familial case of 46,XX ovarian dysgenesis (175), disturbing the co-activation function of this gene in oestrogen- mediated transcription. The Newborn ovary homeobox protein (Nobox) gene shows a clear ovarian phenotype when knocked out in mice, with XX null mice showing infertility, increased loss of oocytes postnatally and atrophic ovaries (176). Heterozygous mutations in the NOBOX gene have been found in 46,XX ovarian dysgenesis and POI cases (177), although large-scale screening studies show this is not a common cause (178, 179). Finally, a homozygous missense mutation in Nucleoporin 107kDa (NUP107) was found in four familial cases of 46,XX ovarian dysgenesis (180). This gene is conserved between Drosophila and human, and introduction of this mutation in flies reduced fertility and caused the ovariole (fly ovary) to disintegrate. It is thought that NUP107 may act during meiosis in a similar manner to two known POI genes, Minichromosome Maintenance Complex Component 8 (MCM8) and MCM9, repairing the double-stranded breaks induced by homologous repair (181).

18 Table 1.1 Genes associated with 46,XX Disorders of ovarian development in humans.

Gene Gain of function Loss of function Types of mutation Publications

BMP15 46,XX ovarian dysgenesis Heterozygous missense (Di Pasquale, 2004) mutations

ESR2 46,XX ovarian dysgenesis Heterozygous missense (Lang-Muritano, 2018) mutation

FGF9 46,XX testicular DSD Large duplication (Chiang, 2013)

FSHR 46,XX ovarian dysgenesis Homozygous mutations (Aittomäki, 1995; Lalioti, 2011)

NOBOX 46,XX ovarian dysgenesis Heterozygous mutations (Bouilly, 2011)

NR0B1 46,XX ovotesticular DSD 80 kb deletion Xp21.2 (Dangle, 2017)

NR2F2 46,XX testicular DSD, heart Heterozygous frameshift (Bashamboo, 2017) defects and BPES mutation

NR5A1 46,XX testicular DSD and Heterozygous missense (Baetens, 2016; Bashamboo, 2016; 46,XX ovotesticular DSD mutation Igarashi, 2016; Swartz, 2016)

NUP107 46,XX ovarian dysgenesis Homozygous missense (Weinberg-Shukron, 2015) mutation

PSMC3IP 46,XX ovarian dysgenesis, POI 3 bp microdeletion (Zangen, 2011) and POF

19 RSPO1 46,XX ovotesticular DSD and Splice donor site mutations, (Parma, 2006; Tomaselli, 2008; 46,XX testicular DSD with Large deletion. Homozygous Naasse, 2017) hyperkeratosis, hypospadias frameshift and stop mutation and hypogenitalism

SOX3 46,XX testicular DSD Large duplication (Sutton, 2011; Vetro, 2014)

SOX9 46,XX testicular DSD and 46, Duplication/triplication of (Huang, 1999; Cox 2011, Benko, XX ovotesticular DSD gene and upstream region. 2011; Vetro, 2014) Balanced translocation.

SOX10 46,XX testicular DSD, 46,XX Large duplication (Polanco, 2010) ovotesticular DSD and 46,XX ovarian dysgenesis

SRY 46,XX testicular DSD and Translocations (Margarit, 2000; Sharp, 2005) 46,XX ovotesticular DSD

WNT4 46,XX testicular DSD and 46, Homozygous and (Biason-Lauber, 2004; Mandel, 2008) XX ovotesticular DSD, heterozygous missense Mullerian aplasia and mutations hyperandrogenism. SERKAL syndrome

WT1 Syndromic and isolated 46,XX Frameshift and missense (136, 137) testicular DSD mutations

BPES, blepharophimosis-ptosis-epicanthus inversus syndrome; POF, Primary Ovarian Failure; POI, Primary Ovarian Insufficiency; SERKAL, Sex Reversion Kidneys Adrenal and Lung dysgenesis.

20 1.2.4 Identifying genetic factors underlying Disorders of Sex Development There have been significant advances in our understanding of DSD in recent years. This is largely attributed to improved genetic analysis tools. Massively parallel sequencing (MPS) and/or comparative genomic hybridisation (CGH) arrays are now widely used to establish molecular diagnoses in DSD patients. Until recently, a molecular diagnosis was made for just 20% of all DSD patients (117). In our recent study we applied MPS to a 1000 gene panel in a cohort of 326 DSD patients (46,XX and 46,XY DSD phenotypes), achieving a diagnostic rate of ~38% (2). However, if we look specifically at 46,XX DSD phenotypes, this diagnostic rate is just 18%. This is likely attributed to a number of factors, such as our minimal understanding of ovarian development and its regulatory genes, some of which are likely to underlie 46,XX DSDs. As highlighted above, we need to identify novel candidate genes that can be screened for in MPS studies on DSD cohorts.

An additional challenge lies in the interpretation of large quantities of potentially pathogenic variants. This is not unique to 46,XX DSDs, for any rare disease phenotype there is greater diagnostic power and ability to identify novel genes when studying more patients. Obtaining larger cohorts and sharing results on candidate genes among researchers will increase our power to detect novel DSD genes. This problem also highlights the need for improved bioinformatic tools to interpret this data as well as disease-specific models that are higher throughput.

To date, research on sex determination has focused on coding regions, however we know that non-coding regulatory regions and epigenetic modifications also play a part in 46,XX DSDs. The most well characterised example is the upstream enhancer region of SOX9 (182). Techniques used to identify epigenetic modifications or regulatory regions, like bisulphite sequencing and CHIP-seq, rely on large amounts of input material and consequently have not yet been applied to the human embryonic gonad, as this tissue is difficult to source. Increasingly sensitive technology requiring less input material will improve the identification of biologically meaningful non-coding regions. This coupled with additional datasets (e.g. gene/protein expression, whole genome sequencing) will give us clues into regions of the genome important in gonad differentiation and DSD. A recent study (183) demonstrated this,

21 combining complex large datasets to generate candidate gene lists specific to cell types in the testis.

In summary, despite significant advances in the diagnostic rate for DSDs, we are still far from completing the picture. Uncovering novel causes of DSD will come from a collaborative approach combining clinical, bioinformatic and functional expertise.

1.3 In vitro differentiation of gonadal lineages Our understanding of mammalian sex determination has progressed rapidly in the last 25 years; studies on mice have contributed largely to this. However, it has become apparent that increasing the diagnostic rate of DSD will require delineation of the human-specific aspects of sex determination. Creating a human model, such as an embryonic gonad cell line, would provide a means of DSD disease modelling and testing new candidate genes. It would allow fine-tuning of our understanding of known DSD genes in a setting more closely resembling human gonadogenesis.

Since discovering the pluripotent potential of embryonic stem cells (ESCs) (184, 185) and more recently, induced pluripotent stem cells (iPSCs) (186), we have seen a huge amount of research aimed at differentiating tissue-specific cell types and organ models (termed organoids). These have huge power as disease models, e.g. the ability to screen drugs and thus develop individualised patient therapies. Differentiation protocols for many human cell types and tissues have already been established, such as brain (187) and kidney (188), yet few attempts have been made to differentiate gonad lineages. Here I will review studies that have differentiated male gonad-like lineages and discuss how these inform future research.

A range of strategies and starting cell populations have been used in an effort to differentiate male gonadal cells. Most of this work has been aimed at differentiation of steroidogenic lineages, given the need for therapies in patients lacking steroidogenic capacity, e.g. adrenal insufficiency or male hypogonadism. All of these studies have relied on overexpression of the SF1 transcription factor. Mesenchymal stem cells (MSCs) were some the earlier cell types employed; these cells are known for their plasticity and differentiation capacity and can be

22 isolated from a wide array of adult and embryonic tissues. Yazawa et al. (2006) transfected mouse MSCs with SF1 and induced steroid synthesis via cAMP (189), generating both Leydig- like and adrenocortical-like cells. Gondo et al. (2008) used a similar approach, directing differentiation of MSCs to steroidogenic lineages via adenovirus-mediated forced expression of SF-1 (190), however interestingly they showed that adipose-derived MSCs gave rise to steroidogenic lineages representative of adrenal, while bone marrow MSCs gave rise to lineages closer to gonadal steroidogenic lineages. Two other studies generated gonad steroidogenic-like cells via overexpression of SF1 in mouse ESCs, however these cells were reliant on cAMP treatment to begin steroid production (191, 192). Only one study has taken a directed differentiation approach to steroidogenic cell differentiation, Sonoyama et al. (2014) (193) first induced human ESCs/iPSCs into mesoderm lineages with GSK-3 inhibitor (BIO) then overexpressed SF1 (via an expression plasmid). The steroidogenic-like cells generated showed characteristics closest to adrenal cortical cells and had limited capacity for proliferation. Another study explored transcription factors and growth factors beyond SF1 that may be important in differentiation of steroidogenic lineages. Yang et al. (2015) initially derived Leydig-like cells from mouse ESCs by overexpression of SF1 and treatment with two small molecules, 8-Br-cAMP and Forskolin (194). They then developed an approach where Mouse embryonic fibroblasts (MEFs) were differentiated to Leydig-like cells by overexpression of several transcription factors. Starting with a set of 11 Leydig-promoting transcription factors, they narrowed this down to define a minimal set of three essential factors (Nr5a1, Dmrt1 and Gata4) (195). This enabled rapid reprogramming (2 days) to Leydig-like cells, with these cells closely resembling adult mouse Leydig cells in terms of transcriptional profile, epigenetic remodelling and function (Testosterone production). Furthermore, when transplanted these cells were able to rescue the Testosterone levels of testosterone-deficient mice after just 2 weeks.

The in vitro differentiation of Sertoli-like cells has not been explored as extensively. The earliest study differentiated Sertoli- and germ-like cells from hESCs in a simple strategy that altered the growth conditions commonly used in the hESC expansion (reduced feeding cycles and ESC colonies <50 cells) (196). Cultured colonies were composed of approximately 30 and 40% Sertoli- and germ-like cells respectively. Indicators of Sertoli-like qualities included expression of the marker FSHR and Sertoli cell-specific morphological features, such as irregular shaped nuclei and an abundance of endoplasmic reticulum and lysosomes (197-199).

23 Contrasting this simple differentiation approach, Buganim et al. (2012) used genetic manipulation. Differentiation of MEFs into Sertoli-like cells (200) was directed via ectopic expression of five transcription factors, Nr5a1, Wt1, Dmrt1, Gata4 and Sox9, using viral vectors. High levels of these introduced factors resulted in expression of Sertoli, Leydig and germ cell markers, however as Nr5a1 regulates Leydig cells and Dmrt1 is expressed in germ cells, it is possible that these two factors are causing non-specific expression. To test function of the MEF-derived Sertoli-like cells, testicular cells (including germ cells) from 1 day old mice were introduced in a co-culture system. The Sertoli-like cells were able to support the survival of introduced germ cells in vitro for much longer than the MEFs or testicular cells alone. A similar approach was used by Liang et al. (2018) (201), where they reprogrammed human fibroblasts to adult Sertoli-like cells by overexpressing transcription factors GATA4 and NR5A1. A further study by Kjartansdóttir et al. (2015) differentiated three hESC lines (HS207, HS360, HS401) to male gonad-like cells in a basic media with addition of just human recombinant BMP7 (202). They showed that the HS360 line, cultured in suspension with BMP7, showed the greatest potential for a testis-like fate, with upregulation of Sertoli cell markers (SOX9 and WT1) and simultaneous downregulation of pluripotency markers. These cells also showed similar morphology to immature Sertoli cells (e.g. irregular elongating nuclei) and structurally grouped into seminiferous cord-like structures, which had not been reported before. The most recent study (203) used human umbilical cord perivascular cells (HUCPCs) as a source of mesenchymal cells and attempted to mimic the conditions of spermatogenesis (203). They performed five weeks of differentiation, consecutively adding growth factors important for the testicular niche (RA, LIF + GDNF, Putrescine, Testosterone + FSH) followed by co-culture with adult mouse Sertoli cells, generating both Sertoli-like (FSHR, SOX9 and AMH positive) and germ-like cells (VASA, DAZL positive). Furthermore, when HUCPC-derived Sertoli-like cells were injected into mouse they localised to the seminiferous tubules. The main limitations of this protocol are the cell type not being easy to source (from umbilical cords) and a long differentiation procedure requiring many growth factors.

These studies highlight several challenges associated with successful differentiation of gonadal cells. First, many of the regulatory proteins expressed during gonadal differentiation (used as markers for cell lineages) are also highly expressed in closely related tissues. Kjartansdottir et al. (2015) used SOX9 and WT1 as Sertoli cell markers, however these factors are highly expressed in the differentiating kidney, meaning that the true fate of these cells is

24 contentious. Identifying a panel of markers that are specific to the differentiating gonad is essential for thorough characterisation of the resulting cell populations. Second, an appropriate control is required when measuring changes in gene or protein expression following differentiation. The most suitable control would be human fetal gonad, however this is difficult to obtain. Kjartansdottir et al. (2015) and Bucay et al. (2009) measured changes in gonad marker expression levels relative to expression in hESCs. Yet given the wide role of many gonadal markers in the differentiating embryo, it is unsurprising that these markers show upregulation. An appropriate control could establish if changes in expression are within a comparable magnitude to gonadal cells. Finally, prolonged culture of differentiated Sertoli- like cells has so far indicated that they only retain typical Sertoli characteristics for a few days in vitro before de-differentiating. Narrowing down the combination of factors required for their maintenance is necessary for the success of these studies.

These studies also highlight gaps where the differentiation method could be improved. For example, many of these protocols require long differentiations (up to five weeks) (203) and even with this report low efficiency of gonadal cell induction (12.6% FSHR+ Sertoli-like cells (203)). Furthermore, several of these studies report simultaneous differentiation of Sertoli- and germ-like cells in one differentiation. This is questionable given that these cells arise from entirely different parts of the embryo and thus require very different signalling environments. This draws into question the characterisation methods used to define these differentiated cell populations.

Aside from the somatic gonadal cells, there has been significant work in the field of in vitro germ cell differentiation. Until recently this was limited to the differentiation of germ cell precursors, the primordial germ-like cells (PGLCs) (204-206), as no one could recapitulate the epigenome-wide hypomethylation that occurs in their differentiation. Mitinori Saitou and team recently overcame this by co-culturing iPSC-derived human PGLCs with dissociated mouse embryonic gonad tissue for 70 days, stimulating global demethylation (207). Combining this with approaches to differentiation of male somatic gonad cells in future will help to generate a more complete model of the testicular niche. This will have wide ranging implications in disease modelling, as well as in vitro fertility research and therapies.

25 In summary, previous studies using a variety of techniques to differentiate gonad-like cells have taught us that more thorough characterisation of the resulting cell populations is necessary for future studies. Furthermore, a differentiation protocol that closely recapitulates the signalling factors pre-gonadal cells are exposed to may yield a more successful outcome. Development of these protocols will inform the signalling pathways and conditions important in early gonadogenesis and help to form a new in vitro model for studying DSD.

1.4 Conclusion and project aims DSD conditions, like those discussed above, have a prevalence of 1.7% of live births (1). These conditions are inherently difficult for the individual and their families. Understanding the genetic basis for these conditions can have several benefits. It can inform clinical management and often reduce the need for expensive and invasive diagnostic tests. In some cases, it can help clinicians predict whether there may be later co-morbidities such as cancer or infertility. And further, understanding the biological basis of these conditions can help families come to terms with a condition and help to reduce stigma. Despite advances in our understanding of gonad development and how this contributes to DSD, there are still a significant proportion of cases (up to 60% (2)) that go undiagnosed.

Advances in genetic analysis approaches, such as MPS, have allowed us to uncover a wide spectrum of candidate gene variants that may provide explanations for these disorders. However, at present the validation of such variants is very difficult as there is no good in vitro human model system. The overarching aim of this project is therefore to improve the diagnostic pipeline for DSD cases. My specific aims are:

Aim 1: To identify the molecular diagnoses in a cohort of 46,XX DSDs.

Aim 2: To characterise 46,XX DSD gene variants functionally and identify the role they play in patient phenotypes.

Aim 3: To develop a stem cell-based model for the characterisation of DSD genes, specifically the differentiation of human pluripotent stem cells to gonad lineages.

26

27 Chapter 2: Genetic analysis of 46,XX Disorders of Sex Development

2.1 Introduction As outlined in Chapter 1, 46,XX disorders of ovarian development are a rare subset of DSD with wide-ranging phenotypes. In these genetic females gonadal phenotypes can range from streak gonads to ovotestis to testis. Defining subcategories within 46,XX disorders of ovarian development has been instrumental to our understanding of the genetic basis of each condition; these are 46,XX testicular DSD (T-DSD), 46,XX ovotesticular DSD (OT-DSD) and 46,XX ovarian (gonadal) dysgenesis. 46,XX OT-/T-DSDs are characterised by the activation of the testis pathway in genetic females. This results in a switch from ovarian to testis fate during embryonic development and depending on how forceful this signal is will either cause partial or complete sex reversal, meaning either ovotestes (46,XX OT-DSD) or testes (46,XX T-DSD) are formed respectively. Broadly this developmental switch is caused by gain-of-function variants in testis pathway genes or loss-of-function variants in ovarian pathway genes (reviewed in (64)). 46,XX ovarian dysgenesis is caused by alterations in hormone signalling or folliculogenesis pathways. At present, in the majority of cases with 46,XX disorders of ovarian development the molecular cause remains unknown, highlighting gaps in our understanding of ovarian development.

Professor Andrew Sinclair’s research group at the Murdoch Children’s Research Institute (MCRI) has an international cohort of over 1,000 DSD patients, with phenotypes ranging from 46,XY gonadal dysgenesis and hypospadias to primary ovarian insufficiency and Mayer- Rokitansky-Küster-Hauser syndrome. Identifying a genetic diagnosis for these individuals can be very beneficial; directing clinical management, indicating possible comorbidities and providing a rational explanation for patients and families. Furthermore, by contributing to the understanding of these variations in sex development, we hope to reduce the stigma that is associated with these conditions. We therefore work closely with clinicians both nationally and internationally to identify molecular diagnoses from patient DNA samples. One of the challenges is finding a genetic test platform that can cover the range of causative DNA mutations/aberrations for these varied phenotypes. This led to our development of a targeted DSD gene panel (2) for analysis by MPS. This gene panel includes 64 diagnostic DSD genes and

28 a further 967 candidate DSD genes. This panel does not allow us to easily detect large CNVs. Consequently, we developed an in-house DSD CNV screen using Multiplex Ligation-dependent Probe Amplification (MLPA) that allows us to look for duplications or deletions across 14 of the most commonly implicated DSD genes.

Within our DSD patient cohort we have a group of 34 individuals with 46,XX (ovo)testicular DSDs or closely related phenotypes. These patients have been pre-screened for SRY translocations, which are the most common cause of these conditions. They therefore represent a valuable resource for identifying novel DSD genes and learning about gonadal development. The primary aim of this chapter was to obtain molecular diagnoses or find novel DSD genes in this cohort of 34 undiagnosed 46,XX DSD cases. I used three contrasting genetic approaches: MLPA, MPS sequencing of the targeted DSD gene panel and Whole genome sequencing (WGS). I will present diagnostic findings and show what these teach us about 46,XX DSD phenotypes and sex development more broadly. Furthermore, I will discuss candidate genes/variants and show how cellular assays and mouse models support or exclude their role in the associated DSD case. Finally, by comparing these three approaches I will discuss what this project has taught us about genetic analysis of these rare phenotypes.

2.2 Methods

2.2.1 Patient cohort All patients in our DSD cohort gave informed consent when seeing collaborating clinicians both within Australia and internationally. Ethical approval for this study was obtained from the Human Ethics Committee of the Royal Children’s Hospital, Melbourne, Victoria, Australia (HREC22073). Genomic DNA was extracted from EDTA-blood samples in independent laboratories including the Victorian Clinical Genetics Service (VCGS) and local hospitals. DNA quality was checked using the Agilent gDNA Screen-Tape run on a 2200 TapeStation and concentration was measured using the Broad range DNA quantification Kit (Thermo Fisher, #Q32850) on a Qubit 3.0 Flurometer. Clinical information was provided by clinicians.

29 2.2.2 MLPA analysis An in-house MLPA probe mix was designed by Dr. Thomas Ohnesorg. This consisted of 22 probes targeting 14 genes in which CNVs have been implicated in DSD (Table 2.1). A 1 μM stock of individual MLPA probes was made and stored at -20oC until use. 500 μL of probe mix was made and the MLPA reaction was set up using MRC-Holland reagents (#EK5-FAM) according to the manufacturer’s instructions. Concentration and integrity of patient DNA was checked prior to MLPA using the Invitrogen Qubit™ dsDNA High Sensitivity Assay Kit (Life Technologies, #Q32851) on the Qubit 3.0 Flurometer, 100 ng patient DNA was added to each reaction. MLPA reaction products were sequenced at Melbourne University (Melbourne Translational Genomics Platform) and results were analysed by Dr. Thomas Ohnesorg using GeneMarker AFLP/Genotyping software (version 1.97). Any CNVs identified were confirmed in an independent MLPA analysis or our custom DSD CGH array.

Table 2.1. MLPA probes targeting 14 diagnostic DSD genes.

Gene Size Probe sequence targeted (bp)

GGGTTCCCTAAGGGTTGGAGTTCCGCTTGCACATCTGCACCTG WNT4 87 [Phos]CCTCTGGATCAGGCCCTTGAGTTTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGAGCAGGTCTTCGCCGCCGAGTGCAT CBX2 89 CCTGAGCAAGCGGCTCCGCAAGGTTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGTCCAGCTGTTTGTTGACTGACTGC NR5A1 91 [Phos]CTGACTGTTGAGCTCCTGCTTCAAAATCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGACGTGCGCTACTCACTTCACCGAAATT GATA4 95 GCCCAACCCCTGCTCTGCTTTTGACTTTTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACTGGCCTCACCCGATACTATGTTTCTC SOX3 98 CCATTCACTCCTTGGCTAACTGCAAACTAACTCTAGATTGGATCTTGCTGGC

SOX9 GGGTTCCCTAAGGGTTGGAGGGACCACTCAGCACTACTTTGGTG 99 upstream 1 [Phos]GGGTTGTTGTGTAATTGGTAGTAGATTACGTTGCTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGATCGGGTTTCACATTTCTCCTTCCCAAGGTG FOXL2 102 [Phos]GGAGAAGCAGGAGGTTTGAAAAACAAAAAGCAGTCTAGATTGGATCTTGCTGGC

SOX9 GGGTTCCCTAAGGGTTGGAAAGTGTTCACCCAGGCAAGAGATGC 104 upstream 3 [Phos]CACCTTCTATTCATTTCTGATGCATCTAGTTTCCTagcaTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACATTTCTGATCAGTTGTACTTCATCCTATATCAGCA FGF9 106 CAGCTGCCATACTTCGACTTATCAGGATTCTCTAGATTGGATCTTGCTGGC

30 SOX9 GGGTTCCCTAAGGGTTGGAGTAACTGGAATTCACTCACAGAACATGCTG 109 upstream 2 [Phos]CACTCTTCCTCAACTCAAACTGAGTATCCAGtatcggatTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACCAGCTAGTGGAATTCAAAACACAATTGGTTCTGTTGGCACA Control 1 112 GGGCAACAGAATGCCACTTCTTTAAGTAACTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACAAATGCTGGAGTCTGAACATCAGTACCAAGGAGTACGC NR0B1 114 CTACCTCAAGGGGACCGTGCTCTTTAACCCGGGTATCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACCAACCTAAGCACTGTTAGTCAGATTGATCCCAGCTCCAT Control 2 116 AGAAAGAGCCTATGCAGCTCTTGGACTACCCTATCATCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGCTGCTAGAGGAGCTCACCAAAATTCAAGACCCTTCTC MAMLD1 118 CAAATGAGCTAGATCTTGAGAAGATACTGGGGACGAAGCCTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAgattcagttacCCCAAAAGTGGCATTTCTTTGCAAAGG SOX9 TESCO 120 [Phos]GCTACACAGACAAGGGCTTTAGCTAGAAAAGCaagtcacatacTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACACCATTTGTCAAAGAGGCCTTTGAAGAGACC DMRT2 [Phos]CCTAAGAAACACAGAGAGTGTTTAGTTAAGGACAACCAGAAGTACACATTTATCTAGATTGG 124 ATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGTTTTCCTGCAAGGTGGGCTAATAAAGGGAAGGCTTTCAGGA DHH 126 GCTTACTGGAAAACACTTCCATTCAGAGGGTGATTCACCTCCAGTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGTGTTTTGGCAAAGCTGATTCTGGAGTGCTGGAGGATGACTCATTGT DMRT1 C 128 GTGTGCTTCCAGGTGGCCCTGAGAAGGCAGCAGGCCCAGGTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGAGATCAGTCGTCAACGTTGCAGGGTAACATTGGCTACAAAGACCT SRY 130 ACCTAGATGCTCCTTTTTACGATAACTTACAGCCCTATGCCGTAAGTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACTTGAATGTGAGCCTGCGTCGGAGCCCAGCAGCTTCACAGTCACTCC DMRT1 CGT 134 CATCGAGGAGGACGAGTGAGCAGTGCCTGCTGCCGATGGCGGTTTCTAGATTGGATCTTGCTGGC

GGGTTCCCTAAGGGTTGGACATCAAGACGGAGCAGCTGAGCCCCAGCCACTACAGCGAGCAGCAG SOX9 CAGCA 136 CTCGCCCCAACAGATCGCCTACAGCCCCTTCAACCTCCCACACTATCTAGATTGGATCTTGCTGGC

2.2.3 Custom DSD CGH array A DSD-targeted CGH array (Agilent) was designed by Jocelyn van den Bergen (Reproductive Development group, MCRI); probes were selected to detect CNVs at exon level for 74 genes known to cause DSD. Some SOX9 regulatory regions were also targeted, chr17: 69475000- 69560000 (XYSR/RevSex) and chr17: 70102435-70105514 (TESCO), to detect CNVs at a functional resolution of 500 bp. A whole genome backbone resolution of ~120 kb was maintained. Computational analysis was performed by Jocelyn van den Bergen using CytoGenomics (Agilent), CGH algorithms ADM2 were used at a threshold of 6.0.

31 2.2.4 Targeted DSD gene panel The targeted DSD gene panel (2) was designed by Professor Andrew Sinclair’s research group at Murdoch Children’s Research Institute using Agilent SureDesign software (https://earray.chem.agilent. com/suredesign/). This panel consists of 64 known diagnostic genes for DSD and 967 DSD candidate genes, identified from human and animal studies as well as DSD and steroid pathways. Agilent HaloPlex technology was used for the preparation of samples; details of library preparation and sequencing are outlined in (2). MPS was performed at the Translational Genomics Unit at MCRI/VCGS (using either the Illumina MiSeq, NextSeq500, or HiSeq4000), the Centre for Translational Pathology, The University of Melbourne (using an Illumina HiSeq2500) or at Australian Genomics Research Facility (using an Illumina HiSeq or MiSeq). Bioinformatic analysis was performed using CPipe (208), with modifications outlined in (2).

Variant files were filtered to include rare variants (1000 Genomes Project ≤0.01 and ESP5400 or ESP6500 ≤0.01) in coding regions (excluded intronic, 5’/3’ regions and synonymous variants). Variant frequency in the relevant ethnic population was checked on ExAC and discounted if common (>0.01). Variant frequency within our entire DSD cohort was also tracked, allowing us to identify variants resulting from sequencing error or variants over- represented in ethnic sub-populations. Variants were inspected for coverage depth and read quality, and visually evaluated using the Integrated Genomics Viewer (IGV) (http://www.broadinstitute.org/igv/) if they were flagged for functional analysis. In flagged variants where coverage or read depth was low, Sanger sequencing was used to confirm their presence (see 2.2.9). Variants were classified as pathogenic if the variant had been previously found in a patient with a similar clinical phenotype. Likely pathogenic variants were novel missense variants in known DSD genes that fit the phenotype and are predicted to be damaging by in silico algorithms. Remaining variants were classified as variants of unknown significance (VUS).

2.2.5 LGR5 in vitro assay The human LGR5 open-reading frame was cloned from control human DNA using primers shown in Appendix 1. This was then sub-cloned into the expression vector pCMV to add a FLAG tag at the C-terminal of the gene. The DSD04 LGR5 variant (c.C8T) was introduced by

32 site-directed mutagenesis using the QuikChange II XL Site-directed Mutagenesis Kit (Agilent, #200521) according to the manufacturer’s instructions (see primer sequences in Appendix 1). The wild-type and variant LGR5 expression vectors were transfected into HEK 293-T cells (80% confluent) at a ratio of 1:5 (0.25 μg DNA:125 μg Lipofectamine) with Lipofectamine 2000 (Invitrogen, #11668019). Following 24 hours of transfection cells were washed with PBS, fixed with 4% PFA, permeabilised with 1% Triton-X and then blocked with 2% BSA in PBS. Cells were incubated overnight with primary antibodies (anti-FLAG antibody (1:5000; Sigma, #F9291), actin (1:200; Sigma, #5060)) in 1% BSA. Cells were washed three times and incubated in secondary antibodies (Alexa 488 and Alexa 594 (both 1:1000, Invitrogen)) in 1% BSA/PBS. DAPI was used for nuclear counterstaining. Cells were imaged using the Leica DM1000 microscope.

2.2.6 RXFP2 in vitro assay The cDNA constructs of full-length human RXFP2 containing a FLAG tag at the 5’ end were cloned into the mammalian expression vector pcDNA3.1 (Stratagene) and the p.Ser721fs variant was introduced by site-directed mutagenesis. HEK 293-T cells in 96 well plates were transfected with either wild type or mutant RXFP2 and cAMP was measured as previously described (209). This assay was performed by the Bathgate group (Florey Institute).

2.2.7 WNT9A cloning An 598 bp region (chr1:228109148-228109745) of human WNT9A was cloned from patient DNA (DSD10) (primer sequences in Appendix 1) and introduced into the pGEM T-easy vector using the pGEM T-easy system kit (Promega, #A1360) according to manufacturer’s instructions. Competent bacterial cells were transformed with the WNT9A vector and plated onto Agar plates with Ampicillin (1:500, Sigma, #A1593), pre spread with 50 μl X-Gal and 5 μl IPTG per plate. Transformed colonies were screened using Sanger sequencing to identify if WNT9A patients variants segregated (i.e. on different alleles) or were present in the same colonies (on same allele).

33 2.2.8 CRISPR mutant mice The Wnt9a exon 2/3 KO mouse was generated by Dr. Liang Zhao (Institute for Molecular Bioscience, University of Queensland) and Wnt9a-/- embryos characterised by Dr. Anthony Bird (Hudson Institute for Medical Research, Monash University). The Emx2-p.A94V knock-in (KI) mouse was generated by Dr. Liang Zhao (Institute for Molecular Bioscience, University of Queensland) and Emx2- p.A94V+/+ embryos characterised by Dr. Ella Thompson (Institute for Molecular Bioscience). CRISPR guide RNAs for these mutant mice can be found in Appendix 2.

2.2.8 Whole genome sequencing analysis Concentration and integrity of patient DNA was checked as outlined in 2.2.1. 5 μg of patient DNA was required for the WGS reaction. MPS was performed at the Kinghorn Centre for Clinical Genomics Sequencing Laboratory (Garvan Institute) on an Illumina HiSeq X Ten. Bioinformatics analysis was performed by Ben Lundie (Garvan Institute). inputFastq files were generated using Illumina bcl2fastq (2.15.0.4) and FastQC (0.10.1) software. A Raw Reads Sequencing Project report was generated using Illumina Isaac aligner (01.14.07.17) and Picard CollectWgsMetrics (v1.119) software. Alignment and variant calling was performed using BWA-MEM (V0.7.10-r789), Novosort (V1.03.01), Samtools (V1.1), GATK tools (V3.3-0- g37228af) and Variant Effect Predictor 76 software. Data was assessed for CNVs and large structural rearrangements using ClinSV (Minoche, unpublished) (performed by Ben Lundie) and Schism (performed by Dr. Simon Sadedin, MCRI). Data was analysed for SNVs and INDELs using Seave software (210). Seave query settings for each of the three analyses are shown below:

Analysis 1

 All variants and impacts in candidate 46,XX sex reversal, POI or diagnostic DSD gene list (Appendix 3)  0.5% prevalence  Minimum variant quality = 100  Minimum scaled CADD score = 2 Analysis 2

 Loss of function and high impact variants (SNPs and INDELs)  0.5% prevalence  Minimum variant quality = 100  Minimum scaled CADD score = 2 Analysis 3

34  Compound heterozygous or homozygous recessive analysis  0.5% prevalence  Minimum variant quality = 100  Minimum scaled CADD score = 2 Variants were then filtered and curated in Excel to prioritise likely pathogenic variants. The following settings were used:

1. Impact – remove intronic, synonymous, 3/5’ UTR region, downstream/upstream, inframe deletions/insertions variants 2. ‘Variant samples’ – select patient only 3. Filter by those that have frequency of 0 or 1 in KCCG Genomes AF/KCCG Exomes AF 4. Assess gnomAD, ClinVar freq., constraint scores (EXAC) and IGV quality 5. CADD score > 20

2.2.9 Routine molecular techniques

2.2.9.1 Polymerase chain reaction DNA was amplified using Phusion High Fidelity DNA Polymerase (Life Technologies, #F-530). Reactions were carried out in a total volume of 20 μl, with a final concentration of 0.25 mM for each primer, 2 mM for each dNTP and 0.4 U Phusion per reaction. For plasmid DNA 10 ng was used as a template or for genomic DNA 10-500 ng was used. The PCR reaction included an initial denaturation step (98oC for 30 seconds), followed by 35 cycles of denaturation (98oC for 10 seconds), annealing (55-65oC for 15 seconds) and extension (72oC for 30 seconds) and a final extension at 72oC for 5 minutes. Amplification was performed using a thermal cycler (Applied Biosystems, Veriti). Successful amplification was confirmed by gel electrophoresis.

2.2.9.2 Visualisation of nucleic acids Nucleic acids were visualised by gel electrophoresis in 0.5-2% agarose gels (Promega, #V3125) containing 1X GelRedTM Nucleic Acid Gel Stain (Biotium, #41003). Size of DNA fragments was determined using 200 ng of 1 kb Plus DNA Ladder (Sigma, #11721933001). Gels were visualised using the Syngene G:Box machine and its associated software for image analysis.

35 2.2.9.3 Sanger sequencing Sanger sequencing was used to confirm the presence of variants. Primers (see sequences in Appendix 1) were designed manually and optimised in a Phusion PCR reaction. Phusion PCR was performed to amplify the region of interest in patient genomic DNA (50-100 ng), followed by visualisation on a 1% Agarose gel. PCR product underwent a clean-up step, where 5 μl PCR product was combined with 2 μl ExoProStar (GE Healthcare Life Sciences, #77705) and this was run on the thermal cycler for 15 minutes at 37oC followed by 15 minutes at 80oC. The product underwent Sanger sequencing at the Centre for Translational Pathology (The University of Melbourne) or the Australian Genomics Research Facility. Sequences were aligned and analysed using CLC Main Workbench software.

2.2.9.4 In silico protein structure analysis HOPE analysis was used to analyse the structural and functional consequences of patient variants (211).

2.2.10 Transcription factor binding site analysis The duplicated DNA region downstream of WNT4 was assessed using Roadmap Epigenomics Visualization Hub (https://vizhub.wustl.edu), with tracks for DNase Hypersensitivity sites from fetal ovary, fetal testis and adult ovary turned on. Based on enrichment of DNase I hypersensitive sites from chr1:22,431,801-22,432,600, I analysed this 800 bp region for transcription factor binding sites using the program MATinspector (www.genomatix.de), with core similarity and matrix similarity settings at 1.00.

2.2.11 Whole Exome Sequencing Concentration and integrity of patient DNA was established as outlined in 2.2.1. The Agilent SureselectXT 2 protocol was used for sample preparation, requiring 1 μg of input patient DNA. MPS was performed at the Australian Genome Research Institute on an Illumina HiSeq or MiSeq. Bioinformatics analysis was performed by Katrina Bell (MCRI) using CPipe (208), with modifications outlined in (2). Variants were loaded onto the seqr platform (Broad Institute) for analysis of inheritance mode.

36 2.3 Results

2.3.1 DSD patient cohort This project focused on a cohort of 34 patients with 46,XX disorders of ovarian development (Table 2.2). This includes ten cases of 46,XX ovotesticular DSD (OT-DSD), 13 cases of diagnosed and suspected 46,XX testicular DSD (T-DSD). There were five syndromic 46,XX DSD cases where non-DSD phenotypes were also present, including neurological, skeletal and adrenal phenotypes. A further six cases presented with varied 46,XX DSD phenotypes including virilisation (not including Congenital Adrenal Hypoplasia) and gonadal dysgenesis. These were classified as ‘Other’. This is an international cohort, with patients primarily from Australia but also from Indonesia, Vietnam, Canada, New Zealand and the Netherlands. As some of these patient DNA samples were collected up to 20 years ago, comprehensive phenotypic data is not available for all patients (e.g. DSD12) meaning we cannot know their definitive clinical diagnosis. This also meant that parental DNA was not available for 19 of the 34 samples. For the remaining 15 cases, parental DNA was run alongside patient DNA in genetic analyses where possible to establish mode of inheritance. These patients were pre-screened for the most common cause of 46,XX disorders of ovarian development, SRY translocation. This was performed at their clinic of origin or in our lab via an SRY- based PCR (performed by Jocelyn van den Bergen, MCRI). This is a large cohort for these rare phenotypes, which gives us better power for identifying molecular diagnoses and learning about the genetic factors underlying ovarian development and dysregulation.

37 Table 2.2. Clinical information in a 46,XX DSD cohort.

Patient DSD phenotype Other phenotypes DSD sub- Family Genetic tests ID group DNA performed available

DSD01 Raised as a male; effeminate features; breast development; soft voice; testes T-DSD Singleton Targeted DSD pre-pubertal in size; penis adult in size with pubic hair Tanner stage 5, before gene panel, MLPA testosterone administration.

DSD02 SRY-ve male with small testes. Ultrasound on testes revealed heterogeneous T-DSD Duo - Targeted DSD echotexture with a few subcentimeter hypoechoic foci scattered throughout maternal gene panel, MLPA, both. FSH: 54.52 IU/L (N:1.4-18.1); LH: 20.54 IU/L (N: 1.5-9.3); Testosterone: CombiSNP array 10.4 nmol/L (N:5.8-28).

DSD03 Bilateral testes with immature Sertoli cells, no germ cells. T-DSD Singleton Targeted DSD gene panel, MLPA

DSD04 Complete sex reversal, two testes, penoscrotal hypospadias. T-DSD Singleton Targeted DSD gene panel, MLPA, custom DSD CGH array

DSD05 46,XX male, testicular tissue identified in biopsy. Inguinal testes, micropenis T-DSD Trio Targeted DSD and hypospadias (ambiguous genitalia). gene panel, MLPA

DSD06 46,XX testicular DSD, raised male. T-DSD Trio Targeted DSD gene panel, MLPA

DSD07 46,XX male with bilateral testes, azoospermia, SRY absent. T-DSD Singleton Targeted DSD gene panel, MLPA

38 DSD08 46,XX male; Gonads include seminiferous cords, vasa and epididymus; T-DSD Singleton Targeted DSD Prader stage IV; infertile. gene panel, MLPA, custom DSD CGH array

DSD09 At birth he had penoscrotal hypospadias with chordee and left Normal motor function but T-DSD Trio Targeted DSD cryptorchidism. Based on right-sided testis and genitogram demonstrating a significant learning difficulties gene panel, MLPA, primarily male urethra with a stretched penile length of ~3 cm, decision and psychosocial/emotional CGH microarray made to raise child as a male. He has since entered puberty with tanner delay. stage 4 breasts with substantial glandular tissue. The phallus has regressed ~3.5 cmx 1 cm.

DSD10 Small penis and scrotal hypospadias, gonads not palpable. No Müllerian Possible T- Singleton Targeted DSD structures on ultrasound. No gonadal histology available so it is uncertain if DSD gene panel, MLPA he has 46,XX testicular or ovotesticular DSD.

DSD11 46,XX male with severe hypospadias. Possible T- Singleton Targeted DSD DSD gene panel, MLPA

DSD12 46,XX male. Possible T- Singleton Targeted DSD DSD gene panel, MLPA

DSD13 46,XX male. Possible T- Singleton Targeted DSD DSD gene panel, MLPA

DSD14 46,XX male with ovotestes, severe hypospadias and penoscrotal Developmental delay and short OT-DSD Trio Targeted DSD transposition. stature. gene panel, MLPA

Karyotype is 46,XX with balanced translocation of 10p to 12q.

39 DSD15 Born with ambiguous genitalia, left gonad is normal ovary, right gonad is OT-DSD Duo - Targeted DSD ovotestis, normal uterus. Underwent vaginoplasty, clitero-reduction, right maternal gene panel, MLPA gonadectomy and inguinal hernia repair during infancy.

DSD16 Histopathology showed left ovotestis. Raised LH & FSH and low testosterone, OT-DSD Duo - Targeted DSD spermatozoa. maternal gene panel, MLPA, Chromosome microarray, custom DSD CGH array

DSD17 Ambiguous genitalia, bilateral ovotestes, SRY absent in blood and testes, OT-DSD Singleton Targeted DSD phallus 2 cm at birth, no Müllerian structures, high testosterone, raised as gene panel, MLPA female.

DSD18 46,XX ovotesticular DSD. OT-DSD Trio Targeted DSD gene panel, MLPA, custom DSD CGH array

DSD19 46XX ovotesticular DSD. OT-DSD Duo - Targeted DSD maternal gene panel, MLPA

DSD20 Mild clitoromegaly (1.1 x 0.5 cm) at birth, labial folds were separate and Born extremely premature at Possible Singleton Targeted DSD there were no palpable gonads. Follow up at 10 years revealed significant 24 weeks gestation, mother OT-DSD gene panel, MLPA clitoromegaly (4.0 x 1.3 cm), had entered spontaneous puberty with Tanner received intramuscular stage 3 breast development, pubic hair stage 3. HCG stimulation test showed progesterone injections for 2 doubling of serum testosterone (2.56 nmol/L to 4.62 nmol/L), doubling of months in 1st trimester of basal androstenedione (9.7 nmol/L to 19.6 nmol/L), and basal pregnancy for threatened dihydrotestosterone increased (0.88 nmol/L to 0.97 nmol/L), suggesting abortion. The mother does not there is functional testicular tissue. However pelvic MRI and gonadal biopsies recall that she had virilisation during her pregnancy (no facial

40 suggest she has normal ovaries. Current diagnosis remains unestablished, hair, acne or deepening of could be either 46,XX ovotesticular DSD or partial aromatase deficiency. voice). Bilateral hernia repair in infancy.

DSD21 Ambiguous genitalia, small phallus, separate urethra and vagina (4 cm deep). OT-DSD Singleton Targeted DSD Right gonad is an ovotestis with sporadic germ cells found in tubules, gene panel, MLPA calcifications and many primordial follicles in ovarian part. Left gonad is ovarian tissue with primordial and developing follicles.

DSD22 46,XX ovotesticular DSD. OT-DSD Singleton Targeted DSD gene panel, MLPA

DSD23 46,XX ovotesticular DSD. Left gonad is testis 9 x 6.3 mm, no right testis. OT-DSD Singleton Targeted DSD Suspected uterus 21 x 6.5 mm. gene panel, MLPA

DSD24 Ambiguous genitalia. Sprengel shoulder, absent Syndromic Trio Targeted DSD uterus and left kidney. gene panel, MLPA

DSD25 Virilised genitalia and large ovarian cyst. Born at 24 weeks gestation as a Syndromic Trio Targeted DSD DCDA twin, the twin died 3 days gene panel, MLPA of age with hyperkalaemian. Adrenal insufficiency and mild skeletal anomalies. Possible P450 oxidoreductase deficiency.

DSD26 Fetal ultrasound showed normal male genitalia, in 46,XX karyotype with no Pregnancy was terminated Syndromic Trio Targeted DSD SRY present. gene panel, insufficient DNA for MLPA

DSD27 Ambiguous genitalia. FSH was elevated in the neonatal period at 56.6 IU/L Severe scoliosis. Syndromic Singleton Targeted DSD and LH 9.6 IU/L; suggesting gonadal dysgenesis. Twin sister had no signs of gene panel, MLPA virilisation.

41 DSD28 46,XX male with hypospadias, under developed scrotum with small gonads, Seizure disorder being managed Syndromic Trio Targeted DSD micropenis (stretched penile length 3vcm at 5 years), no evidence of by neurologists. (with gene panel, MLPA Müllerian structures. Gonadal biopsy not yet performed, although HCG tests possible T- show a positive testosterone response to stimulation indicating testicular DSD) tissue is present.

DSD29 Testicular regression. Other Duo - Targeted DSD paternal gene panel, MLPA

DSD30 Dysgenic right gonad. Other Singleton Targeted DSD gene panel, MLPA

DSD31 Vaginal atresia, clitoris hypertrophy, bicornate uterus. Severe facial dysmorphism, Other Singleton Targeted DSD auditory canal stenosis/hearing gene panel, MLPA impairment, microtia, microdontia, brachymesophalangia, normal intellectual development.

DSD32 Perineal hypospadias, impalpable gonads, ovaries and uterus on Other Singleton Targeted DSD laparoscopy. Raised as male. gene panel

DSD33 Raised as female, underwent feminising genitoplasty. Now vaginal stenosis, Other Singleton Targeted DSD no diagnosis established. gene panel

DSD34 Virilised female born with ambiguous genitalia. 17-OH P and Cortisol normal, Other Trio Targeted DSD Testosterone increased (1.3) suggesting there is a testicular source of gene panel, MLPA testosterone. USS revealed gonad in right inguinal canal, suspected uterine remnant. Pelvic MRI: No gonads seen, uterine remnant seen, possible vaginal canal.

FSH, Follicle stimulating hormone; HCG, Human chorionic gonadotropin; LH, Luteinizing hormone; N, normal; USS, Ultrasound scan.

42 2.3.2 MLPA analysis for copy number variants in DSD genes

2.3.2.1 VAMP7 MLPA is a multiplex PCR-based method that allows quantification of gene copy number, allowing us to detect CNVs in patient DNA. MLPA analysis was performed on 31 patients in this cohort (Table 2.2); any CNVs identified were confirmed by a second MLPA reaction. A duplication in Vesicle-Associated Membrane Protein 7 (VAMP7) was found in one patient (DSD14) (Fig. 2.1), and this was found to be maternally inherited. VAMP7 has been previously implicated in hypospadias (212) and it is known to co-localise with estrogen-responsive gene ESR1. Increased VAMP7 activity may enhance ESR1 expression and thus lead to estrogen-responsive genes (ATF3, CYR61, CTGF) being upregulated. Given there is no DSD phenotype reported in the mother I do not think that this CNV can explain the DSD phenotype alone, although it may act as a genomic modifier.

Figure 2.1. A maternally inherited VAMP7 duplication in DSD14. Graphs of relative gene copy number for 20 MLPA-targeted regions. Blue dotted lines indicate the upper and lower limits of normal variation. (A) Duplication of VAMP7 probe in DSD14. (B) Duplication of VAMP7 probe in DSD14-maternal sample. (C) No CNVs were identified in DSD14-paternal sample.

43 2.3.2.2 MLPA identifies duplications in the upstream enhancer region of SOX9 Diagnostic duplications were also identified in the SOX9 upstream enhancer region in four patients (DSD08, DSD16, DSD18 and DSD19) (Figure 2.2). Duplications of the upstream enhancer region of SOX9, which result in ectopic SOX9 expression, have been found in approximately 20 cases of 46,XX T-DSD and three cases of 46,XX OT-DSD (Table 2.3, Fig. 2.3), while deletions within this region result in 46,XY sex reversal due to decreased SOX9 expression (153, 182). Fine- mapping of these CNVs using microarray technology have resulted in the refinement of two enhancer regions, RevSex and XYSR (reviewed in (213)), which when duplicated in XX patients (upregulating SOX9) or deleted in XY patients (suppressing SOX9) results in 46,XX or 46,XY sex reversal respectively. Further investigation of the interactions between these enhancers can help to explain SOX9 regulation in the developing gonad.

The duplications identified here were unrelated, each patient duplication covered different regions of the RevSex and XYSR enhancer regions (Fig. 2.2). Duplications in DSD16 and DSD18 were selected for further analysis based on DNA availability; these were fine-mapped using a targeted DSD CGH array (see Methods 2.2.3, performed by Jocelyn van den Bergen, MCRI). This confirmed the exact region of these duplications (Table 2.3). The targeted DSD CGH array was also run on a small subset of the 46,XX disorders of ovarian development cohort (see Table 2.2), identifying a further SOX9 upstream duplication in DSD08. This duplication was shown to have a 3’ breakpoint laying 5’ of the XYSR MLPA probe, therefore could not be picked up via MLPA.

In DSD16 (Case 1, Table 2.3), the duplicated region localised to chr17: 69,311,111–69,558,832 (154), covering the XYSR and RevSex enhancer regions. The 3’ breakpoint of this duplication was within the RevSex region and could be called to within a 450 bp window. By comparing the minimal region of overlap for this and previous 46,XX duplications that included RevSex (Case 1, Fig. 2.3), we were able to refine the RevSex enhancer region from 42 kb to 24 kb (154). By fine- mapping the duplications in DSD08 and DSD18 (Case 19 and 20 respectively of Table 2.3, Fig. 2.3) we highlighted that neither of these covered the RevSex enhancer region, instead covering the XYSR enhancer region. These defined a minimal critical region of 5.2 kb for the XYSR region and demonstrated that duplications of this region cause both XX and XY sex reversal, therefore it is not a 46,XY-specific enhancer as previously suggested (182). These findings are now being combined with in vitro gene regulation assays to advance our understanding of SOX9 regulation during human gonad development (see Croft et al., (2018), Nature Communications 9:5319).

44 Our MLPA analysis also identified a duplication encompassing all three probes of the SOX9 enhancer region in a 46,XX T-DSD patient (DSD06 (Fig.2.2C)), while the other three duplications covered 1-2 of the SOX9 probes and were in 46,XX OT-DSD cases (DSD16, DSD18, DSD19, Fig. 2.2A, C, D). This suggests a possible correlation between the size or location of the duplication and the DSD penetrance (degree of sex reversal). Table 2.3 summarises all reported cases of 46,XX DSD resulting from SOX9 enhancer duplications and Figure 2.3 shows their approximate genomic regions. Table 2.3 highlights the wide phenotypic spectrum associated with SOX9 duplications, for example gonadal phenotypes range from streak to ovotestis to azoospermic testis. Correlating duplication region with DSD phenotype is difficult. Case 9 and Case 12 show a highly similar region of duplication around the RevSex region and each of these patients classify as 46,XX OT-DSD. However in Case 9 the gonadal tissue is ovarian in the cranial region and testicular in the caudal region, while in Case 10 the left gonad is ovarian and the right is testicular. Meanwhile, duplications covering a similar region of RevSex (Case 2, 3, 4 and 10) but not extending as far in the 3’ direction are found in cases of 46,XX T-DSD, with some of these cases showing complete sex reversal (Cases 2 and 3). Furthermore duplications covering RevSex and XYSR are found in the full spectrum of phenotypes, Case 7 presents with ovary/ovotestis and ambiguous genitalia, while Case 8 has normal male secondary sexual characteristics. And finally, duplications encompassing RevSex/XYSR have been found in 46,XX brachydactyly- anonychia patients (129) (Fig. 2.3, black bars), who present with short fingers or toes (brachydactyly) and nail abnormalities (anonychia) but lack DSD phenotypes, suggesting that SOX9 dysregulation can be modified by other factors (genomic or environmental modifiers) or the 5’ region of these duplications includes a repressor of SOX9 (counteracting enhancer overexpression). Overall, there does not appear to be a correlation between duplication location and DSD penetrance. It seems that like many other gonadal development genes (e.g. NR0B1 and NR5A1), SOX9 is sensitive to changes in dosage of regulatory elements and environmental factors influence how far sex development is tipped towards the testicular pathway. The phenotype of Case 9 in particular suggests that other paracrine signalling pathways, which are differentially expressed in the cranial and caudal gonad regions, interact with SOX9 regulation to decide the sex fate of developing gonadal cells.

45

Figure 2.2. MLPA identifies SOX9 enhancer duplications in four 46,XX (ovo)testicular DSD patients. Graphs of relative gene copy number for 20 MLPA-targeted regions. Blue dotted lines indicate the upper and lower limits of normal variation. (A) Duplication of SOX9 upstream region 1 and 2 probes in DSD16. (B) Duplication of SOX9 upstream region 1 probe in DSD18. (C) Duplication of SOX9 upstream region 1, 2 and 3 probes in DSD06. (D) Duplication of SOX9 upstream region 2 and 3 probes in DSD19. (E) Relative genomic location of the four SOX9 probes upstream of the SOX9 transcription start site.

46 Table 2.3. Clinical presentation of patients with 46,XX (ovo)testicular DSD caused by SOX9 regulatory region duplication.

DSD clinical phenotype

Case Duplication Classifi- External Internal Histology Hormonal profiling Sex of Reference region cation rearing

1 Chr17: 46, XX Right cryptorchidism Seminal vesicles Ovotestis, hemorrhagic High FSH and LH, M (154, 214) (DSD 69,311,111– OT-DSD (atrophic testis 1 normal, azoospermia. corpus luteum and Low T. 16) 69,558,832 mL), left testis 8 mL, corpus albicans, Tanner stage 4 primordial follicles, pubertal Sertoli cell nodules, development. seminiferous tubules with no spermatogenesis.

2, 3 Chr17: 46,XX Case 1 and 2: Normal Normal male genital Atrophic seminiferous M, M (215) 69,491,366– T-DSD male genitalia tract without tubules containing only 69,575,195 without any signs of Müllerian remnants, eosinophilic Sertoli cells undervirilisation. azoospermia. suggestive of testicular dysgenesis, Leydig cell hyperplasia.

4 Chr17: 46,XX Bilaterally Azoospermia. Testes showed M (215) 69,435,809– T-DSD hypotrophic testes calcifications, seminal 69,588,345 (right 15 mm × 7 mm vesicle hypoplasia. and left 18 mm × 12 mm).

47 5 516–659 kb 46, XX Isolated OT-DSD, Azoospermia. M (182) upstream OT-DSD had neither SOX9 hypospadias nor signs of gynaecomastia.

6 259–703 kb 46, XX Ambiguous genitalia, Possible vaginal Left gonad was an - (182) upstream OT-DSD hypospadias and a remnant, midline ovary, right gonad was SOX9 bifid scrotum, intra- urethral opening an ovotestis, consisting abdominal gonads. below hypertrophied of both ovarian and clitoris or hypoplastic testicular tissue with penis. Fallopian tube primitive seminiferous on left side. tubules.

7 Chr17: 46, XX Ambiguous genitalia Rudimentary uterus. Right gonad was an F (182) 69,370,916– OT-DSD with clitoromegaly, ovary, left was either a 69,855,932 vagina. dysgenetic testis or an ovotestis.

8 Chr17: 46,XX Normal male Azoospermia. Low T, high FSH M (125) 69,401,099– T-DSD secondary sexual and LH. 69,878,197 characteristics and bilateral gynecomastia.

48 9 Chr17: 46, XX Ambiguous genitalia Absent uterus, vaginal Caudal region of both At 8 months: low - (125) 69,510,367– OT-DSD with hypertrophic atresia, and two gonads were testicular LH and T, normal 69,764,059 clitoris, single ovoidal gonads tissue with prepubertal FSH. Testosterone meatus, and detected in the seminiferous tubules, response to hHCG urogenital sinus. inguinal canal. whereas the cranial administration was region showed ovarian low (76 ng/dl). tissue with oocytes.

10 Chr17: 46,XX Hypotrophic testes, Azoospermia. High FSH and LH, M (216) 69,533,305– T-DSD hypospadias, other low T. 69,606,825 secondary sexual characteristics were normal.

11 Chr17: 46, XX Bifid scrotum, Epididymal structures, Ovotestes. M (153) 69,069,079– OT-DSD incurved short penis bilateral fallopian 69,764,059 with hypospadias. tubes.

12 Chr17: 46, XX Perineal Fallopian tubes, Left gonad is ovary with M (153) 69,521,863– OT-DSD hypospadias, rudimentary vagina fallopian tube, right 69,670,036 asymmetric scrotum. and uterus. gonad is testis.

49 13 Chr17: 46, XX Incurved penis with Vagina, uterus. Left gonad is an F (153) 68,829,028– OT-DSD perineal ovotestis with 69,609,453 hypospadias, epididymal structure asymmetric scrotum. and fallopian tube, Right side is a streak gonad partially differentiated toward ovary.

14, 778–600 kb 46,XX Normal male Azoospermia. Leydig and Sertoli cells, M, M, M (81) 15, upstream T-DSD secondary sex severely diminished and 16 SOX9 characteristics. atrophied seminiferous tubules, and no spermatogenesis.

17, Chr17: 46,XX Bilaterally Germinal cell aplasia. Low T, high FSH M (123) 18 69,507,344– T-DSD hypotrophic testes and LH. 69,603,142 and mild bilateral triplication gynaecomastia.

19 Chr17: 46,XX Prader stage IV. Azoospermia. Gonads include Croft et (DSD 69,458,883– T-DSD seminiferous cords, al., 2018 08) 69,482,850 vasa and epididymis.

20 Chr17: 46, XX 46,XX ovotesticular Croft et (DSD 69,475,275 – OT-DSD DSD. al., 2018 18) 69,499,520

NB: DSD cases from our cohort are shown in brackets in the first column.

50 Figure 2.3. Representation of duplications identified upstream of SOX9 in 46,XX individuals. Reported duplications (and one triplication) upstream of SOX9 resulting in 46,XX ovotesticular DSD (orange bars) or 46,XX testicular DSD (blue bars). Two proposed SOX9 enhancer regions are shown, RevSex (green region) and XYSR (red region), located approximately 500 kb upstream of the SOX9 transcription start site. Black solid bars represent duplications in 46,XX brachydactyly-anonychia patients who lack DSD phenotypes (129).

51 2.3.3 Targeted DSD gene panel Following analysis with MLPA (and in some cases the targeted DSD CGH array) there were 29 patients without a molecular diagnosis. To identify potentially causative genetic variants patient DNA was sequenced on our targeted DSD gene panel. This gene panel contains the 64 diagnostic DSD genes, four of which are associated with 46,XX (ovo)testicular DSDs (in SNV form), WT1, WNT4, RSPO1 and NR5A1 (2). This targeted gene panel, unlike MLPA or arrays, will identify small genetic changes such as SNVs or changes up to 20 bp in size. Variant filtering (outlined in 2.2.4) resulted in 10-50 variants per patient. These variants were then curated and classified using the American College of Medical Genetics and Genomics (ACMG) guidelines as ‘pathogenic’, ‘likely pathogenic’, ‘Variant of uncertain significance - VUS’, ‘likely benign’ or ‘benign’ (217). From this analysis, the same pathogenic variant was identified in three patients (DSD02, DSD17 and DSD28), this was a heterozygous missense variant in the NR5A1 gene (p.Arg92Trp). In addition, a likely pathogenic variant was identified in NR5A1 (p.Ala260Val) in DSD21. This variant had not previously been identified in 46,XX DSD. In vitro characterisation of these NR5A1 gene variants is presented in Chapter 3.

The remaining variants were classified as VUS. Further curation of these included examining mutant mouse reproductive phenotypes and searching the literature for possible links to gonad development. Six candidate genes are discussed here, EMX2, FOXL2, Leucine-rich Repeat- containing G Protein-coupled Receptors 5 and 8 (LGR5 and LGR8/RXFP2), Wingless-type MMTV integration site family, member 5A (WNT5A) and WNT9A (Table 2.4).

52 Table 2.4. Candidate variants from the targeted DSD gene panel.

ExAC constraint scores

Gene DSD ID Location dbSNP ID DNA change Protein Mutation Zygosity pLI z score gnomAD in change type (inheritance) (ExAC) freq. silico

EMX2 DSD01 Chr10:119303056 NA NM_001165924:c.C p.Ala93Val missense het (NA) 0.9 3.02 NA 2/3 278T 4

FOXL2 DSD34 Chr3:138665354 rs75757121 NM_023067.3:c.G2 p.Ala71Pro missense het (paternal) 0.8 NA 0.0000203 1/3 9 11C 8 5

LGR5 DSD04 Chr12:71833868 NA NM_003667:c.C8T p.Thr3Ile missense het (NA) 0 -1.48 0.0000649 2/4 9 (South Asian)

LGR8/ DSD32 Chr13:32376439 NA NM_130806:exon1 p.Ser721fs frameshift het (NA) 0 NA NA NA 8:c.2162_2165del RXFP2 WNT5A DSD34 Chr3:55504151 NA NM_003392.4:c.C1 p.Thr371Met missense het (de novo) 0.9 2.46 0.0000081 3/4 112T 7 23

WNT9A DSD10 Chr1:228109381 rs14440008 NM_003395:c.G936 p.Arg312Ser missense het (NA) 0.6 1.36 0.0005029 2/4 0 T 8 (African)

Chr1:228109506 rs14838438 NM_003395:c.C811 p.Arg271Trp missense het (NA) 0.0005438 2/4 2 T (African)

In silico: PolyPhen2, MutationTaster, SIFT, LRT; fraction indicates proportion predicted deleterious or possibly deleterious. pLI, Probability of LoF intolerance; NA, not available.

53 2.3.3.1 A candidate variant in ovarian regulator FOXL2 DSD34 presented with ambiguous genitalia, uterine remnant and possible vaginal canal, with normal levels of 17-OH P and Cortisol but increased Testosterone, suggesting the presence of testicular tissue (Table 2.2). In this patient I identified a missense variant in exon 1 (c.211G>C, p.Ala71Pro) of FOXL2, corresponding to the highly conserved DNA binding domain (Fig. 2.5 A, B). There is only one FOXL2 transcript (NM_023067.3) and this shows high expression specifically in human ovarian tissue (GTEX). This variant showed a heterozygous paternal inheritance pattern, which is being validated separately in follow up whole exome sequencing (WES), due to low MPS read depth. It also showed low frequency in gnomAD database (2.04E-05), and in the ExAC database this variant has not been identified in females, only in males at a frequency of 1.49E-05. This variant was of further interest because of the known role of FOXL2 in ovarian development.

Foxl2 plays an essential role in ovarian differentiation (55); it is expressed specifically in the ovarian granulosa and interstitial cells from E12.5 (Fig 2.5D, E). FOXL2 continues to be expressed postnatally, actively repressing testis elements (e.g. Dmrt1) so that the ovarian fate is maintained (54, 58). Mutations in Foxl2 cause sex reversal (or polledness) in XX goats (52). In humans, over 130 heterozygous missense variants in FOXL2 have been associated with human BPES syndrome, characterised by primary ovarian insufficiency and eyelid malformations (56). Based on external phenotype and hormonal profiling, DSD34 may have ovotestes, however a gonadal biopsy would be needed to confirm this. No virilisation or testicular development has yet been reported in 46,XX individuals with FOXL2-associated BPES, however they do present with small uterus and atrophic ovaries, as described in DSD34.

A large majority of BPES-associated FOXL2 variants localise to the DNA binding domain of FOXL2. The FOXL2-p.Ala71Pro variant in DSD34 is also found in this domain, located in a mutational hotspot (Fig. 2.5A) indicating that this region is intolerant to variation. A pathogenic amino acid change at the next residue (p.Ser70Ile) (Human Gene Mutation Database) suggests that this region is essential for protein function. Indeed, protein modelling (Fig. 2.5C) revealed that the residue change from Alanine to Proline at position 71 introduces a residue of greater mass, and given its location on the surface of the protein, this may impact binding to target genes and disturb interactions with other molecules or other parts of the protein. A recent study characterised the effects of variants at neighbouring residues 69 and 70, showing that they

54 result in mislocalisation, protein aggregation and impaired transactivation (218). Applying these cellular assays to the FOXL2-p.Ala71Pro variant could confirm its pathogenicity and determine whether it may underlie the phenotype observed in DSD34.

In addition to the FOXL2 variant, I identified a heterozygous missense variant in WNT5A (NM_003392.4:c.C1112T:p.Thr371Met) in DSD34. This variant shows a de novo inheritance pattern, has a low gnomAD frequency (8.12E-06) and is predicted highly damaging (CADD score = 26.8). Heterozygous variants in WNT5A have been implicated in Robinow syndrome, a rare condition characterised by dysmorphic facial features, hypoplastic external genitalia (in males), renal and vertebral anomalies (OMIM). The WNT5A gene encodes a WNT protein involved in canonical WNT signaling. Given that the WNT pathway and FOXL2 are the key regulators of ovarian development, it is possible that these two variants act in combination to disrupt sexual differentiation in DSD34.

Figure 2.4. The targeted DSD panel identifies a paternally inherited missense variant in FOXL2 in DSD34. IGV visualisation of the NM_023067.3(FOXL2):c.G211C:p.A71P variant in DSD34. Total allele count = 7, C allele count = 4, G allele count = 3, N (unknown) allele count = 0.

55

Figure 2.5. A heterozygous missense variant in FOXL2 is located in the highly conserved DNA binding domain. (A) Schematic of human FOXL2 major protein domains indicating approximate position of the DSD34 variant (black box) in the DNA binding domain. BPES-associated variants (red box) in close proximity highlight this area as a mutational hotspot. DBD; DNA binding domain; Poly-ALA, Poly alanine domain. (B) The location of the variant in the FOXL2 protein sequence is highly conserved among mammalian species. (C) Predicted structural consequences of the variant in the FOXL2 protein from HOPE analysis (211). FOXL2 protein is shown in grey, with the side chains of the wild-type and variant residues shown in green and red respectively. (D) Graph of Foxl2 expression in mouse gonad between stages E10.5-13.5. Data is from (219). Data points represent the mean of the raw counts per million. Error bars show the standard deviation. (E) Graph of Foxl2 expression in mouse gonad cell lineages between stages E11.5- 13.5. Data points represent log-transformed, normalised intensity values from microarray

56 analysis (220). Colours represent different cell lineages, with XX shown as dashed lines and XY shown as solid lines. Error bars represent standard error.

The remaining four candidate variants from the targeted DSD gene panel have been followed up with functional studies, discussed below.

2.3.4 Candidate variants from the targeted DSD panel – LGR genes Two variants were identified in LGR genes, LGR5 and LGR8/RXFP2 in patients DSD04 and DSD32 respectively (Table 2.4). The LGR genes are a family of seven-transmembrane receptors. They share a similar membrane-spanning structure, composed of a leucine-rich, ligand-binding extracellular domain and a short cytoplasmic domain that binds to accessory molecules (221). These proteins have a well understood mechanism for signal transduction whereby binding of hormones to the ectodomain results in a conformational change that triggers activation of cAMP and induction of intracellular signalling pathways (222). A role in the developing gonad has been established for several LGR proteins, and one member of this gene family, LHCGR, has been associated with DSD (Leydig cell hypoplasia (223)).

In DSD04 I identified a heterozygous missense variant in LGR5 (Table 2.4) (Fig. 2.6). This patient presented with complete sex reversal (46,XX T-DSD), two testes and penoscrotal hypospadias (Table 2.2). This case was a singleton, therefore inheritance mode of this variant is unknown. The variant had a low frequency in the relevant population cohort on gnomAD database (0.00006499, South Asian) and was predicted damaging by in silico predictors (2/4). This variant is located in the highly conserved signal peptide region (Figure 2.7A, B); a change from Threonine to Isoleucine at residue 3 introduces a larger and more hydrophobic residue at this site. This region of the peptide is important as it is recognised and cleaved by other proteins to generate the mature LGR5 protein. The introduction of the variant residue is predicted to disturb recognition of the signal peptide, potentially resulting in a loss of hydrogen bonds and/or disturbing correct folding (HOPE analysis (211)).

Lgr5 is specifically expressed in the developing ovary from E11.5 onwards (Fig. 2.7C, D) and plays a central role during the differentiation of pre-granulosa and germ cells in the XX mouse gonad

57 (224). LGR5 acts as a cell surface receptor for R-spondins. Binding of R-spondins to LGR5 induces canonical WNT signalling through LRP5/6 complexes (159, 225), the activation of this pathway is essential for normal ovarian development. Lgr5-/- KO is neonatal lethal in mice (226), however Lgr5+/- mice show a phenotype similar to that of Rspo1 KO, with subtle changes in ovarian structure and tongue malformations (Dr. Dagmar Wilhelm, personal communication, University of Melbourne). In humans RSPO1 is a diagnostic gene for syndromic 46,XX (ovo)testicular DSD (Table 1.1).

Given the location of this variant in the signal peptide of LGR5, I wondered whether this variant may affect LGR5 cell surface localisation and thus recognition of R-spondins. I assessed this by looking at protein localisation and expression in wild-type and mutant LGR5 (Fig. 2.8). I generated expression vectors with the human LGR5 open-reading frame both in wild type and with the DSD04 variant, with a FLAG tag at the C-terminal (outlined in 2.2.4), these were transfected into HEK 293-T cells. In cells transfected with wild type LGR5 I observed staining of intracellular puncta in the cytoplasm and perinuclear regions (Fig. 2.8 A, B), as shown previously (227, 228). In mutant LGR5 the same pattern of localisation and expression was observed (Fig. 2.8C, D), indicating that the variant is not having a deleterious effect on the LGR5 signal peptide. As such, this variant was considered likely benign in this DSD case.

58

Figure 2.6. A heterozygous missense variant identified in LGR5 in DSD04 via the targeted DSD gene panel. IGV visualisation of the NM_003667(LGR5):c.C8T:p.T3I variant in DSD04. Total allele count = 262, C allele count = 150, T allele count = 112, N (unknown) allele count = 0.

59

Figure 2.7. A missense variant in LGR5 is located in the highly conserved signal peptide region. (A) Schematic diagram of human LGR5 major protein domains indicating approximate position of DSD04 variant. (B) The location of the p.Thr3Ile variant in the LGR5 protein sequence is highly conserved among mammalian species. (C) Graph of Lgr5 expression in mouse gonad between stages E10.5-13.5. Data is from (219). Data points represent the mean of the raw counts per million. Error bars show the standard deviation. (D) Graph of Lgr5 expression in mouse gonad cell lineages between stages E11.5-13.5. Data points represent log-transformed, normalised intensity values from microarray analysis (220). Colours represent different cell lineages, with XX shown as dashed lines and XY shown as solid lines. Error bars represent standard error.

60

Figure 2.8. Protein localisation and expression of variant LGR5 is unaffected. Protein expression of variant and wild type LGR5 was assessed in HEK 293-T cells with an anti-FLAG antibody (green). Cells were transfected with an equal amount of FLAG-tagged LGR5 expression vector. Nuclear counterstaining was performed with DAPI (blue) and cytoskeleton was stained with an Actin antibody (red). (A, B) Wild type LGR5 was expressed in both the nucleus and cytoplasm, with presence of punctates. (C, D) The variant LGR5 protein expression and localisation was unaffected.

In addition, I identified a heterozygous deletion in another member of the LGR protein family, LGR8 or RXFP2 (Table 2.4, Fig. 2.9A, B). This was identified in DSD32, an individual with 46,XX DSD with virilisation, perineal hypospadias, impalpable gonads, with uterus and ovaries (Table 2.2, Fig. 2.7). This 4 bp deletion in exon 17 causes a frameshift that results in replacement of the last 27 amino acids in the C-terminal domain of the RXFP2 protein with nonsense protein sequence. This patient was a singleton so the inheritance mode is unknown, however this variant was absent from the gnomAD database.

61 This variant was located in the C-terminal intracellular domain (Fig. 2.10A), conserved among mammalian species (Fig. 2.10B). This region responds to ligand-binding which leads to downstream initiation of intracellular signalling pathways (222). RXFP2 is an important regulator of testicular descent via INSL3 binding (229). Indeed, a heterozygous missense variant (p.Thr222Pro) in RXFP2 has been found in a case of human cryptorchidism, however the variants presence in unaffected individuals significantly weakens this genetic link (230). In the mouse, Rxfp2 is highly expressed in the XX gonad from E10.5-12.5 (Fig. 2.10C, D) and in the Soay sheep variation at the Rxfp2 locus has been associated with XX sex reversal (polledness) (231). Until now, the role of RXFP2 in ovarian development and differentiation has not been assessed; I wondered whether this RXFP2 frameshift variant might affect response to ligand binding.

To investigate the function of this variant RXFP2 in vitro we collaborated with the Bathgate group (Florey Institute) as they have a well-established RXFP2 assay (209). An RXFP2 expression vector was cloned based on the full-length receptor (NM_001166058), with site directed mutagenesis being used to generate the frameshift variant. HEK 293-T cells were transfected with wild type or variant RXFP2 and treated with a dose response of the ligand INSL3 (methods outlined in 2.2.6)(Fig. 2.11). Intracellular cAMP was measured and calibrated to the forskolin response, this showed that cells transfected with variant RXFP2 displayed an almost identical response to wild type RXFP2. The variant RXFP2 receptor thus maintains its response to ligand binding at levels equivalent to wild type. It is possible that there are differences in signalling in the native (ovarian) cells or that the variant protein has impaired trafficking to the cell surface. However based on these data this variant was considered likely benign and not investigated further.

62

Figure 2.9. A heterozygous frameshift variant identified in RXFP2 in DSD32 via the targeted DSD gene panel. (A) IGV visualisation of the NM_130806:exon18:c.2162_2165del:p.S721fs variant in DSD32. Total allele count = 155, A allele count = 73, G allele count = 82, N (unknown) allele count = 0. (B) Sanger sequencing was performed on DSD32 to confirm the presence of the NM_130806:exon18:c.2162_2165del:p.S721fs variant (black dotted box).

63

Figure 2.10. A heterozygous frameshift variant in RXFP2 is located in the C-terminal intracellular domain. (A) RXFP2 protein structure showing the location of the p.S721fs variant in the intracellular domain. SP; signal peptide; IC, C-terminal intracellular domain. (B) The location of the variant in the RXFP2 protein sequence shows conservation among human, mouse and macaque. (C) Graph of Rxfp2 expression in mouse gonad between stages E10.5-13.5. Data is from (219). Data points represent the mean of the raw counts per million. Error bars show the standard deviation. (D) Graph of Rxfp2 expression in mouse gonad cell lineages between stages E11.5-13.5. Data points represent log-transformed, normalised intensity values from microarray analysis (220). Colours represent different cell lineages, with XX shown as dashed lines and XY shown as solid lines. Error bars represent standard error.

64

Figure 2.11. Testing receptor signalling function of variant RXFP2 compared to wild type RXFP2. HEK 293-T cells were transfected with wild type and variant RXFP2 cDNA and treated with increasing levels of Forskolin. Intracellular cAMP was measured and this was calibrated to the Forskolin response, indicating that wild type and variant RXFP2 exhibit the same response. Data points represent the mean and standard error of data from at least three independent experiments performed in triplicate. This assay was performed by the Bathgate group at The Florey Institute.

2.3.5 Candidate genes from the targeted DSD gene panel – CRISPR mouse models for Wnt9a and Emx2

2.3.5.1 WNT9A Two heterozygous missense variants were identified in the WNT9A gene in patient DSD10 (Table 2.4) (Fig. 2.12A). DSD10 presented with a small penis with scrotal hypospadias, impalpable gonads and absent Müllerian structures. Gonadal histology was not assessed so he was classified as possible 46,XX T-DSD, however the presence of ovotestes cannot be ruled out. This individual was of African ethnicity, therefore gnomAD frequencies from African populations were taken into account, and this was low (0.0005) for both WNT9A variants. These variants were each predicted damaging by in silico algorithms (2/4) and this gene showed a z missense constraint score of 1.36 (ExAC), indicating that this gene is intolerant to missense variants (232). These two variants are located in a region important for binding of other molecules that shows high conservation among eutherian species (Fig. 2.13B, C). The lack of sequence conservation with chicken (Fig. 2.13B, C) suggests that WNT9A may have a divergent structure or function in other vertebrates. The p.Arg271Trp variant introduces an amino acid that is more hydrophobic, larger and differentially charged (positive to neutral charge) (Fig. 2.13A), which can result in loss of interactions with other molecules, loss of hydrogen bonding and disrupted folding (211).

65 Similarly, the p.Arg312Ser variant introduces an amino acid that is smaller, more hydrophobic and differentially charged (positive to neutral charge) (Fig. 2.13A), this is predicted to disturb interactions with other molecules due to disrupted folding or hydrogen bonding (211).

Parental DNA was not available for DSD10 therefore I could not establish whether these variants were compound heterozygous or present on the same allele from MPS sequencing data. However as these variants are in close proximity, I was able to clone a 598 bp region encompassing the two variants from patient DNA into the pGEM vector (see 2.2.7). This was introduced into bacterial cells and colonies were screened for the presence of WNT9A variants by Sanger sequencing. The two variants did not segregate in colonies screened (Fig. 2.12C), indicating that they are located on the same allele. The inheritance scenario could therefore be that one variant was inherited from one parent and the other arose de novo or that one parent also harbours these same two variants. Getting access to parental DNA in future would be beneficial to understand this scenario better.

WNT9A encodes a secreted protein that binds to surface receptor Frizzled and regulates WNT/β- catenin signalling. Wnt9a is highly expressed in the developing ovary, becoming sexually dimorphic as the gonads begin to differentiate (Fig. 2.13D); this is almost identical to Wnt4 expression in ovarian differentiation (233). Wnt4/Wnt9a have a synergistic relationship in the developing joints (234, 235). Furthermore, Wnt9a is downregulated in Rspo1-/- XX gonads (38); therefore it may be regulated by key ovarian factor RSPO1. Based on this, I hypothesised that Wnt9a acts synergistically with Wnt4 to promote ovarian differentiation and/or suppress the testis pathway. Given the evidence indicating potential pathogenicity of WNT9A variants in DSD10 and the unexplored role of WNT9A in ovarian development, I wanted to characterise a KO mouse for Wnt9a to test whether the ovaries are disrupted.

A mutant mouse model has been made previously for Wnt9a, where exon 2 was knocked down (235), this phenotype was neonatal lethal with skeletal abnormalities, however gonadal phenotype was not assessed. For our study we collaborated with Prof. Peter Koopman’s (Institute for Molecular Bioscience, University of Queensland) and Prof. Vincent Harley’s (Hudson Institute of Medical Research, Monash University) groups who have expertise in mouse modelling. CRISPR guide RNAs were designed to induce breaks in the DNA and consequent

66 deletion of exons 2 and 3 (Appendix 4, Methods 2.2.8). These Wnt9a-/- mice are currently being bred so that the mutant phenotype can be examined.

Figure 2.12. The targeted DSD panel identifies two missense variants in WNT9A in DSD10. (A) IGV visualisation of the NM_003395 (WNT9A): c.G936T:p.R312S variant in DSD10. Total allele count = 156, A allele count = 51, G allele count = 105, N (unknown) allele count = 0. (B) IGV visualisation of the NM_003395 (WNT9A): c.C811T:p.R271W. Total allele count = 121, A allele count = 66, C allele count = 55, N (unknown) allele count = 0. (C) Sanger sequencing confirms

67 WNT9A variants c.G936T (green box) and c.811T (red box) are on the same allele in DSD10. Alignment of sequences of human WNT9A genomic DNA and cloned colony with pGEM-WNT9A plasmid.

Figure 2.13. Two heterozygous missense variants in WNT9A located in a conserved binding domain. (A) Schematic structures of the original (left) and variant (right) amino acids. The backbone is coloured red, the side chain, unique for each amino acid, is coloured black. (B, C) The location of the variants in the WNT9A protein sequence are highly conserved among eutherian species. (D) Graph of Wnt9a expression in mouse gonad between stages E10.5-13.5. Data is from (219). Data points represent the mean of the raw counts per million. Error bars show the standard deviation.

68 2.3.5.2 EMX2 A heterozygous missense variant was also identified in the EMX2 gene of patient DSD01 (Table 2.4, Fig. 2.14). DSD01 presented with 46,XX T-DSD, was noted to have normal male genitalia, pre-pubertal sized testes and gynecomastia before Testosterone treatment (Table 2.2). Histology has not been performed on patient gonads and as the patient has undergone testosterone treatment we do not know if they have functional Sertoli cells. The EMX2 variant is located in exon 1, corresponding to the Homeobox protein chain region, which is highly conserved among amniote species (Fig. 2.15A, B). This variant is found in two of three transcripts for human EMX2 (NM_001165924, NM_004098), including the transcript most abundantly expressed in human ovary and testis (NM_004098) (GTEX). HOPE analysis revealed that the change from an Alanine to a Valine at residue 93 might alter the conformation of EMX2 (211). This patient was a singleton so inheritance is unknown, however this variant has not been reported in humans before, being absent from gnomAD and ClinVar databases. EMX2 has an ExAC missense constraint z score of 3.02, indicating this gene is highly intolerant to missense variants (232).

Emx2 is essential for the development of the bipotential gonads, with Emx2-null mice showing absence of gonads, genital tracts, kidneys and ureters (36). Emx2 is highly expressed in the bipotential gonad then becomes specifically upregulated in the granulosa and interstitial cells of the mouse ovary by E12.5 (Fig. 2.15C, D), suggesting a functional role for Emx2 in ovarian differentiation. EMX2 haploinsufficiency is a rare cause of DSD; 10q deletions have been identified in approximately 15 46,XY individuals (reviewed in (236)), with genital anomalies (including sex reversal in two cases) reported in all of these cases alongside other syndromic phenotypes. One example is a 3.8 Mb microdeletion encompassing EMX2, found in a 46,XY individual with partial testicular dysgenesis (Leydig and Sertoli cell function deficiency), micropenis, hypospadias as well as developmental delay, absent left kidney and scoliosis (237). SNVs in EMX2 have not been associated with DSDs and no 46,XX DSDs have been attributed to EMX2. To investigate whether this EMX2 variant in DSD01 was indeed disrupting ovarian development, we generated a CRISPR KI mouse harbouring the equivalent variant (Emx2.NM_010132(3 exons):c.281C>T(p.A94V)) (see Methods 2.2.8 and Appendix 5).

69

Figure 2.14. The targeted DSD panel identifies a missense variant in EMX2 in DSD01. (A) IGV visualisation of the NM_001165924 (EMX2):c.C278T:p.A93V variant in DSD01. Total allele count = 26, C allele count = 7, G allele count = 1, T allele count = 18, N (unknown) allele count = 0.

70

Figure 2.15. A heterozygous missense variant in EMX2 is located in a highly conserved region of exon 1. (A) Schematic representation of EMX2 indicating location of DSD01 variant. Boxes indicate coding regions; grey shading indicates the region corresponding to the EMX2 homeodomain. (B) The location of the variant in the EMX2 protein sequence is highly conserved among amniote species. (C) Graph of Emx2 expression in mouse gonad between stages E10.5- 13.5. Data is from (219). Data points represent the mean of the raw counts per million. Error bars show the standard deviation. (D) Graph of Emx2 expression in mouse gonad cell lineages between stages E11.5-13.5. Data points represent log-transformed, normalised intensity values from microarray analysis (220). Colours represent different cell lineages, with XX shown as dashed lines and XY shown as solid lines. Error bars represent standard error.

Mice were generated that were homozygous for the Emx2-p.A94V variant. We assessed expression of testis and ovarian markers on E12.5 gonads of both wild-type (wt) and KI mice (Fig. 2.16). There was no change in Emx2 expression between KI and wild-type, indicating that Emx2 transcription is unaffected by this variant. No significant changes were observed in ovarian marker Foxl2 or testis markers Amh, Hsd3b1 and Nr5a1. There was a small but significant decrease (p<0.05) in Sox9 expression in Emx2 KI ovary, however it is unlikely that this would cause a functional change. Immunofluorescence staining of XX gonads revealed no obvious morphological changes (Fig. 2.17G, H); KI and wt ovaries showed the same staining pattern for FOXL2+ granulosa cells and no evidence of testicular somatic cells (SOX9-/AMH-) (Fig. 2.17D, H).

71 Wild-type XX ovaries were negative for Leydig cell marker 3βHSD+ (Fig. 2.17C) yet in KI XX ovaries we observed a number of 3βHSD+ cells (Fig. 2.17D, see arrows). In immunofluorescence staining of XY KI gonads there was no change in expression of Sertoli cell markers SOX9 and AMH and the testis cords appeared normal (Fig. 2.17B, F). However interestingly, we again observed an increase in 3βHSD+ Leydig cells in KI testes compared to wild-type (Fig. 2.16A, B, see arrow). This staining pattern suggests that in both XX and XY gonads the Emx2 variant enhances the male steroidogenic pathway. The presence of Leydig cells in an XX gonad could result in masculinisation of an XX reproductive system, for example ectopic Dhh expression in mouse ovary results in fetal Leydig cell differentiation, external virilisation and ovarian descent (238).

However, complicating these findings, 3βHSD+ cells were also observed in the embryonic kidney (Fig. 2.17D, see arrows), a structure in which steroidogenesis does not occur. It is possible that this represents tissue-specific non-specific staining, as parts of the kidney show high levels of auto-fluorescence (Dr. Alex Combes, personal communication, University of Melbourne). Alternatively, this may indicate that the 3βHSD staining in all tissues of this experiment is non- specific rather than the male steroidogenic pathway being activated in the KI gonads. Indeed, we saw no significant increase of Hsd3b1 in qRT-PCR screening of KI XX/XY gonads (Fig. 2.16). Furthermore, no masculinisation of KI XX mice was observed in later embryonic time points (data not shown, Ella Thomson, personal communication, University of Queensland).

To better understand these findings we will repeat qRT-PCR and immunofluorescence staining at this time point and at a bipotential gonad stage (E11.5). We will also perform co-staining of gonads with 3βHSD and additional Leydig markers (e.g. CYP217, SF1, STAR) to prove or disprove presence of Leydig cells in XX KI ovaries. In Emx2 KO gonads the Epidermal Growth Factor Receptor (Egfr) gene is highly induced (6), resulting in the persistence of tight junctions between cells and inhibiting epithelial-to-mesenchymal transition during early gonad development. We will also look for changes in Egfr expression in KI gonads via qRT-PCR.

We know that mutant mouse models often show a milder or absent gonadal phenotype compared to humans. Given the possible functional change observed in this mouse model is so minor, there may be additional genomic modifiers contributing to the phenotype of this patient.

72 This patient was therefore included in a pilot study using WGS to identify DSD variants (see 2.3.6).

Figure 2.16. Gene expression in Emx2-p.A94V+/+ mouse gonads. qRT-PCR data showing relative gene expression levels of markers of the testis and ovary in E12.5 mouse ovaries (XX) or testes (XY) from wild-type (wt) and Emx2 knock-in strains. Each sample represents a biological triplicate (Mean ± S.E.M). A student’s t-test was applied to obtain p-values, *p<0.05. Gene expression is quantified relative to reference gene Tbp.

73 Figure 2.17. Immunofluorescence staining in Emx2-p.A94V+/+ mouse gonads. Immunofluorescence staining of E12.5 mouse testes (A, B, E, F) and ovaries (C, D, G, H) with antibodies for Sertoli cells (AMH, green or SOX9, red), Leydig cells (3βHSD, red) and granulosa cells (FOXL2, green), with DAPI (blue) marking cell nuclei.

74 2.3.6 Whole genome sequencing With the genetic analyses discussed so far, diagnoses were made in known DSD genes (SOX9, NR5A1) across nine cases (9/34, 26%), and candidate variants or genes (EMX2, WNT9A and FOXL2) are being investigated with functional assays in a further three cases (3/34, 9%) (Fig. 2.18). This leaves a cohort of 22 patients (64%) in which the underlying genetic aberrations are unknown. These cases will be followed up using the targeted DSD CGH array or WES, however this is outside the scope of this project. Five undiagnosed cases from the 46,XX DSD cohort were selected for a pilot study to test the efficacy of WGS for genetic diagnosis of these phenotypes.

Pathogenic variant SOX9 (5) NR5A1 (p.R92W) (3) 8 Diagnosis

Likely pathogenic variant 1 22 patients Following up NR5A1 (p.A260V) 3 Undiagnosed Variant of unknown significance WNT9A FOXL2 EMX2

Figure 2.18. Status of the 46,XX (ovo)testicular DSD cohort (August 2018).

The five cases selected for WGS were DSD01, DSD02, DSD04, DSD05 and DSD14; these patients were preferentially selected based on DNA quantity, parental DNA availability and phenotypic notes. These five individuals have comprehensive phenotypic notes (Table 2.2), making it easier to make phenotype-genotype correlations. Furthermore for two of these patients we have maternal and paternal DNA samples, for one we have maternal DNA and for the remaining two we have close contact with the clinicians so obtaining parental DNA in future is possible. Each of these patient DNA samples had undergone MLPA and targeted DSD gene panel analysis. No likely candidate genes were found in DSD02, DSD04, DSD05 and DSD14. DSD01 had a candidate variant followed up in EMX2 (see 2.3.5.2). Two of these patients also had microarray analysis that returned negative results. DSD02 has had a CombiSNP array and DSD04 went on the custom DSD CGH array. Together this made these five patients the best candidates for this trial study.

75

To date, WGS analysis has only been published for one DSD patient, a 46,XX DSD individual, where WGS was used to confirm a CNV identified by a CGH array (239). Previous studies have used WES on a cohort of nine 46,XX (ovo)testicular DSD cases (146), achieving a diagnostic rate of 33%. I hoped to surpass this, given the ability of WGS to identify SNVs, CNVs and structural rearrangements in both coding and non-coding regions of the genome.

2.3.6.1 Whole genome sequencing: CNV analysis WGS and bioinformatic analysis was performed at the Kinghorn Centre for Clinical Genomics Sequencing Laboratory (Garvan Institute). Ben Lundie (Garvan Institute) performed an analysis for large CNVs and structural rearrangements using ClinSV software (Minoche et al., manuscript in preparation) and Dr. Simon Sadedin (MCRI) analysed the data for CNVs using Schism software (240). This identified five CNVs and structural rearrangements of potential interest, summarised in Table 2.5.

In DSD04 I found a heterozygous 4.4 kb deletion encompassing exons 1-5 of Dynein Axonemal Assembly Factor 3 (DNAAF3) (Appendix 6). Homozygous mutations in DNAAF3 cause Primary ciliary dyskinesia-2, which is characterised by bronchiectasis and infertility (241). DNAAF3 encodes the SEF protein, which regulates FGF signalling (242); specifically the intracellular domain of SEF interacts with FGFR2 and FGFR1 (243). Dnaaf3 is highly expressed in the ovary and we know that FGF signalling is an important component of the testis pathway. This is an interesting candidate as it may downregulate FGF signalling in the ovary, therefore loss-of- function deletions could prevent repression of FGF signalling in the ovary.

Another CNV of potential interest was a 41 kb heterozygous duplication (chr1:22391601- 22432700) in DSD14. This duplication encompasses the Cell division cycle 42 (CDC42) gene and is located 12 kb downstream of WNT4. WNT4 has an essential role in human ovarian development, with homozygous and heterozygous mutations resulting in syndromic 46,XX partial/complete sex reversal and Müllerian aplasia (142, 156). There are four known XY human patients with duplications of chromosome 1p35 that includes the WNT4 locus, their symptoms range from isolated cryptorchidism to severe genital ambiguity (244). CDC42 has no known link

76 to reproductive processes and there are no reported CNVs associated with human reproductive phenotypes (ClinVar). In XX mice where Wnt4 was overexpressed, fertility and vasculature was noted to be normal and Cdc42 overexpression was associated with hyper-branching in the developing mammary gland but no gonadal phenotype (245).

This duplicated region was assessed on the UCSC genome browser with tracks displaying data for DNaseI hypersensitivity in fetal/adult ovary and fetal testis. Interestingly, there was a DNaseI hypersensitive peak in these three tissue samples, indicating a region of open chromatin and thus a potential regulatory region in gonads (Fig. 2.19A, B). Within this same region, there is also a positive profile for an active enhancer site, shown by ChromHMM data from penile foreskin fibroblast tissue. MATinspector was used to search for transcription factor binding sites within this 800 bp region of interest (Fig. 2.19B, C). I did not identify any known transcription factor regulators of sex determination, however relatively little is known about how the WNT4 locus is regulated during ovarian development. If this is a repressor element for WNT4, duplication at this locus may result in silencing of the WNT4 locus. Loss of WNT4 function (heterozygous and homozygous missense mutations) has been implicated in syndromic 46,XX (ovo)testicular DSDs previously (142, 156). Conversely, if this region contains an enhancer element it could lead to overexpression of WNT4. Wnt4 overexpression in mice has been shown to disrupt the NR5A1/β- catenin complex (246), an important regulator of anti-testis gene Nr0b1 and canonical WNT signalling. Future work could investigate this further by cloning this potential regulatory region into a reporter vector and testing for basal enhancer/repressor activity.

77 Table 2.5. Candidate variants from whole genome sequencing.

SNVs only

DS Gene Genomi cDNA Mutation Protein Zygo- db SNP ID CADD RVIS pLI/z gNo- Previous disease MGI mouse Inherita- D c change type change sity score- ExAC scores mAD associations reproductive nce ID location scaled (ExAC) freq. phenotype (whole exome)

DSD EMX2 chr10:g.1 ENST00000 Missense p.Ala93Val Het NA 19.94 38.66 z=3.02, Not Deletions Abnormal sex NA 01 1930305 442245.4:c. pLI=0.94 found encompassing determination, 6C>T 278C>T EMX2 identified in gonadal ridge 46,XY syndromic hypoplasia DSDs

FOXO3 chr6:g.10 ENST00000 Missense p.Asp62Asn Het rs532258926 19.65 45.95 None 0.0007 None Abnormal NA 8882595 406360.1:c. 637 oogenesis & G>A 184G>A folliculogenesis

DSD IL17RD chr3:g.57 ENST00000 Missense p.Lys131Thr Het rs184758350 22.8 35.79 z=-0.35, Not A SNP is associated None reported NA 04 144258T 427856.2:c. pLI=0 found with XX & XY >G 320A>C Hypogonadotropic hypogonadism, either hom or het with additional mutation in HH- associated gene, e.g. FGFR1.

DNAAF3 chr19:55 Deletion Het NA Primary ciliary None reported NA 673592- dyskinesia-2 (AR 5567790 inheritance) 3 resulting in bronchiectasis and infertility (241)

78 SCARA5 chr8:g.27 ENST00000 1 bp p.Val479Glyfs Het NA None NA z=0.7 , Not CNVs including Abnormal testis NA 729503A 354914.3:c. Insertion, Ter7 pLI=0.07 found SCARA5 – one XY morphology, >AC 1435dupG frameshift patient has reduced male and micropenis and fertility truncation of cryptorchidism in last 16 aa 46,XY (but CNV also includes STAR)

DSD Includes 8:106796 771 bp Het NA Deletion of RP11- Zfpm2-/- mice NA 05 ZFPM2 533- deletion 152P17.2(ZFPM2) have male to 1067973 associated with female sex 05 46,XY sex reversal reversal (92).

USP25 chr21:g.1 ENST00000 Missense p.Pro781Arg† Het rs190641000 2.65 -0.85 z=0.7, 0.0002 None reported None reported Paternal 7219986 400183.2:c. - In pLI=0.99 408 C>G 2342C>G NM_0012830 41.2 isoform only

chr21:g.1 ENST00000 Missense p.Ala940Thr Het rs34979861 6.03 NA 0.0014 Paternal 7250133 400183.2:c. 13 G>A 3028G>A

Includes 10:50192 Balanced Het NA Associated with None reported NA AKR1C2 28- translocation 46,XY females (AR 5033008 inheritance).

NOBOX chr7:g.14 ENST00000 Missense p.Pro609Leu Het rs115882574 None NA z=-1.80, 0.0009 Associated with POI Abnormal ovarian Maternal 4094583 223140.5:c. (mater pLI=0 689 (AD inheritance) follicles, G>A 1475C>T nal) (177) and 46,XX sex abnormal seminal reversal vesicle, (McElreavey group) absent oocytes

GALT chr9:g.34 ENST00000 Missense p.Phe86Ser Het rs111033715 26.8 23.53 z=2.09, 0.0000 Galactosemia (AR Abnormal Paternal 648116T 556278.1:c. pLI=0 1804 inherited), includes spermatogenesis >C 257T>C ovary abnormality

79 CITED2 chr6:g.13 ENST00000 Missense p.Pro202Thr Het rs146180399 20.9 20.49 z=0.8, 0.0001 Associated with Absent adrenal Maternal 9694478 537332.1:c. pLI=0.71 733 heart septal defects G>T 604C>A (AD inheritance)

CHD7 chr8:g.61 ENST00000 Splice region NA Het rs141314884 None NA z=1.92, 0.0025 Hypogonadotropic Abnormal female Maternal 763323A 527921.1:n variant pLI=1 82 Hypogonadism, genitalia, >G .25+3A>G ovarian hypoplasia reduced fertility

DSD CDC42 1:223916 41kb Het NA 46,XY patients with Fertility & NA 14 and 12 01- duplication duplications vasculature kb 2243270 including WNT4 sho normal in Wnt4 downstre 0 w cryptorchidism to overexpression am of severe genital XX gonad (246) WNT4 ambiguity (244) Cdc42 overexpression disturbs mammary gland, gonadal phenotype unknown (245)

Breakpoi 10:91168 Translocation Het NA None None reported in NA nt on 12 253 to Sox5-null is located 12:23700 within 877 SOX5

GALT chr9:g.34 ENST00000 Splice region NA Het rs61735984 None None z=2.09, 0.0031 Galactosemia Abnormal Paternal 648114C 556278.1:c. variant pLI=0 38 spermatogenesis >A 255C>A

CHD7 chr8:g.61 ENST00000 Splice region NA Het NA 10.08 1.43 z=1.92, 0.0001 Hypogonadotropic Abnormal female Unknown 653820C 526846.1:c. variant pLI=1 614 Hypogonadism, genitalia, – no >G -172C>G ovarian hypoplasia reduced fertility coverage on WES

80 DSD LMNA chr1:g.15 ENST00000 Missense p.Met290Val Het NA None NA z=0.5, Not Malouf Syndrome, Small gonad, Maternal 02 6105866 392353.3:c. pLI=0.99 found Atypical Werner azoospermia A>G 868A>G (includes polycystic ovaries & testis abnormalities

USH2A chr1:g.21 ENST00000 Splice NA Het NA 18.76 99.79 z=-5.12, Not Retinitis None reported Maternal 5972462 366943.2:c. acceptor pLI=0 found Pigmentosa (w C>T 9746-1G>A variant testis abnormality) Bardet-Biedl Syndrome (w ovary hypoplasia)

81

82 Figure 2.19. A heterozygous duplication at chr1: 22391601-22432700 includes a potential enhancer/repressor downstream of WNT4. (A) The 41 kb genomic region duplicated in DSD14, located 12 kb downstream of WNT4. Bioinformatic tracks are shown including the ENCODE track of enhancers present in human mammary epithelial cells (HMEC) (yellow denotes active enhancer) and DNaseI hypersensitivity data from human fetal testis and ovary (ROADMap). This shows a peak at the 3’ end of the duplication (red asterisks, far top right of figure), enlarged in (B). (B) DNaseI hypersensitivity peak in fetal/adult ovary and fetal testis, the central region of this peak is highly conserved among mammalian species (100 vertebrate conservation track). (C) Binding sites of transcription factors found within the 800 bp potential regulatory region (chr1: 22,431,801-22,432,600). The factors that bind to the labelled sites are as follows: E2F; MASH1, Achaete-Scute Family BHLH Transcription Factor 1; SREBP, Sterol Regulatory Element Binding Protein; TCF11, Transcription Factor 11.

83 2.3.6.2 Whole genome sequencing: Single nucleotide variant analysis The WGS data was then assessed for SNVs and small INDELs using Seave (210) (Methods 2.2.8). I performed several analyses, focusing on candidate genes, highly damaging variants and inheritance mode (Analysis 1-3 respectively) (Fig. 2.20). Variants from each of these analyses underwent the same filtering on Seave in terms of prevalence (<0.5%), quality (≥100) and predicted pathogenicity (CADD Scaled score ≥ 2). Candidate variant lists from a single analysis on a single patient included up to 5000 variants, I therefore performed additional filtering on Excel, removing those which have been identified twice or more in the Kinghorn genome/exome database (MGRB) and variants not located in coding regions of DNA (except intronic splicing regions). This does mean we may exclude pathogenic variants in non-coding regulatory regions, however given how little is known about regulatory regions of the genome as a whole, this simplified our initial analysis. After applying this filtering, around 100-300 variants remained for each patient. To select the best possible candidate variants I used a combination of variant/gene scoring systems: Combined Annotation-Dependent Depletion (CADD) score (247), Residual Variation Intolerance Score (RVIS) (248), pLI/z constraint scores (ExAC) (232), gnomAD frequency and VarElect (http://varelect.genecards.org). I also curated genes with data from the MGI mouse database, whether they show expression in reproductive tissues and whether there is a reproductive phenotype in mutant mice, and human disease association from Online Mendelian Inheritance in Man (OMIM). Top candidates are summarised in Table 2.5.

84

Figure 2.20. Whole genome sequencing analysis pipeline. Overview of the three approaches to variant filtering in 46,XX DSD whole genome sequencing data.

In the first analysis I used three gene lists to filter the data (Appendix 3). The first list includes diagnostic genes for all DSD phenotypes (n = 64, annotated by Reproductive Development group). The second contains all candidate genes for 46,XX (ovo)testicular DSDs based on mouse models and relevant signalling pathways (n = 75, annotated by myself). The third list contains candidate genes for POI (n = 94, annotated by Dr. Elena Tucker), as there is overlap in underlying molecular mechanisms for these phenotypes and 46,XX disorders of ovarian development. The first analysis generated the lowest number of variants overall; one of these was the EMX2 variant in DSD01, confirming its predicted pathogenicity (CADD = 19.94, RVIS = 38.66) as shown previously from the targeted DSD panel analysis (see Results 2.3.5.2). A further variant of interest was identified in the NOBOX gene in DSD05. This was a maternally inherited heterozygous missense variant which showed a low frequency on gnomAD (0.0009689). This gene encodes a transcription factor that plays an essential role in the first steps of ovarian

85 folliculogenesis and in the maintenance of the follicular pool, heterozygous loss-of-function variants underlie up to 6% of POI cases (177). This variant was flagged because a heterozygous variant has been recently found in a 46,XX (ovo)testicular DSD patient in a cohort from Prof. Ken McElreavey’s lab (Pasteur Institute, Paris).

The following analyses followed a more unbiased approach, in the second analysis I filtered variants for loss-of-function and high impact variants, e.g. frameshift and truncating variants. Using this approach I obtained between 8 and 74 variants per patient. These were prioritised using VarElect, which ranked variants on features including mouse phenotype, expression in relevant tissue (mouse) and pLI scores. From this there were several candidates of interest. In DSD04 I identified a 1 bp heterozygous insertion in the Scavenger receptor class A, member 5 (SCARA5) gene. This results in a frameshift and truncation of the last 16 amino acids, impacting the SRCR (C-terminal scavenger receptor cysteine-rich domain) domain of the protein. RT-PCR and histological staining have shown high levels of Scara5 expression in the Sertoli and epithelial cells of the testis (249). Furthermore, Scara5-/- mice have been described as having abnormal testis morphology and reduced male fertility.

The third analysis focused on double-hits in genes, looking at variants inherited as homozygous or as potential compound heterozygous scenarios. This analysis was hampered by the fact that compound heterozygous status could not be elucidated given these patients were sequenced as singletons. However following curation I found two variants of interest in the Ubiquitin specific protease 25 (USP25) gene in DSD05. This gene encodes a deubiquitinating enzyme and was flagged given its role as a positive regulator of WNT/β-catenin signalling. Deficiency of this enzyme may result in the degradation of tankyrases, leading to stabilisation of Axin that would antagonise WNT signalling (250). Each of these variants was missense. While they were not predicted to be highly pathogenic (CADD = 2.65 and 6.03), this gene does show a degree of intolerance to missense and LOF variation (z = 0.7, pLI = 0.99). However it should also be noted that this is an exceptionally large gene, spanning over 150 kb at 21q11.2. There is no mouse reproductive phenotype or reports of human disease association for this gene.

Given the difficulty in narrowing down candidate genes from WGS analysis, we performed follow-up trio/duo WES in three patients where parental DNA was available (DSD02, DSD05 and

86 DSD14) (see Methods 2.2.11). This allowed us to establish inheritance mode in the candidate SNV list (shown in Table 2.5). This is a valuable piece of information to help assess pathogenicity of variants. A number of variants were maternally inherited, including NOBOX (DSD05), LMNA and USH2A (DSD02) (Table 2.5). Given there is no known DSD phenotypes reported in the mothers of these patients, these genes could only be involved in the phenotype in cases of incomplete penetrance/variable expressivity or as genomic modifiers. In the case of NOBOX, we know that variants in this gene are associated with POI in an autosomal dominant inheritance model (177). Incomplete penetrance and variable expressivity have not been shown in association with these reported POI cases, therefore this variant is likely benign.

Another example where predicted pathogenicity altered was the USP25 gene in DSD05 (Table 2.5). Each variant showed paternal inheritance; indicating that only one allele of the USP25 gene is affected rather than a compound heterozygous scenario. This means that USP25 would have to be haploinsufficient for one or both of these variants to have a pathogenic effect. Indeed, given that this protein has a potential role in the ovarian pathway, via WNT signalling, it is possible that reduced function in the 46,XY father would not affect reproductive function. Further evidence for the role of this gene in ovarian development must be sought to assess whether this it is a good candidate.

As shown in these examples, the addition of parental WES data assists our analysis of the WGS data greatly; highlighting which variants we ultimately want to follow up with in vitro/vivo analyses.

2.4 Discussion Our understanding of the factors controlling human gonad development and their role in DSD has advanced significantly since the discovery of the first DSD gene 28 years ago. Advances in microarray and MPS technology have contributed largely to this, where the diagnostic rate in 46,XY DSD cohorts can now be as high as 60% (2). However, we consistently see lower rates of diagnosis in 46,XX DSD cohorts, likely due to gaps in our knowledge on ovarian development, and the contribution of large CNVs which may not be picked up by targeted sequencing methods. In this chapter I have demonstrated a diagnostic pipeline for a cohort of 34 cases with

87 46,XX disorders of ovarian development, with the aim being to identify the underlying molecular genetic cause of these DSD phenotypes.

SRY-negative 46,XX (ovo)testicular DSDs, making up 70% of this cohort, are caused by both CNVs and SNVs. To capture this underlying variation, I used two targeted gene approaches, MLPA and the targeted DSD gene panel. Candidate genes from each of these studies were followed up using in vitro cellular assays and by generation of mutant mouse models. This gave us insight into how these genes function in their wild-type and variant form, as well as confirmed pathogenicity in several cases. In the final study I took a wider-scale untargeted approach to genetic analysis, trialling WGS on five 46,XX DSD patients.

A definitive diagnosis was reached in five patients following MLPA (and for some cases CGH- array) analysis. This added to the expanding pool of DSDs caused by SOX9 enhancer CNVs, which is defining the precise coordinates of this regulatory region and giving insight into the regulation of this key testis gene (154, Croft et al., (2018), Nature Communications 9:5319). In the context of our cohort, duplications of the SOX9 enhancer region accounted for 15% (5/34) of the total cases, but more specifically 15% of 46,XX T-DSD (2/13 cases) and 30% of 46,XX OT-DSD (3/10 cases). For 46,XX (ovo)testicular DSDs, MLPA has a relatively high diagnostic yield, it is also quick and cost-effective, demonstrating it’s utility as a diagnostic screening tool for these phenotypes. However, in a research setting it is less effective, the main downsides being the small amount of data retrieved and the high quantity of DNA required (100 ng per reaction). For us to garner useful information on SOX9 regulation from the duplications we identified (i.e. fine-mapping of genomic region), we needed to perform a follow up CGH array. This demonstrates the transitional stage we are in with old and new technology, in future it is likely that we would obtain all necessary information from WGS.

In the application of the targeted DSD gene panel to 46,XX DSDs I showed a diagnostic rate of 8.8% (3/34), with a further ‘likely diagnosis’ in one case. These were all in the NR5A1 gene. Taking the diagnostic CNVs and SNVs together, I was able to reach an overall diagnostic rate of 26% (9/34 patients) in this 46,XX DSD cohort. Our recent study (2) presented the outcomes of the targeted DSD gene panel across our entire DSD cohort, highlighting the differential in diagnostic rate between 46,XX and 46,XY DSD (18% versus 43% respectively). This differential is

88 likely attributed to the fact that a large proportion is caused by CNVs (up to 30% in SRY-negative 46,XX OT-DSD), which cannot be picked up using this approach. Furthermore, over 20 genes have been implicated in 46,XY sex reversal or gonadal dysgenesis, while only 11 genes have been implicated in 46,XX (ovo)testicular DSD cases, just four of these in SNV form (NR5A1, RSPO1, WNT4 and WT1). This gap likely stems from these being much rarer and less-studied phenotypes as well as us knowing less about the ovarian differentiation pathway compared to testis. Furthermore, given the wide phenotypic variation in 46,XX DSDs (e.g. ovotestes and gonadal fate asymmetry) it is likely that mosaicism contributes to a portion of undiagnosed cases. Proving mosaicism would require sampling gonadal tissue paired with high coverage or long- read sequencing approaches. Such investigation may reveal more diagnoses but is outside the scope of our study.

Nonetheless, this analysis resulted in the identification of three candidate genes (EMX2, FOXL2 and WNT9A). In DSD01, who presented with 46,XX T-DSD, I found a heterozygous missense variant in EMX2. This was located in the highly conserved Homeobox protein chain region and not previously reported in human population databases. A CRISPR mutant mouse model for this variant (Emx2-p.A94V+/+) has shown that ovaries develop normally but the male steroidogenic pathway may be activated in a subset of cells. Additional characterisation of these mutant mouse gonads will aim to identify whether the EMX2-p.A93V variant can cause XX sex reversal. Follow up whole genome sequencing of DSD01 did not reveal any strong candidates, although a missense variant in found in FOXO3 (Table 2.5) may act in combination with EMX2 in this phenotype.

In DSD34 the targeted DSD gene panel identified a paternally inherited missense variant in ovarian gene FOXL2. This variant is located in the highly conserved DNA binding domain, a mutational hotspot for BPES syndrome (affecting the ovary and eyelid). Protein modelling predicted this p.Ala71Pro change to affect FOXL2 binding to target genes. In vitro cellular assays will be used to confirm pathogenicity of this variant.

Finally, I identified two missense variants on the same allele of the WNT9A gene in DSD10, an individual with virilisation and possible 46,XX T-DSD. Wnt9a shows strong evidence of being a regulator of ovarian development, based on mouse expression and mutant models, as well as its role in the WNT/β-catenin pathway. Each of these variants in DSD10 introduced amino acids of different properties in terms of size, charge and hydrophobicity. Given the location of these variants, it is predicted that they would disturb interaction of WNT9A with other molecules. The

89 results of our Wnt9a-/- KO mouse model will help determine whether this gene may play a role in DSD phenotypes.

The remaining cases where strong candidates were not identified will be analysed for CNVs and structural rearrangements using the targeted DSD CGH array, which is out of scope for my thesis.

I also applied WGS to a subset of these undiagnosed cases. This pilot study began with great hope for an increased diagnostic yield, given the ability of WGS to simultaneously identify SNVs, CNVs and structural rearrangements across the entire genome. Several interesting candidates came out of CNV/rearrangement analysis, including a duplication downstream of WNT4. At the 3’ end of this duplication I identified a region where DNaseI hypersensitive sites were highly enriched in human fetal gonad tissues, suggestive of a potential enhancer/repressor region. Luciferase assays could be used to test whether there is WNT4-related enhancer activity within this region.

Analysis for SNVs identified a large number of candidate variants (>200) for each patient. I found several variants that are predicted to be pathogenic and appear to function in pathways important for ovarian differentiation, e.g. USP25, GPR89B and NOBOX. Additional information on inheritance mode following parental WES has aided prioritisation of these variants.

These three approaches to genetic analysis each highlight the challenges involved in identifying molecular diagnoses in 46,XX DSDs. The key limitation of this cohort is the predominance of singleton cases and low sample size due to their rarity. In cases where promising candidate genes have been identified (e.g. WNT9A) we only have one patient affected by this gene, making it is difficult to argue pathogenicity as well as publish. Looking forward, we expect that ongoing analysis of the patient datasets generated here will ultimately resolve diagnoses as we learn more about human gonad development as well as gene variant interpretation. Top candidate variants identified from the targeted DSD gene panel and WGS will be shared with collaborating DSD researchers and uploaded to SDgene and MatchMaker exchange (https://www.matchmakerexchange.org), to hopefully identify genes involved in similar DSD phenotypes.

90 While diagnoses are still undetermined in the vast majority of cases described in this chapter, what I have generated is a valuable resource for future research. In these 22 undiagnosed patients I have eliminated the known causes of 46,XX DSD, moving us closer to diagnoses and leaving us with datasets that hold great potential for novel gene discovery.

91

92 Chapter 3: Characterisation of variants in the NR5A1 gene: implications for 46,XX Disorders of Sex Development

3.1 Introduction NR5A1, encoding the NR5A1/SF1 protein, is a transcription factor and key regulator of the hypothalamus–pituitary–gonadal and hypothalamus–pituitary–adrenal axes during development (97, 251, 252). Given its important role in gonadal development (253, 254), variants in NR5A1 are implicated in a wide range of DSD phenotypes, including 46,XY DSD (gonadal dysgenesis, hypospadias, under virilisation and male infertility (161, 162))) and 46,XX POI (163). SNVs had not previously been diagnostic for non-syndromic 46,XX (ovo)testicular DSDs until 2016 when three reports simultaneously identified the same NR5A1 variant (p.Arg92Trp) in ten individuals with 46,XX (ovo)testicular DSD (146-148). This variant has now been reported in twelve cases of 46,XX (ovo)testicular DSD (146-148, 255, 256).

The findings described in Chapter 2 showed that four 46,XX (ovo)testicular DSD patients (DSD07, DSD17, DSD21 and DSD28) carried pathogenic (p.Arg92Trp) or likely pathogenic (p.Ala260Val) heterozygous variants in the NR5A1 gene. The aim of this chapter is to characterise these novel and known NR5A1 variants to elucidate the mechanism by which variant NR5A1 mediates sex reversal in the patients described. A further aim was to detect potential causes of the wide phenotypic variability and incomplete penetrance observed in individuals with the NR5A1 p.Arg92Trp variant. We assessed intracellular localisation of wild-type and variant NR5A1 protein to see whether trafficking was affected. Protein modelling was performed to demonstrate how each of these variants impact conformation and interaction with other molecules. To determine which sex differentiation pathways are affected in these patients we used luciferase assays. Finally, the MPS data was analysed for potential modifiers that may contribute to the variable expressivity observed.

93 3.2 Methods

3.2.1 Massively Parallel Sequencing Sequencing data was generated and analysed as outlined in Methods 2.2.4. NR5A1 variants were inspected for quality using IGV (http://www.broadinstitute.org/igv/). Sequencing data was further analysed for genomic modifiers by filtering variants using a list of 116 genes alongside previously described filtering criteria. An initial gene list of potential NR5A1 interactors (N = 576) was compiled from NR5A1 overexpression/knockdown assays (257) and data in STRING (https://string-db.org/). 116 of these genes were covered by our targeted DSD gene panel.

3.2.2 Plasmid construction Variant NR5A1 expression vectors carrying the c.C274T or c.C779T variants were generated in the mammalian expression vector pCMV6-Entry-hNR5A1 (OriGene, #RC207577) using the QuikChange II XL Site-directed Mutagenesis Kit (Agilent, #200521) according to the manufacturer’s instructions (see primer sequences in Appendix 7).

The pGL4-hNR0B1 reporter plasmid (Fig. 3.1) was generated via cloning 994 bp of the promoter region of the human NR0B1 gene (chrX:30327432-30328425) into the pGL4.10.luc2 reporter plasmid (Promega, #E665A) (see primer sequences Appendix 7). Colonies were pre-screened for the presence of insert using a restriction enzyme digest for XhoI and HindIII (cut sites located either side of hNR0B1 insert, Fig. 3.1). Sanger sequencing (see Methods 2.2.9) was then performed on positive colonies to check for correct orientation, sequences were aligned using Benchling (https://benchling.com).

94

Figure 3.1. Generation of a reporter vector for the human NR0B1 promoter. (A) The 994 bp human genomic region cloned, containing the NR0B1 promoter. (B) Map of the pGL4.10.luc2 plasmid including the hNR0B1 promoter region. (C) Sanger sequencing confirms cloning of the human NR0B1 promoter region into the pGL4.10.luc2 plasmid in the correct orientation in three colonies (C3, 4, 5). Alignment generated using Benchling software.

3.2.3 Protein immunofluorescence COS-7 cells were seeded onto eight-well chamber slides (Lab-Tek) and transfected with NR5A1 expression vectors (WT, p.Arg92Trp, p.Ala260Val) using Lipofectamine 2000 (Invitrogen,

95 #11668019). Following 24 hours of transfection cells were washed with PBS, fixed with 4% PFA, permeabilised with 1% Triton-X and then blocked with 2% BSA in PBS. Cells were incubated overnight with primary antibodies (polyclonal SF1 (E18) antibody (1:300; Santa Cruz, #SC10976), actin (1:200; Sigma, #5060)) in 1% BSA. Cells were washed three times with PBS and incubated in secondary antibodies (Alexa 488 and Alexa 594 (both 1:1000, Invitrogen)) in 1% BSA/PBS. DAPI was used for nuclear counterstaining. Cells were imaged using the LSM 780 confocal microscope (Zeiss) at 40X magnification.

3.2.4 In silico protein modelling A HOPE analysis was performed to assess the structural and functional consequences of the variants identified (https://www.cmbi.ru.nl/hope/)(211). Predictions of the NR5A1 variant protein structure were generated in silico using I-TASSER, (https://zhanglab. ccmb.med.umich.edu/I-TASSER/) (258, 259). Predicted crystal structures were then visualised using PyMOL Molecular Graphics System (v1.7.6.6) (https://www.pymol.org).

3.2.5 Luciferase assays Luciferase reporter assays were performed in two cell lines: HEK 293-T and COS-7. Luciferase activity was assessed in 96-well plates using Lipofectamine 2000, with co-transfection of Renilla (pRL-TK) as a marker of transfection efficiency. To assay the ability of NR5A1 to transactivate the SOX9 gene (via its enhancer TESCO), COS-7 cells were co-transfected with a reporter construct, pGL4-mTESCO (75 ng/well), pRL-TK (10 ng/well), SOX9 (50 ng/well) and NR5A1 (WT, p.Arg92Trp, p.Ala260Val) (50 ng/well) and harvested 24 hr post transfection. The TESCO reporter was also used to assay the NR0B1-mediated repression of SOX9, as previously described (148). We further investigated the ability of NR5A1 proteins to interact with NR0B1 using a pGL4-hDAX1 reporter, this was co-transfected into HEK 293-T cells at 20 ng/well with pRL-TK (10 ng/well), pcDNA3-β-catenin (30 ng/well) and NR5A1 (WT, p.Arg92Trp, p.Ala260Val) (10 ng/well), cells were harvested 36 hours post transfection. To assay the ability of NR5A1 proteins to regulate the canonical WNT pathway, we used the TOPFlash-TCF reporter, pTOPflash (50 ng/well), and co-transfected HEK 293-T cells with pRL-TK (5 ng/well), pcDNA3-β-catenin (50 ng/well) and NR5A1 (WT, p.Arg92Trp, p.Ala260Val) (50 ng/well), cells were harvested 48 hours post transfection. Luciferase activity was measured using Dual-Luciferase Reporter Assay System Kit

96 (Promega, #E1980) on an Infinite M200 Pro plate reader (Tecan). Data represent the mean with standard error of three independent experiments, each performed in triplicate.

3.3 Results

3.3.1 Identification of NR5A1 variants via Massively Parallel Sequencing During analysis of the targeted DSD gene panel two pathogenic/likely pathogenic variants (c.C274T:p.Arg92Trp and c.C779T:p.Ala260Val) were identified in the NR5A1 gene across four patients (DSD07, DSD17, DSD21 and DSD28) (Table 3.1). These were both located in exon 4 of NR5A1 and found in heterozygous form. These variants were flagged for functional analysis given their recent implication in 46,XX (ovo)testicular DSD.

DSD07, DSD17 and DSD21 are singletons therefore inheritance mode is unknown; in DSD28 the NR5A1 p.Arg92Trp variant was maternally inherited. Previous reports indicate that this variant has an autosomal dominant mode of inheritance, with variable expressivity and incomplete penetrance; this was also demonstrated here with the mother of DSD28 presenting with sub- fertility. Interestingly, the maternal aunt and uncle of DSD28 were also reported to have fertility issues.

Table 3.1. NR5A1 variant information.

DNA Protein In gnomAD Patient ID Location Zygosity Inheritance Previous reports change change silico freq.

NA DSD07 DSD07, & DSD17. 46,XX & XY DSD DSD17, c.C274T Exon 4 p.Arg92Trp Het 4/4 Not found Maternal (146-148) DSD28 DSD28

DSD21 c.C779T Exon 4 p.Ala260Val Het NA 46,XY DSD (260) 2/4 4.13e-6

In silico: PolyPhen2, MutationTaster, SIFT, LRT; fraction indicates proportion predicted deleterious or possibly deleterious. NA, not available.

97 Each of the individuals with the NR5A1 variants presented with SRY-negative 46,XX (ovo)testicular DSD (Table 2.2), with varying degrees of virilisation. DSD07 was born to non- consanguineous parents and presented in adulthood with azoospermia and bilateral testes. DSD17 presented at birth with ambiguous genitalia, bilateral ovotestes, 2 cm phallus with perineal urethral opening and no Müllerian structures. DSD28 presented with a micropenis, hypospadias and small underdeveloped scrotum, with no evidence of Müllerian structures and a positive testosterone response to stimulation. This patient also had a seizure disorder being managed by neurologists. DSD21 presented with ambiguous genitalia, a small phallus and vagina, with separate urethral and vaginal openings. The right gonad was an ovotestis, while the left gonad was ovarian. DSD07, DSD17 and DSD29 were shown to have normal adrenal function at the time of sample collection; this data was not available for DSD21.

IGV was used to confirm the presence of these variants (Fig. 3.2 and 3.3), this showed that these variants were heterozygous and of good quality, showing both good depth and bidirectional reads spanning the regions. Sanger sequencing to confirm these variants was therefore unnecessary.

98

Figure 3.2. IGV visualisation of the NM_004959.4(NR5A1):c.C274T;p.(Arg92Trp) variant in DSD07, DSD17 and DSD28. For DSD17 total allele count = 155, A allele count = 73, G allele count = 82, N (unknown) allele count = 0. For DSD07 total allele count = 226, A allele count = 108, G allele count = 117, N (unknown) allele count = 0. For DSD28 total allele count = 101, A allele count = 66, G allele count = 35, N (unknown) allele count = 0.

99

Figure 3.3. IGV visualisation of the NM_004959.4(NR5A1):c.C779T;p.(Ala260Val) variant in DSD21. Total allele count = 145, A allele count = 86, G allele count = 59, N (unknown) allele count = 0.

These variants are located in the DNA binding and ligand binding domains (Fig. 3.4). The p.Arg92Trp variant shows high sequence conservation across vertebrate species, indicating that this is an important region of the protein. The p.Ala260Val variant shows lower sequence conservation, conserved only across mammalian species, however this is unsurprising given that sex determination is divergent between human and chicken.

100

Figure 3.4. Location of variants in NR5A1 protein and sequence conservation. The NR5A1 protein consists of the DNA binding domain (DBD), hinge domain (HD) and ligand binding domain (LBD). NR5A1 gene variants are found at residues highly conserved among mammalian species.

3.3.2 NR5A1 protein localisation Expression vectors were generated for variant NR5A1 so that the function and behaviour could be assessed in vitro. We first investigated whether protein localisation or expression was affected by NR5A1 variants. Wild-type NR5A1 is known to show nuclear expression, with characteristic nucleolar exclusions (261). This expression pattern was observed in immunofluorescence staining of COS-7 cells transfected with wild-type and variant NR5A1 (Fig. 3.5). There was also no difference in levels of expression seen between the wild-type and variant NR5A1 (Fig. 3.5iv, vi).

101

Figure 3.5. Immunofluorescence staining for NR5A1 protein in COS-7 cells. Cells were transfected with equal amount of NR5A1 expression vector (wild-type or variant), protein expression was assessed using the NR5A1 (E-18) antibody (green), DAPI (blue) was used for nuclear counter-staining. Wild-type NR5A1 showed strong nuclear staining with nucleolar exclusions (i, ii), expression and localisation of missense variant NR5A1 (p.Arg92Trp and p.Ala260Val) was unaffected (iii-vi). A truncating NR5A1 variant (p.R89Gfs*17), shown for comparison (viii), was not expressed. This analysis was performed by Gorjana Robevska.

3.3.3 Protein modelling Next we investigated the effects of NR5A1 variants on protein conformation, a HOPE analysis was performed to generate the crystal structure of variant NR5A1 and the resulting protein was visualised using PyMol. For the p.Arg92Trp variant this falls at a residue within the highly conserved DNA binding domain. Substitution of an Arginine to a Tryptophan induces a significant change at this functionally critical position; there is a change in charge (positive to neutral) and loss of hydrogen bond forming potential due to the size difference between these residues (Fig. 3.6A, B). Because of the large change and the importance of this region for binding to target genes, it is likely that this change affects the variant proteins ability to interact with DNA. In the p.Ala260Val variant, the wild-type Alanine is part of the evolutionarily conserved alpha helix 3 of the ligand binding domain. The Alanine residue sits at the surface of the protein and in close proximity to a ligand binding site (residue 262, iTASSER), therefore its replacement with a bulkier Valine could affect interactions of NR5A1 with other molecules (Fig. 3.6C, D). Together these indicate that these variants may impact the ability of NR5A1 to interact with its cofactors and regulate target genes. I next investigated variant NR5A1 function in vitro to see what stage of sex differentiation these variants may be affecting.

102

Figure 3.6. Protein structure modelling of NR5A1 variants. In silico predictions of protein conformation of wild-type and variant NR5A1. (A) Wild-type Arginine (Arg, R) at position 92 falls within the DNA binding domain of the protein. (B) The variant Tryptophan (Trp, W) is larger than the Arginine and has less Hydrogen bonding potential. (C) The residue at position 260 falls within alpha helix 3 of the ligand binding domain. (D) The wild-type Alanine (Ala, A) is smaller than the variant Valine (val, V); this is located on the protein surface. In silico predictions were generated using iTASSER and PyMol modelling. This analysis was performed by Gorjana Robevska.

3.3.4 Luciferase assays Using the variant NR5A1 expression vectors I tested several sex differentiation pathways for changes in response to the p.Arg92Trp and p.Ala260Val forms of NR5A1. A previous report showed that the NR5A1 p.Arg92Trp variant has less ability to upregulate male pathway genes (Sox9, Amh and Cyp11a1) (147). We assessed if NR5A1 variants can still upregulate the SOX9 mTESCO enhancer. This showed that like the NR5A1 p.Arg92Trp variant, the p.Ala260Val variant also shows decreased ability to transactivate this male pathway enhancer (Fig. 3.7A). The

103 reduced SOX9 activation was also evident in the presence of SOX9-inhibiting factor FOXL2 (Fig. 3.7A), suggesting that these NR5A1 variants do not affect FOXL2-mediated repression of SOX9.

As the testis-specific SOX9 does not appear to be abnormally activated by these NR5A1 variants, other assays focused on ways that SOX9 may escape repression during ovarian development. NR0B1 is a key repressor of the testicular pathway and its dysregulation may underlie the XX sex reversal. During ovarian development NR0B1 represses NR5A1 transactivation of SOX9 (91). Igarashi et al. (2016) proposed that the p.Arg92Trp variant NR5A1 is less responsive to NR0B1. To assess this, HEK 293-T cells were co-transfected with constructs for NR5A1, NR0B1, SOX9 and the mTESCO Sox9 enhancer. When wild-type NR5A1 was present, transfection with NR0B1 resulted in dosage-dependent repression of Sox9 mTESCO (Fig. 3.7B). Likewise, when each variant NR5A1 was present Sox9 mTESCO remained repressed in a dosage-dependent response to NR0B1. This shows that these variants are still responsive to NR0B1-mediated repression of SOX9. Bashamboo et al. (2016) also investigated NR0B1 dysregulation. In the differentiating gonad, the NR5A1/β-catenin complex upregulates several targets, one of the more well characterised being DAX1 (encoded by NR0B1) (262). Bashamboo et al. (2016) suggested that the NR5A1 p.Arg92Trp variant has less synergy with β-catenin, meaning NR0B1 is dysregulated and SOX9 repression is lost in the XX gonads (147). To investigate the direct effects of variant NR5A1 on NR0B1 activity, I generated a reporter construct containing 994 bp of the upstream promoter region of NR0B1 (Methods 3.2.2, Fig. 3.1). HEK 293-T cells were co-transfected with the NR0B1 reporter along with constructs for wild-type and variant NR5A1, in the presence and absence of β-catenin (Fig. 3.7C). NR0B1 promoter activity was upregulated by wild-type NR5A1 and each variant showed a similar level of upregulation, consistent with what I had seen in the previous assay (Fig. 3.7B) where variant NR5A1 did not change NR0B1 activity on SOX9. To investigate the NR5A1/β-catenin complex in NR0B1 regulation, I introduced β-catenin. Wild- type NR5A1 and β-catenin caused a 5-fold increase in NR0B1 reporter activity compared to empty control, yet introduction of variant NR5A1 significantly repressed this activity (Fig. 3.7D). To further investigate whether NR5A1 variants repress the WNT/β-catenin pathway, I measured the effect of wild-type and variant NR5A1 on canonical WNT activity using the TOPFlash reporter system (Fig. 3.7E). In the presence of β-catenin, the TOPFlash reporter showed a 20-fold induction (Fig. 3.7E). WNT signalling was repressed when wild-type NR5A1 and β-catenin were transfected and this was repressed further in the presence of the p.Arg92Trp and p.Ala260Val NR5A1 (p=0.0002 and p<0.0001 respectively). Since each of these variants were heterozygous

104 in the DSD patients described, I also transfected each variant NR5A1 with an equal amount of wild-type construct. Repression of WNT signalling was still observed at similar levels when each variant NR5A1 was transfected with and without the wild-type, indicating that there is no dominant negative effect. Taking these results together, it appears that these variants in NR5A1 disrupt the NR5A1/β-catenin complex, impacting its normal ovarian function to activate NR0B1 and the WNT signalling pathway (Fig. 3.8). These disrupted regulatory relationships during ovarian development may allow SOX9 to escape repression, leading to testis differentiation in an XX individual.

105

106 Figure 3.7. NR5A1 variants show altered function in sex differentiation pathways in vitro. (A) Transactivation of SOX9 mTESCO is reduced when each variant NR5A1 (as well as a positive control variant, LoF from 46,XY DSD) is transfected into COS-7 cells with SOX9, compared to wild-type NR5A1. This was also observed in the presence of female pathway repressor FOXL2. (B) Co-transfection of HEK 293-T cells with NR5A1, SOX9 and increasing amounts of NR0B1 showed that variant NR5A1 does not impact NR0B1-mediated repression of SOX9. SOX9 activity measured using the mTESCO reporter. (C) Co-transfection of COS-7 cells with wild-type or variant NR5A1 results in no change in NR0B1 promoter activity. (D) Co-transfection of COS-7 cells with wild-type or variant NR5A1 and β-catenin causes repression of the NR0B1 promoter for both NR5A1 variants. (E) TOPFlash activity is reduced when HEK 293-T cells are co- transfected with β-catenin and variant NR5A1 compared to wild-type NR5A1. Data represent the mean with the standard error of three independent experiments performed in triplicate. An unpaired t-test was applied to obtain p-values, **** p < 0.0001; *** p < 0.001; ** p < 0.01; * p < 0.05; ns = p > 0.05.

107

Figure 3.8. Variants in NR5A1 disrupt the ovary-promoting effects of the NR5A1/β-catenin complex. (A) In the developing ovary, NR5A1 and β-catenin form a complex that upregulates NR0B1 activity, leading to repression of testis-specific SOX9. The NR5A1/β-catenin complex also regulates the WNT signalling pathway, which results in the initiation of ovarian differentiation and repression of SOX9 (via RSPO1). (B) When NR5A1 is mutated to the p.Arg92Trp or p.Ala260Val variant it impacts the regulatory effects of the NR5A1/β-catenin complex during ovarian development, enhancing repression of NR0B1 and WNT signalling. It is predicted that this allows SOX9 to escape repression and testis differentiation to occur.

3.3.5 Additional genomic variants may contribute to oligogenic inheritance Based on the wide phenotypic variation observed in individuals with NR5A1 variants, it is hypothesised that oligogenic inheritance plays a role (162, 263). I reanalysed MPS data from the targeted DSD gene panel to look for potential modifiers that may have an additive effect on gonadal development with NR5A1 variants. Sequencing data was filtered using a list of 116 NR5A1-related genes (see Methods 3.2.1). Sixteen potentially pathogenic variants were identified, summarised in Table 3.2. In DSD07, variants were found in CREBBP, GDF9, HSD3B1, STAR and TG. DSD17 had variants in AR, DACH1 and ZFPM2. DSD28 had two NR5A1-related variants; these were in the FRAS1 and MTSS1 genes. DSD21 had variants in BMP15, MSX2, PGR, POR, PTCH1 and RARA. These variants may act additively with NR5A1 to generate the DSD

108 phenotype, for example DSD21 presented with a heterozygous variant in Cytochrome P450 Oxidoreductase (POR) in addition to the NR5A1 p.Ala260Val variant. Variants in this gene have been associated with genital anomalies and combined POR (NM_000941: c.G1370A; p.Arg457His) and NR5A1 (p.Arg92Trp) variants have been reported in a case of 46,XX T-DSD before (148). Given their role in steroidogenesis, NR5A1 and POR variants may have an additive impact on steroidogenic function. This may also apply to other variants I identified in known steroidogenic genes, e.g. STAR, PTCH1 and HSD3B1.

109 Table 3.2. Additional genomic variants identified in 46,XX (ovo)testicular DSD cases.

Protein Inheritan In gnomAD DSD ID Gene Chr DNA change Consequence Zygosity dbSNP ID ClinVar change ce silico freq.

Conflicting interpretations of NM_001079846 rs6175338 DSD07 CREBBP 16 p.L513I missense Het NA 3/4 pathogenicity. Reported in: 0.009668 , c.C1537A 1 Rubinstein-Taybi syndrome.

NM_005260, rs6175458 DSD07 GDF9 5 p.P103S missense Het NA 3/4 Not found 0.002925 c.C307T 3

NM_000862, rs7747381 0.000166 DSD07 HSD3B1 1 p.Y225C missense Het NA 3/4 Not found c.A674G 58 2

NM_000349, rs1381612 DSD07 STAR 8 p.R274C missense Het NA 3/4 Not found 6.37E-05 c.C820T 53

NM_003235, rs1147818 0.000727 DSD07 TG 8 p.R152H missense Het NA 3/4 Not found c.G455A 69 1

Conflicting interpretations of NM_000044, rs2019346 pathogenicity. Reported in: DSD17 AR X p.P392S missense Het NA 3/4 0.004138 c.C1174T 23 Hypospadias, Partial androgen insensitivity syndrome.

NM_004392, DSD17 DACH1 13 p.N115K missense Het NA 0/4 NA NA NA c.C345A

NM_012082, rs2022172 Benign. Reported in: 46,XY sex DSD17 ZFPM2 8 p.D98N missense Het NA 2/4 0.002608 c.G292A 56 reversal 9.

110 Conflicting interpretations of pathogenicity, risk factor. NM_025074.6, p.R3269 rs6172936 DSD28 FRAS1 4 missense Het Maternal 3/4 Reported in: Cryptophthalmos 0.005271 c.G9806A Q 6 syndrome, Congenital diaphragmatic hernia.

NM_014751.4, rs5487929 DSD28 MTSS1 8 p.I396T missense Het Paternal 1/4 Not found 0.00124 c.T1187C 52

NM_005448, p.S261de inframe rs1118897 DSD21 BMP15 X c.782_783insTC Het NA NA Not found NA linsSL insertion 93 T

NM_002449, rs7805935 DSD21 MSX2 5 p.E32V missense Het NA 0/4 Not found 2.57E-05 c.A95T 93

NM_000926, rs2003221 DSD21 PGR 11 p.V221D missense Het NA 3/4 Not found 0.000291 c.T662A 78

NM_000941, rs3729552 0.000178 DSD21 POR 7 p.R570H missense Het NA 4/4 Not found c.G1709A 96 8

NM_000264, DSD21 PTCH1 9 p.G538R missense Het NA 3/4 NA NA NA c.G1612C

NM_000964, rs1165386 DSD21 RARA 17 p.T43S missense Het NA 0/4 Not found 0.001987 c.C128G 51

In silico (only available for missense variants) predictors: Mutation Taster, PolyPhen2, SIFT, LRT; fraction indicates proportion predicted deleterious or possibly deleterious. NA, not available.

111 3.4 Discussion NR5A1 acts at multiple points in gonad differentiation, functioning as both an activator and repressor of target genes. Variants in NR5A1 are associated with a wide range of DSDs from 46,XY gonadal dysgenesis to 46,XX POI, and now a SNV (p.Arg92Trp) contributes to 46,XX (ovo)testicular DSDs. Despite these recent reports, there is still no robust mechanism explaining how this NR5A1 variant activates testis differentiation in 46,XX individuals. In this chapter I described three additional 46,XX (ovo)testicular DSD patients with the p.Arg92Trp variant. In addition to this, I found a further NR5A1 variant (p.Ala260Val) in a patient with 46,XX OT-DSD. This is the first instance of an NR5A1 variant occurring in 46,XX (ovo)testicular DSD which does not affect codon 92.

The aim of this chapter was to characterise these two NR5A1 variants to establish a mechanism for how they might mediate XX sex reversal. I first showed that there was no major changes in protein expression or localisation (Fig. 3.5), however the in silico protein modelling indicated that binding to other molecules and targets was likely to be affected (Fig. 3.6). I therefore studied several sex differentiation pathways with in vitro luciferase assays to identify changes resulting from the NR5A1 variants.

As the bipotential gonads begin to differentiate, the Y-linked SRY gene activates the testis pathway; this allows the first male-specific cells, Sertoli cells, to form. We know that the SRY switch can be bypassed if the factor immediately downstream, SOX9, is upregulated (82, 264) and the NR5A1 product, SF1, transactivates SOX9. Previous functional analysis has shown that NR5A1-p.Arg92Trp is unable to activate the testis pathway via Sox9, or other male pathway genes Amh and Cyp11a1 (147). We showed that the NR5A1 p.Ala260Val variant also loses the ability to activate Sox9 (Fig. 3.6A), thus in vitro there is no evidence for these variants causing abnormal activation of the testis pathway. This is consistent with the fact that both of these variants have been identified in 46,XY gonadal dysgenesis, where testis pathway activation is lost (147, 260).

When there is no Y-chromosome present, the bipotential gonads differentiate towards an ovarian fate and testis development is actively repressed. The NR0B1 gene is a repressor of testis

112 development; specifically it antagonises the NR5A1 transactivation of SOX9 during ovarian development (91). A previous paper (148) hypothesised that the NR5A1 p.Arg92Trp variant is unresponsive to NR0B1, meaning SOX9 repression is lost in these cases. Conversely, I showed that each variant form of NR5A1 retains sensitivity to NR0B1 and Sox9 TESCO is repressed (Fig. 3.7B), suggesting this anti-testis function is retained in our patients. Bashamboo et al. (2016) proposed an alternate mechanism in which the NR5A1 p.Arg92Trp has less ability to activate the anti-testis NR0B1, perhaps via a loss of synergy with β-catenin. This would mean that testis factors like SOX9 would escape repression. I found that both NR5A1 variants maintained ability to upregulate the NR0B1 promoter at levels comparable to wild-type NR5A1 (Fig. 3.7C). However, when NR5A1 and β-catenin were simultaneously introduced, variant NR5A1 caused repression of NR0B1 compared to wild-type. This suggested that the WNT signalling pathway may be affected by variant NR5A1.

NR5A1 also plays a role during ovarian differentiation. A hypomorphic Cited2-/- XX mouse which models Nr5a1 depletion showed a reduction in expression of Wnt4, Rspo1 and Foxl2 (149), suggesting that Nr5a1 initiates ovarian differentiation either directly or indirectly. It was previously shown that the NR5A1 p.Arg92Trp variant causes reduced WNT signalling compared to wild-type NR5A1, which is thought to be via reduced synergy with β-catenin (147). I also show that WNT signalling is repressed in both variant forms of NR5A1 (Fig. 3.7E). However I also observeD that introduction of wild-type NR5A1 and β-catenin represses WNT signalling compared to β-catenin alone, suggesting that the NR5A1/β-catenin complex can both activate and repress WNT signalling. This may be influenced by subtle changes in gene dosage or environment, highlighting how these in vitro assays are not the optimal system for characterising variants. Despite this, in vitro analysis of these variants indicates that two sex differentiation pathways are likely to be impacted in the patients described, WNT signalling and NR5A1/β-catenin upregulation of NR0B1 (Fig. 3.8). These impacts would result in male pathway factors escaping repression, allowing testis development to proceed in an XX individual.

As NR5A1 has a large number of target genes it is possible that other pathways not investigated here are involved in the phenotype too. Inaccessibility of gonadal tissue makes this difficult to assess, as do the significant differences in mouse versus human NR5A1 during sex differentiation. Indeed, the CRISPR generation of a mouse model carrying the p.Arg92Trp variant exhibited no gonadal phenotype (265). Studying additional DSD patients with these variants may

113 provide insight into the underlying mechanism. Interestingly another heterozygous variant has been reported at codon 92 (p.Arg92Gln) in 46,XX individuals with adrenal insufficiency with or without sex reversal (266, 267). As with the p.Arg92Trp variant, the p.Arg92Gln loses the ability to activate male target genes such as Amh and Cyp11a1 (268), yet it does not repress WNT signalling like the p.Arg92Trp variant does (147). Also, the expression of adrenal phenotypes only in individuals with the p.Arg92Gln variant suggests that these two residue changes at codon 92 can differentially impact target genes. Long-term follow up of cases could confirm this, as adrenal insufficiency may not present until later in life. DNA binding studies could be useful to compare how each of these A-box motif variants affects binding specificity.

So far 46,XX individuals with the p.Arg92Trp variant have covered a wide spectrum of phenotypes, for example in gonadal phenotype this has ranged from normal ovary, streak gonad to testis (146-148). Variable expressivity and incomplete penetrance are characteristics of NR5A1-associated disorders, because adreno-gonadal development is exquisitely sensitive to NR5A1 gene dosage (19), modified by genetic or environmental factors. These variants are absent or at low frequency in large population datasets (Table 3.1) and individuals with the p.Arg92Trp have diverse ethnic backgrounds (African, Hispanic, European and Asian), suggesting that genetic background or founder effect do not explain the variable expressivity. A recent paper reported 46,XX siblings with the p.Arg92Trp variant but different gonadal and genital phenotypes (256), suggesting that environment is an important modifier. In our analysis for NR5A1-related genomic modifiers I revealed an additional 2-6 variants per patient (Table 3.2). While a more rigorous variant filtering strategy is necessary to control for false discovery, it is plausible that such variants could act additively with NR5A1 or modify its expression to cause the wide phenotypic variability. Functional assessment would be required to confirm their pathogenicity and interaction with NR5A1 in these phenotypes. In future studies, whole genome or exome sequencing on these and additional individuals (affected and unaffected) with NR5A1 variants would enable a more thorough and unbiased approach to understanding oligogenic inheritance.

I have shown here that pinpointing the exact pathway(s) affected in the patients described is inherently difficult as we cannot easily examine the tissue of interest (human embryonic gonad) and environmental or genomic modifiers likely play a significant role. This highlights the need for improved models to study pathogenicity of DSD variants in vitro. Despite this, these results

114 advance our understanding of the underlying mechanism and add novel data to the pool of patients with NR5A1 variants and 46,XX sex reversal. Further, by screening a large cohort of SRY- negative 46,XX (ovo)testicular DSDs (24 cases), I have revealed that NR5A1 variants contribute to 16% (4/24) of these cases. This is comparable to SOX9 enhancer duplications, which underlie 20% (5/24) of cases in this cohort. Consequently, screening for NR5A1 variants should be included in routine genetic testing for these patients.

115

116 Chapter 4: Differentiation of human pluripotent stem cells to gonadal lineages

4.1 Introduction In chapters 2 and 3 I highlighted our enhanced ability to identify candidate gene variants for DSD. However, at present our ability to assess their pathogenicity or contribution to disease is hampered by a lack of appropriate models in which to test them. This is a common theme across clinical genomics and in recent years has led to extensive research on differentiation of human iPSCs and ESCs to specific cell lineages and tissues. Specifically, many of these studies have taken a directed differentiation approach whereby pluripotent cells are guided through the embryonic cell populations that give rise to the tissue of interest. This approach has led to the successful induction of a wide range of human tissues including cardiac (269), lung (270), pancreatic (271), intestinal (272), cerebral (187), and renal lineages (273). Research at MCRI, led by Prof. Melissa Little, recently differentiated iPSCs to human renal lineages, leading to the generation of kidney organoids that closely model the human embryonic kidney (275). This was done by guiding iPS cells through the developmental cell populations that give rise to the renal lineages. This first begins by differentiating iPSCs to the posterior primitive streak (induced by WNT signalling activator CHIR99021), as assessed by the expression of BRACHYURY/T and MIXL1. The posterior primitive streak gives rise to the mesoderm of the body axis, which are defined from along the mediolateral axis as paraxial, intermediate and lateral plate mesoderm (Fig. 4.1) (274). The differentiating cultures were guided towards an intermediate mesoderm lineage, as determined by the co-expression of LHX1, PAX2 and OSR1, via the addition of FGF9 (188, 273, 275). After 7 days of differentiation, cells are dissociated and reaggregated into aggregates termed ‘organoids’ and cultured on Transwell filters. After 15 days of differentiation these organoids develop distinctly kidney-like structures and functions. Using such approach generates improved disease models and allows us to learn about development of human organs and tissues in much more detail.

117 Figure 4.1. Shared developmental origins of gonad and kidney.

At present, a modification of this approach has not been applied to human gonad development. I believe such an approach has potential for developing an in vitro model for DSD. As well as this, we know that the mammalian gonad has its developmental origins in a region close to the intermediate mesoderm (IM) (Fig. 4.1) (8, 276); we therefore postulated that we could take this human kidney differentiation protocol and adapt it to generate gonad lineages. The aim of this chapter was therefore to develop a novel protocol for the differentiation of testis lineages from human iPSCs. Ideally I would like to differentiate the major cell types of the testis in a single differentiation experiment, with the exception being germ cells, as these arise from the hindgut and migrate to the bipotential gonad (by E10 in mouse or week 7 in humans). Furthermore, I would like to generate human testis organoids that could model the effects of DSD variants on testis development. We chose the testis rather than the ovary as we have a greater understanding of its development and structural characteristics early in development, which will be important for monitoring success within organoid modelling.

In this chapter, I first evaluated whether the testis is compatible with organoid culture by assessing the self-organising potential of embryonic mouse testis cells following dissociation and reaggregation. I then identified specific markers for the gonadal structures that I aimed to develop by analysing gene expression data and performing immunofluorescence staining on human fetal testis sections. I then took the kidney differentiation protocol as a starting point

118 and drew on knowledge of gonad development to explore conditions that may promote the induction of bipotential gonad, followed by testis lineages. This process also led to insights into the origins of the human gonad and important regulatory interactions during this organ’s development.

4.2 Methods

4.2.1 Mouse dissociation and reaggregation Multiple litters of Swiss brown mice were bred at the MCRI animal house facility; embryos were collected then stored and dissected in 1 x PBS. Mouse gonads were dissected at E12.5, E13.5 and E14.5 and sexed visually. Tissues were disrupted with a 21-gauge syringe and treatment with Accutase enzyme (ThermoFisher, #A1110501) for 15 minutes at 37oC. Following complete dissociation of gonadal tissue, cells were pelleted (2000 rpm for 5 minutes) and resuspended. Cell suspensions were strained using a 35 mm cell strainer (In vitro technologies, #FAL352235) and cell density was estimated using a haemocytometer. The cells were then pelleted (2000 rpm for 2 minutes twice, with 180o rotation between spins) and transferred to a Transwell filter (Sigma-Aldrich, #CLS3450-24EA) using P200 wide bore tips (Molecular BioProducts, #3531) in

CO2-independent media (Life Technologies, #18045088) (+ Pen/Strep 25 u/ml (Thermo Fisher Scientific, #15070-063)) media for 48-72 hours, changing media every 2 days. Reaggregated cell pellets (attached to the filter) were stained and imaged for immunofluorescence using protocol for whole mount immunofluorescence (Methods 4.2.2), with primary antibodies shown in Table 4.1.

Table 4.1 Primary antibodies used for immunofluorescence staining.

Antigen Host Supplier Catalogue Dilution number

AMH Mouse Santa Cruz sc-166752 1:100

E-Cadherin Mouse BD Biosciences 610181 1:200

GATA4 Mouse Santa Cruz sc-25310 1:400

HSD3B1 Mouse Santa Cruz sc-515120 1:500

119 Laminin Rabbit Sigma L9393 1:500

MVH Rabbit Abcam ab13840 1:5000

NR5A1 Mouse Santa Cruz sc-393592 1:100

OCT4 Goat Santa Cruz sc-8629 1:600

PECAM Rat Abcam ab7388 1:200

Rabbit Abcam ab5535 1:5000 SOX9 Goat R&D Systems AF3075 1:300

STAR Rabbit Proteintech 12225-1-AP 1:20

Goat R&D Systems AF5729-SP 1:300 WT1 Mouse Santa Cruz sc-393498 1:100

4.2.2 Mouse whole mount immunofluorescence Dissected gonads were transferred to a 2 ml eppendorf tube in 1 x PBS on ice. PBS was removed and tissue fixed in 4% PFA/PBS (Australian Biostain, #30525-89-2) for 10 minutes at room temperature. PFA was removed and tissue and washed with PBS once then PBTX (PBS with 0.1% Triton-X) twice for 30 minutes each. Blocking was performed for 2-3 hours in blocking solution (PBTX with 10% donkey serum (Sigma, #D9663)) rocking at 4oC. Primary antibodies were added in fresh blocking buffer (600 μl total volume) and incubated overnight, rocking at 4°C. Tissues were rinsed twice then washed 5 times for 1 hour each with 1 ml PBTX, rocking at 4°C. Antigen species-specific secondary antibodies were prepared (1 in 1000) in blocking solution and incubated overnight, rocking at 4°C. Secondary antibodies were removed, followed by rinsing and washes as per primary. Tissues were incubated with DAPI (Life Technologies, #D1306) in PBS (5 μL in 200 ml PBS) for 2 hours, rocking at 4°C, then washed three times for 2 minutes each with PBTX. Mounting was done in 80% glycerol on glass bottom dishes (MatTEK, #P35G-0-14-C) and imaging performed using confocal microscopy (LSM780, Zeiss).

4.2.3 Paraffin sectioning and staining Human fetal testis tissues were embedded in paraffin and sectioned by the Copenhagen Hospital (Dr. Anne Jørgensen). Slides underwent De-Wax treatment, consisting of two 2 minute washes in Xylene, two 2 minute washes in 100% ethanol, 1 minute in 90% ethanol, 1 minute in 80% ethanol, 1 minute in 70% ethanol, 1 minute in 50% ethanol and one dip in distilled water, slides

120 were then stored in 1 x PBS. Antigen retrieval consisted of two 5 minute washes in 0.1 M Citrate buffer (2.94 g trisodium dehydrate in 900 ml distilled water, at pH 7) in the microwave on high, replacing citrate buffer between washes. Slides were then allowed to cool while remaining in buffer for at least 20 minutes. Slides were placed in humidified incubation chamber and tissue circled with DAKO (hydrophobic) pen, 100 μL blocking buffer (5% BSA + 10% Horse serum in 0.1% PBTX) was added to each slide, this was incubated at room temperature for 2 hours. Blocking buffer was replaced with primary antibodies in diluent (blocking buffer:PBTX 0.1% at 1:4 ratio) and incubated overnight at 4oC. Slides were washed in 1 x PBS with a magnetic stirrer and secondary antibodies were added in diluent, this was incubated in a humidified chamber at room temperature protected from light for 2 hours. Slides were washed in PBS for 5 minutes and DAPI added (4 μL in 200 ml PBS) in a black box for 3 minutes at room temperature. Slides were washed three times in PBS for 5 minutes each. PBS was removed and 1-2 drops of Fluorsave added (Merck Millipore, #345789) before covering with a cover slip. Slides were stored in a light sensitive slide book until confocal imaging.

4.2.4 Reagent preparation NT2D1 condition media

NT2D1 cells were grown on a T25 cell culture flask (Nunc, #156367) in 10 ml E6 media (supplemented with 10% FCS); spent media was collected daily and stored at 4oC. NT2D1 cells were split (1/3 to 1/4) with Versene (1 ml for T25 flask for 5 minutes at 37oC) (Gibco, #15040066) when they reached 80-100% confluency. Condition media was made up of NT2D1 spent media:E6 media at 1:1 ratio.

iPSC culture media

APEL/APEL2 media

APEL (Stem Cell Technologies, #5210, discontinued)/APEL2 media (Stem Cell Technologies, #05275) was made by adding 0.5 ml Antibiotic-Antimycotic (Thermo Fisher Scientific, #15240- 062) to 100 ml of APEL/APEL2 media. Protein-free hybridioma (PFHM-II) (Thermo Fisher, #12040077) was also added (at 4%) to APEL2 only. Stored at 4oC for up to 2 weeks.

121 E6 media

E6 media was made by adding 500 μL Holo-Transferrin (Sigma, # T4132) and 1 ml of Insulin (Sigma, # 91077C-1G) to 500 ml DMEM-F12 media (Life Technologies, #11330-032), followed by filter sterilisation. This is stable at 4oC for several months.

Swirler culture media

Swirler culture media is made with E6 media supplemented with 0.1% PVA (Sigma, #P8136), 0.1% Methyl cellulose (Sigma, #M0262) (pipette this with wide bore tips). Filter sterilise and store at 4oC. Rho Kinase Inhibitor (ROCKi) Y27632 (Stem Cell Technologies, #72307) (1:1000 dilution) is required for first day in swirler culture to prevent cell death.

Matrigel-coated cell culture flask

To prepare Matrigel-coated tissue culture flasks/plates, for a T25 flask add 30 μL of hESC- qualified Matrigel (Corning, #354277) to 3 ml chilled DMEM/F-12 (Thermo Fisher Scientific, #11320-082). Mix well and transfer to a T25 tissue culture flask and leave at room temperature for at least 30 minutes. NB: Matrigel must be kept on ice or it will solidify.

EDTA (Thermo Fisher, #15575020)

Make 0.5 mM working solution by diluting 1:1000 in PBS and store at room temperature.

CHIR99021 (R&D, #4423/10)

Centrifuge the tube at room temperature for 5-10 seconds at 1000-3000 x g. Reconstitute 10 mg of CHIR99021 in 2.149 ml DMSO (Sigma Aldrich, #D5879) to make a 10 mM stock. Prepare 15 μL aliquots and store at -20oC for up to a year.

122 Heparin (Sigma Aldrich, #H4784-250MG)

Reconstitute to 1 mg/ml in MilliQ water and filter sterilise it through a polyethersulfone 0.22 μm syringe-driven filter unit (Merck Millipore, #SLMP025SS). This can be stored at 4oC for over a year.

FGF9 (R&D, #273-F9-025)

Centrifuge the tube at room temperature for 5-10 seconds at 1000-3000 x g. Reconstitute to 100 μg/ml in filtered DPBS (Thermo Fisher, #14190-144) containing 0.1% (weight/volume) human serum albumin (Albucult) (Novozymes, #230-005).

BMP4 (R&D, #RDS314BP010)

Centrifuge the tube at room temperature for 5-10 seconds at 1000-3000 x g. Add 100 μL HCl (with 0.1% Albucult) to 10 μg of BMP4, flick gently and swirl to resuspend. Spin again and make 5 μL aliquots, freeze at -80oC for up to 3 months.

NOGGIN (R&D, #6057-NG-025)

Reconstitute at 250 μg/ml in PBS containing 0.1% Albucult, freeze at -80oC for up to 3 months.

PGD2 (Cayman chemical, #12010)

Reconstitute to a 1mg/ml stock in DMSO and store at -20oC.

Smoothened agonist (SAG) (Merck, #566660)

Reconstitute to 200 uM in PBS and store at -20oC.

123 Retinoic acid (RA) (Sigma, #R2625)

Reconstitute to 3 mg/ml solution in DMSO and store in light protected vials at -20°C.

4.2.5 Maintenance of iPSCs

4.2.5.1 iPSC lines Experiments were performed in two feeder-free human iPSC lines: CRL1502 clone C32 (female) and PCS_201_010 iPS clone 5 (male), these were checked for SNPs and tested for mycoplasma infection regularly, with effort made to avoid reaching high passage number and iPSCs from same passage number being used across experiments to reduce variability. PCS_201_010 is derived from a foreskin fibroblast line obtained from ATTC (American Type Culture Collection, USA); CRL1502.3 is from E.J. Wolvetang (The University of Queensland, Australia) generated from patient fibroblasts via episomal reprogramming.

4.2.5.2 Frozen stock of human iPSCs For expanding human iPSCs before cryopreserving them, cells were cultured in xeno-free Essential-8 (E8) medium (Thermo Fisher, #A1517001) on Matrigel-coated surface. Once cells reached the desired confluency they were harvested using TrypLE Select (Thermo Fisher, #12563011), transferred to a 15 ml falcon tube and centrifuged for 3 minutes at 400 x g. Cells were resuspended in 10% DMSO/90% FBS and split into cryo-vials (1 ml per vial), aiming for 1.5- 2 x 106 iPSCs per vial. Vials were placed in Frosty-boy at -80oC overnight then transferred to liquid nitrogen for long term storage. iPSCs were thawed by defrosting cryovial in 37oC water bath. 1 ml E8 media was then added drop-wise to the cryovial with a 2 ml pipette and this was transferred to a 15 ml falcon and centrifuged at 1.5 rpm for 3 minutes. Media was aspirated and cells resuspended in 3 ml E8 medium then transferred to one Matrigel-coated well of a 6 well plate. Cells were dispersed by moving plate twice in figure eight motion in 37oC incubator. iPSCs were expanded in E8 medium on Matrigel-coated culture dishes and passaged using EDTA until reaching seeding confluency required for differentiation (70-80% confluent in a T25 flask).

124 4.2.6 Differentiation of iPSCs into gonadal lineages

4.2.6.1 Plating iPSCs for differentiation (Day -1) iPSCs were washed twice with 3 ml PBS and harvested by incubating in TrypLE Select (1 ml per T25) at 37oC for 3 minutes. Once cells were dislodged, 3 ml E8 was added and pipetted up and down 5-6 times to obtain single cell suspension followed by transfer to 15 ml falcon. T25 flask was rinsed with a further 3.5 ml E8 and this was added to same 15 ml falcon, then centrifuged tube for 3 minutes at 1.5 rpm. Pellet was resuspended in 3 ml E8 and 10 μL aliquot taken for cell count using a haemocytometer. iPSCs were plated at ~10,000 cells per cm2. Required cell number was removed and placed in a separate 15 ml falcon, topped up with required volume of total E8 (for a total of 600 μL resuspended cells per 24-well or 3 ml per 6-well, etc.) as well as Revita cell (at 1:100 dilution (Life Technologies, #A2644501). Cells were gently mixed and aliquoted into wells ensuring cells were distributed by moving plate in figure eight pattern. iPSCs incubated at 37oC for approximately 24 hours before differentiation began.

4.2.6.2 Inducing the embryonic mesoderm (Days 0-7) E8 media was aspirated and replaced with room temperature E6 media containing CHIR99021 at desired concentration (3-5 μM). Cells were cultured in 37oC incubator for 3-5 days changing media every 2 days, shorter duration and lower concentration of CHIR99021 will favour anterior mesoderm development, while longer duration/higher concentration favours posterior. After CHIR treatment culture medium was changed to E8 supplemented with FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml), keeping these growth factors on ice is essential. Cells cultured for 7 days total to reach embryonic mesoderm, with media changed every two days.

4.2.7 Organoid culture 4.2.7.1 Air liquid interface organoid culture (NB: calculations are for cells grown in a 6 well plate)

Media was aspirated and cells washed twice with PBS. 1 ml trypsin EDTA (0.05%) (Thermo Fisher Scientific, #25300-054) was added to cells followed by incubation at 37oC for 2 minutes, checking regularly to see when cells are detached from surface. To neutralise, 1 ml FDMEM (Thermo Fisher Scientific, #11320-082) was added and this pipetted up and down then added to a new 15 ml falcon. Another 2 ml FDMEM was added to rinse plate and this was added to same falcon.

125 A 10 μL aliquot was taken for a cell count using a haemocytometer. Organoids were made of 0.1 – 0.25 x 106 cells, volume required for desired organoids was removed plus extra (n = desired organoids + 2) and spun in a 15 ml falcon for three minutes at 1.5 rpm. Media was aspirated and cells resuspended in 300 μL x n, this was then added to n tubes. Tubes were spun at 1.8 rpm three times for 3 minutes each, twisting 180o between each spin. Costar Transwell filters were placed into 6 well plates (Corning, #3450) and p200 wide bore tips were used to transfer cell pellets onto Transwell while minimising medium carry over. Up to four organoids were placed on each filter. 1.2 ml E6 media (supplemented with growth factors) was added beneath the filter for subsequent differentiation and this was incubated in 37oC incubator, with media changed every two days. To harvest for RNA, organoids attached to filter were removed using a scalpel blade and submerged in 1 ml Trizol reagent. For immunofluorescence, organoids attached to filter were submerged in 4% PFA/PBS, staining was then performed as per mouse whole mount immunofluorescence (Methods 4.2.2).

4.2.7.2 Swirler culture iPSCs were differentiated according to the optimised protocol for the first 10 days in monolayer culture in 6 well plates, aiming to reach at least 5-10 x 106 cells as high cell density achieves more uniform organoids. Swirler media was prepared as outlined in 4.2.4. Cell monolayers were washed with 1 ml of EDTA solution per well of a 6 well plate, gently swirling. EDTA was aspirated and replaced with 1 ml fresh EDTA, followed by incubation at 37oC for 2-3 minutes, cells would become circular but not completely detached. EDTA solution was removed and 2 ml swirler media immediately added to well, cells were detached by gentle pipetting twice with Gilson pipette. Cells were passed through a 40-μm sieve (In vitro Technologies, #FAL352340) into 6 cm low attachment plates (Corning, #CLS3261). Sieve was washed with 3 ml swirler media to get total of 5 ml per dish. Dishes were placed in a humidified plastic tablet and cultured swirling at 60 rpm in a 37oC incubator. Swirler media was replaced after 24 hours, removing ROCKi; media was then changed every two days.

4.2.8 Immunofluorescence staining of monolayer differentiations Cells were fixed with 4% PFA/PBS for 20 minutes at room temperature then washed in PBS three times for 5 minutes each. Blocking buffer was added (0.1% PBTX + 5% BSA + 10% Horse serum) for 2 hours at room temperature. Primary antibodies were added in diluent (1:4 Blocking

126 buffer:PBTX) overnight at 4oC. Cells were washed three times with PBS for 5 minutes each. Secondary antibodies were diluted as per primaries then incubated for at least 4 hours at 4oC. Cells were washed in PBS for 5 minutes then DAPI added (5 μL in 200 ml PBS), followed by a final PBS wash. Imaging was performed on a confocal microscope.

4.2.9 Generation of reporter iPSCs This iPSC line was generated by Yudha Patria in collaboration with Shireen Lamande, John Bateman, Ed Stanley and Andrew Elefanty. The reporter locus was generated using the GAPTrap system (277), specifically the CRISPR guide was designed to target TdTomato a few base pairs downstream of the SOX9 exon 3 stop codon. This was then introduced into a human male iPSC line (PB001.1), generating the heterozygous reporter line SOX9-TdTT-CRE#13. This line was adapted to feeder-free culture.

4.2.10 RNA extraction and Reverse Transcription RNA was extracted using TRIzolTM Reagent (ThermoFisher Scientific, #15596018), RNA extraction performed according to manufacturer’s instructions. Integrity and concentration of RNA was measured by UV spectrophotometry (NanoDrop ND-1000, Thermoscientific). To avoid amplification of genomic DNA, RNA was DNase treated (Promega, #M6101). RNA quantity was diluted to the lowest concentration RNA in the sample group to normalise for input amount in the downstream qRT-PCR experiment.

4.2.11 Real-time Quantitative Reverse Transcriptase PCR Reverse transcription was carried out using the GoScript Reverse Transcriptase system (Promega, #A5001) with random hexamer primers. qRT-PCR analyses were performed with GoTaq® qPCR Master Mix (Promega, #A6001) on the LightCycler480 (Roche) with GAPDH used as a reference gene and NT2D1 cDNA used as a positive control. Absolute data was normalised to GAPDH and then normalised to a specified sample (ΔΔCt method). Primers were designed using Primer Depot, sequences can be found in Appendix 8. Efficiency of each pair was established using cDNA derived from two populations of human testicular-like cells (278, 279).

127 4.3 Results

4.3.1 Testing the self-organising ability of embryonic testis in mouse To maximise clinical and research utility, this project aims to differentiate and maintain several testis cell types that can also assemble into testis-like structures (testis organoids) during the differentiation process. To demonstrate the potential for this, we performed a dissociation/reaggregation experiment on embryonic mouse gonads, culturing these reaggregates on Transwell filters (as per the human kidney organoid protocol) for 48 hours. This has been performed previously (280), however we wanted to characterise the generated structures using modern techniques (immunofluorescence) to provide a reference point for what kind of structures may be observed in human testis organoids. Previous studies have indicated that later stage gonads (over E14.5) (280) lose self-organising potential; therefore we used gonad tissue from E12.5, E13.5 and E14.5.

128

Figure 4.2. Embryonic mouse testis shows self-organisation potential. E12.5 mouse testis was dissected with (A-E) and without mesonephros (F-J) then dissociated, reaggregated and cultured for 48 hours. Reorganisation of reaggregates was assessed by bright field imaging (A, F) and immunofluorescence staining (B-E, G-H). A non-dissociated male gonad (labelled ‘g’) with mesonephros (labelled ‘m’) attached is shown for comparison (K, L). Immunofluorescence staining included antibodies for Sertoli cells (SOX9 or AMH, green) and germ cells (E-cadherin or OCT4, red), with DAPI (blue) marking cell nuclei. Scale bars in B, D, K, G and I represent 100 μM. Scale bars in C, E, L, H and J represent 50 μM.

Initially we wanted to see whether dissociated testis cells could reassemble when dissected with or without the mesonephros. The mesonephroi are paired structures that lie lateral to the gonads within the urogenital system. Some cells from the mesonephros contribute to the

129 formation of the testis (281). The testis was dissected and dissociated into single cells at E12.5 both with and without the mesonephros (Fig. 4.2). In each of these conditions I saw branching web-like networks forming after 48 hours of culture (Fig.4.2A, F), following immunofluorescence staining I saw that these are made up of SOX9+ Sertoli cells (Fig. 4.2B, D, I). The organisation of these cells may be related to initiation of tubulogenesis, this would be consistent with this time point and structure observed (200). In addition, the Sertoli cells re-organise into bundles with E- Cadherin+ germ cells within them (Fig. 4.2E, H), similar to the testis cords observed in a non- dissociated testis at this time point (Fig. 4.2K, L). The similar results observed when testis was dissociated with or without the mesonephros suggests that the mesonephros is not essential for self-organisation of embryonic testis cells. The mesonephros is known to contribute essential signals and cell types during testis differentiation but I show here that its presence is not required for self-organisation of testis in vitro.

I then assessed how this self-organisation potential changes through development, dissecting testis (with no mesonephros) at later embryonic stages (E13.5 and E14.5). At E13.5 I saw that multiple cell types communicate to re-form distinctly gonad-like structures, with similar branching and cord-like structures observed as in E12.5 (Fig. 4.3A-D). However by E14.5 (Fig. 4.3G-J), the cell-cell recognition between Sertoli (AMH+) and germ cells (MVH+) is significantly reduced, shown by a lack of structures forming between these two cell types, despite a sufficiently high starting cell number (407,500 cells). These results are consistent with previous dissociation-reaggregation studies (280) on later stage mouse gonads, where germ cells were clustered on the periphery of the reaggregates, failing to associate within testis cords. It is thought that this is due to changes on the surface of Sertoli cells, leading to lower affinity with germ cells. In regard to developing a human testis organoid protocol, this indicates that it will be important for cells to be dissociated and reaggregated into organoids at a stage before the human equivalent of E14.5, i.e. when the cells are still early in testis differentiation. As it seems that testis cells at this stage have higher organisation and migration potential, which will be essential if I want to recreate testis-like structures in vitro.

130

Figure 4.3. Self-organisation ability of mouse testis declines after E14.5. Mouse testes were dissected at E13.5 (A-F) and E14.5 (G-L) then dissociated, reaggregated and cultured for 48 hours. Reorganisation potential was assessed by immunofluorescence staining (A-H), with comparison to a non-dissociated testis (E, F, K, L). Immunofluorescence staining included antibodies for Sertoli cells (SOX9 or AMH, green) and germ cells (E-cadherin, OCT4 or MVH, red), with DAPI (blue) marking cell nuclei.

131 In summary, these experiments have demonstrated the self-organising ability of the embryonic testis, here we have seen how embryonic testis cells behave when they lose their cell-cell interactions and must reform a functional structure. As the early gonad cells in our organoid differentiation will undergo a similar stress when forming organoids, this study will act as a useful structural and antibody staining reference.

4.3.2 Identification of markers specific for gonad lineages and cell types To successfully model gonad development from iPSCs we need markers specific for each cell- type or stage of gonad development. During embryogenesis, a relatively limited number of signalling pathways are used for all organs. It is the cross-talk, dosage and spatiotemporal timing of these pathways that leads to the diverse structures that make up the human body. Consequently, characterisation in iPSC differentiation needs to be performed with markers that are highly specific to the desired tissue or cell type. I therefore devised a panel of genes that can be used to characterise the populations I aimed to generate in this differentiation protocol.

In this protocol, I will be guiding cells from the mesoderm lineages to the early gonad progenitor cells (or bipotential gonad lineages) and onwards to distinct testis cell types. The primary aim is to create Sertoli-like cells followed by Leydig cells, as these are the first testis lineages to differentiate. I first drew from literature on mouse gonad development and defined a key list of genes known to be essential for normal gonad development (Table 4.2). I then assessed how specific these markers are to the gonad by evaluating gene expression in closely related tissues. Examining gene expression in human fetal kidney revealed an overlap for several of these genes, such as EMX2 and WT1 (Fig. 4.4), while other markers like GATA4, AMH and NR5A1 were not expressed. In addition to kidney, the adrenal and gonad also arise from the same progenitor population of cells, the adrenogonadal primordium, which divides into the adrenal cortex primordium and bipotential gonad around E10.5 in mouse. A microarray study on human fetal adrenal gland, testis and ovary (282) highlighted GATA4, EMX2, WT1 and LHX9 as markers selectively upregulated in gonad and not adrenal. While NR5A1, NR0B1, HSD3B2 and STAR are also expressed in developing adrenal lineages. Together, this highlights the need for a panel of markers, which when used in combination provides us with more confident characterisation.

In addition to identifying markers specific to gonad, Appendix 9 outlines markers used to define the precursor populations (pluripotency, primitive streak and mesoderm) during differentiation.

132 Ialso selected several Hox genes (Appendix 9) known to be involved in gonad development (discussed below) to give us further insight into embryonic patterning of gonadal lineages. Finally I chose markers of tissues that also arise from the intermediate mesoderm (adrenal, kidney and chondrocyte) (Appendix 9), these will be used to screen differentiations for unwanted cell types.

Table 4.2 Markers of the human fetal gonad for characterisation of gonad differentiations.

Gene Protein Expression Mouse knockout gonadal phenotype Expression in other mouse fetal function localisation tissues (MGI expression database)

Emx2 Transcription Bipotential Emx2-/- mice have absent gonads, genital High expression in nervous, factor gonad tracts, kidneys and ureters. (36). urinary and visceral organ system.

Gadd45g Nuclear Bipotential Male to female sex reversal on B6 High expression in nervous protein, gonad background (283, 284). system. possible cofactor

Gata4 Transcription Bipotential Gata4-/- mice die before E9.5, Gata4 High expression in alimentary, factor gonad conditional KO - coelomic epithelium of cardiovascular and visceral organ mutant embryos remains as a single layer systems. of cells (5).

Hsd3b2 Enzyme Bipotential Not done Low level expression in gonad liver/biliary and visceral organ system.

Lhx9 Transcription Bipotential Lhx9 exon 2/3 deletion - all mice High expression in central neural factor gonad phenotypically female, no gonads (26). system and visceral organ system.

Nr0b1/ Nuclear Bipotential Nr0b1-deficient mice have disorganised High expression in visceral organ Dax1 receptor gonad and incompletely formed testis cords by system. E13.5 (285).

Nr5a1/ Nuclear Bipotential Nr5a1-/- mice do not develop bipotential High expression in nervous and Sf1 receptor gonad gonads or adrenal glands (17). visceral organ systems. Transcription factor

Wt1 Transcription Bipotential Wt1-/- mice fail to develop bipotential High expression in urinary and factor gonad gonads (35). visceral organ system.

Zfpm2/ Cofactor of Bipotential Zfpm2-/- mice have male to female sex High expression in nervous and Fog2 Gata4 gonad reversal (92). visceral organ system.

Amh Hormone Testis Amh-/- XY mice have Leydig cell High expression in visceral organ (Sertoli hyperplasia and develop female system. cells) reproductive organs (286).

Dhh Signalling Testis Dhh-/- mice have abnormal peritubular High expression in visceral organ molecule (Sertoli and tissue leading to disrupted testis cord system. Leydig cells) formation (287).

133 Fgf9 Signalling Testis Fgf9-/- mice have male to female sex High expression in alimentary, molecule (Sertoli reversal and abnormal development of nervous, sensory and visceral cells) Sertoli cells (77). organ systems.

Hsd3b1 Enzyme Testis Not done High expression in visceral organ (Leydig cells) system.

Sox9 Transcription Testis Sox9-/- XY mice show male to female sex High expression in alimentary, factor (Sertoli reversal (288). musculoskeletal, nervous, cells) respiratory, sensory, urinary and visceral organ systems.

Star Hormone Testis Star-/- XY mice have female external High expression in visceral organ (Leydig cells) genitalia and adrenocortical insufficiency system. (289).

Foxl2 Transcription Ovary Postnatal conditional Foxl2 KO causes High expression in visceral organ factor (granulosa transdifferentiation of ovarian somatic cell system. cells) lineages (58).

Foxl2-/- goats show female to male sex reversal (52).

Rspo1 Signalling Ovary Rspo1-/- XX mice show partial female to High expression in nervous and molecule (granulosa male sex reversal (45). visceral organ system. cells)

134

Figure 4.4. Expression heatmaps of gonad markers in mouse and human fetal kidney. (A) Gene expression across six cell clusters within week 16 human fetal kidney (290). (B) Gene expression across 13 cell clusters within human kidney organoids. (C) Gene expression across 14 cell clusters within the embryonic day 18.5 mouse kidney. Scales represent log fold change of differential expression relative to all other cell clusters. Data in B and C is from a single-cell RNA seq study (291). Plots were generated by Dr. Alex Combes (MCRI).

To anticipate the expression patterns of markers I may expect to see in our human testis differentiation I looked at their expression profiles through gonad development in mouse. Figures 4.5 shows the expression profile of genes during mouse gonad differentiation, based on bulk RNA-seq data (219). While we know that there are some differences in human versus mouse gonad development, this gives an idea of how the expression profile of our gonad- specific markers may change over the course of in vitro differentiation.

135 Markers of the bipotential gonad (Fig. 4.5A) consistently show a peak in expression at E11.5 in the mouse testis, then decrease but maintain high expression as the testis begins differentiation. Similarly, Sertoli cell markers Sox9 and Fgf9 peak at E11.5 (Fig. 4.5B), we know that Sox9 is one of the earliest regulators of testis development (peaking at week 7-7.5 in humans (292)) to be expressed downstream of Sry. Despite the high-specificity of Sry to the testis, I avoided this as a marker as it has no introns therefore is difficult to accurately assess via qRT-PCR. Furthermore, we know that testis development can progress in the absence of Sry if Sox9 is upregulated instead (82, 264). Expression of Leydig markers Star and Hsd3b1 peak at later time points, consistent with developmental timing where Leydig cells differentiate in response to Sertoli cell signalling. Some of these markers are highly specific to testis, with no expression shown in ovary for markers Amh, Hsd3b1 and Dhh. Ovarian markers are Foxl2 and Rspo1 (Fig. 4.5C); Rspo1 is expressed at low levels in testis and given that it responds to WNT signalling this pathway may also be active in developing testis.

I also included Hox genes, given their important role in specifying embryonic patterning. Figure 4.5D shows Hox genes reported to be expressed in early gonad lineages, Hoxd9/Hox-5.2, Hoxc5/Hox-3.4 and Hoxc6 (293-295). Previous studies showed that Hoxd9 is expressed from E10.5 in the coelomic epithelium and mesonephric tubular epithelium and its expression is then restricted to the testis interstitium by E14.5 (293). Similarly, Hoxc5 and Hoxc6 have shown expression in the interstitial cells of the testis at E12.5 in mouse (294). This is supported in microarray data on developing cells of the mouse gonad, where expression of Hoxc5 and Hoxc6 is highest in the interstitial cells of each sex (Fig. 4.5E). Hoxc5 has also been shown to directly regulate Nr5a1 in the adrenal primordium via an adrenal specific enhancer in intron 4 of Nr5a1 (FAdE), yet there has been no evidence of this regulatory relationship existing in gonad (296). Interestingly, no Hox genes have yet been functionally implicated in gonadogenesis (reviewed in (297)), only their cofactors such as Pbx1 and Homeobox genes Lhx9 and Emx2. The identification of Hox genes controlling patterning of specific tissues is difficult to establish due to Hox genes being a large gene family with overlapping expression and functional redundancy, meaning single gene KO mouse studies do not display phenotypes. We hope that screening our differentiations for Hox gene expression may shed some light on their role in the human gonad.

136 A Lhx9 Emx2 Gata4 Female 8000 8000 1500 Male

6000 6000

s

s s

t t t 1000

n

n n

u

u u

o

o o

c 4000 c 4000 c

w

w w

a a a 500

R 2000 R 2000 R

0 0 0 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 1 1 1 1 Embryonic day Embryonic day Embryonic day

Nr0b1 Nr5a1 Gadd45g 1000 8000 2500

800 2000 6000

s s s

t t t

n 600 n n 1500

u u u

o o o

c c 4000 c

w 400 w w 1000

a a a R R 2000 R 200 500

0 0 0 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 1 1 1 1 Embryonic day Embryonic day Embryonic day

Wt1 Zfpm2 Hsd3b2 15000 3000 3000

s s s t 10000 t 2000 t 2000

n n n

u u u

o o o

c c c

w w w

a 5000 a 1000 a 1000

R R R

0 0 0 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 1 1 1 1 Embryonic day Embryonic day Embryonic day

137

B Sox9 Amh Fgf9 8000 6000 400

6000 4000 300

s

s

s

t

t

t

n

n

n

u

u

u

o

o

o

c

c 4000 2000 c 200

w

w

w

a

a

a

R

R 2000 0 R 100

0 -2000 0 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 1 1 1 1 Embryonic day Embryonic day Embryonic day

Star Hsd3b1

15000 500

400

s

s

t t 10000 300

n

n

u

u

o

o

c c 200

w

w

a a 5000 100

R

R 0

0 -100 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 Embryonic day Embryonic day

C Rspo1 Foxl2 6000 4000

3000

s s

t t 4000

n n

u u

o o

c c 2000

w

w

a 2000 a

R R 1000

0 0 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 Embryonic day Embryonic day

138 D Hoxd9 Hoxc6 Hoxc5 2500 2000 2000

2000 1500 1500

s s s

t t t

n 1500 n n

u u u

o o o

c c 1000 c 1000

w 1000 w w

a a a R R 500 R 500 500

0 0 0 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 .5 0 1 2 3 0 1 2 3 0 1 2 3 1 1 1 1 1 1 1 1 1 1 1 1 Embryonic day Embryonic day Embryonic day

Figure 4.5. Gonad marker expression profiles in the developing mouse testis and ovary. (A) Expression pattern of bipotential gonad markers in differentiating mouse gonads. (B) Expression pattern of testis markers in differentiating mouse gonads. (C) Expression pattern of ovarian markers in differentiating mouse gonads. (D) Expression pattern of HOX genes in differentiating mouse gonads. Graphs in A-D represent gene expression between stages E10.5-13.5 of mouse embryonic gonad development. Data is from bulk RNA-seq followed by EdgeR analysis (219). Data points represent the mean of the raw counts per million. Error bars show the standard deviation. (E) Graphs of log-transformed, normalised intensity values from microarray analysis on E11.5-13.5 embryonic mouse gonads. Colours represent different cell lineages, with XX shown as dashed lines and XY shown as solid lines. Error bars represent standard error. Data is from (220).

139 4.3.3 Immunofluorescence characterisation of gonad-specific markers in human fetal testis To more fully characterise the differentiated iPSCs I also sought evidence from gonad marker expression at the protein level. I therefore characterised antibodies by testing them on paraffin- embedded human fetal testis (9 week gestation) (Fig. 4.6). This highlighted some interesting differences between human and mouse fetal testes, for example I saw large groups of germ cells within the testis cords in the mouse testis. By contrast I saw mainly individual germ cells within the networks of testis cords in the human fetal testis. This may be in part due to difference in stage of testis development; the embryonic day 12.5 mouse testis is roughly equivalent to that of human testis at week 7-8.

Immunofluorescence staining with single-cell resolution confocal microscopy has not to our knowledge been performed on human testis as early as week 9. This allows us to assess key characteristics of cell staining unique to cell types. Counterstaining with DAPI also allowed characterisation of intracellular staining patterns for different markers. AMH (Fig. 4.6E, G) and STAR (Fig. 4.6c, d) showed cytoplasmic staining, OCT4 (Fig. 4.6E, F), SOX9 (Fig. 4.6U, V), GATA4 (Fig. 4.6s, t) and NR5A1 (Fig. 4.6k, l) showed nuclear staining, while WT1 (Fig. 4.6M, O) was expressed in both the nucleus and cytoplasm of cells. SOX9 (Fig. 4.6U, V) and AMH (Fig. 4.6E, G) are highly specific to Sertoli cells, while WT1 (Fig. 4.6M, O, c, e) is highly expressed in Sertoli but also expressed in interstitial cells. STAR (Fig. 4.6c, d) was specific to Leydig cells but also showed some background staining. Laminin (Fig. 4.6M, N), which is normally highly expressed in the epithelial basement membrane, showed widespread non-specific staining. NR5A1 (Fig. 4.6k, l) and GATA4 (Fig. 4.6s, t) show widespread expression in both Sertoli and interstitial cell populations.

140

141

142

Figure 4.6. Expression patterns of gonad markers in human fetal testis. Sections of parrafin- embedded human fetal testis (9 weeks) were stained for antibodies marking different testis cell populations. (A-H) Antibodies marking Sertoli cells (AMH, red) and germ cells (OCT4, green). (I- P) Antibodies marking Sertoli/interstitial cells (WT1, red) and basement membrane (Laminin, green). (Q-X) Antibodies marking germ cells (OCT4, red) and Sertoli cells (SOX9, green). (Y-f) Antibodies marking Sertoli/interstitial cells (WT1, red) and Leydig cells (STAR, green). (g-n) Antibodies marking Sertoli/interstitial/Leydig cells (NR5A1, red) and germ cells (OCT4, green). (o-v) Antibodies marking Sertoli/interstitial cells (GATA4, red) and germ cells (OCT4, green). DAPI (blue) stains cell nuclei. Scale bars in A-D represent 100 μM. Scale bars in E-H represent 10 μM. Scale bars in I-P, Q-T, Y-b, g-r represent 50 μM. Scale bars in U-X, c-f, s-v represent 20 μM.

4.3.4 Patterning axes of the embryonic mesoderm in iPSCs The bipotential gonad, consisting of gonad precursor cells (GPCs), is widely considered to arise from the IM at six weeks of human embryonic development or E10 in the mouse. The mesoderm itself forms from the elongation of the primitive streak between E8.0-8.5 in mouse or day 15 in

143 human (Fig. 4.7A) (298). The mesoderm gives rise to a wide host of tissues and these distinct cell populations are specified by signalling pathways, with gradients of factors being established along the various axes of the embryo. One example is the lateral-medial axis of the mesoderm, where NOGGIN/BMP4 antagonism helps to define the three regions of mesoderm: paraxial (high NOGGIN/low BMP4), intermediate (intermediate NOGGIN/BMP4), and lateral plate mesoderm (low NOGGIN/high BMP4) (Fig. 4.7B).

During the first 7 days of the kidney differentiation protocol mesoderm lineages are induced (188). I investigated patterning of two embryonic axes to establish which regions of the mesoderm have the best potential to give rise to GPCs; these were the anterior-posterior axis and the lateral-medial axis of the mesoderm.

Figure 4.7. Patterning of the anterior-posterior and lateral-medial axes of the fetal mouse mesoderm. (A) Canonical WNT signalling in the fetal mouse patterns cells migrating from the streak along the anterior-posterior axis of the IM between E8.0-8.5. Cells migrating early from the primitive streak are exposed to less WNT signals and thus become the anterior IM, while late migrating cells become the posterior IM. (B) Transverse section of an E8.5 mouse embryo shows morphogen gradients that pattern the mesoderm along the lateral-medial axis. E, embryonic day; SC, spinal cord; NC, notochord; PM, paraxial mesoderm, IM, intermediate mesoderm; LPM, lateral plate mesoderm. Figure adapted from (298, 299).

4.3.4.1 Testing the anterior-posterior axis of the intermediate mesoderm for gonad marker induction The IM is patterned along the anterior-posterior axis by cells that have been exposed to variable WNT signals as they migrate from the primitive streak (Fig. 4.7A) (298). Earlier migrating cells will be exposed to WNT signalling for a shorter amount of time, giving rise to the anterior IM.

144 Conversely, later migrating cells are exposed to more WNT and go on to become the posterior IM.

In the first few days of the differentiation protocol primitive streak-like cells are generated from iPSCs by induction of WNT signalling (using CHIR99021). By changing CHIR concentration and duration I can control whether these cells give rise to anterior or posterior IM. I varied CHIR treatment from 3, 4 or 5 days, followed by mesoderm induction (FGF9 treatment), and then assessed markers of the bipotential gonad after 7 days of differentiation (Fig. 4.8). A longer duration of CHIR gave higher induction of LHX9 and HSD3B2 (Fig. 4.8), two of the most specific bipotential gonad markers known. Most bipotential gonad markers and even Sertoli cell marker (SOX9) were induced as early as 7 days of differentiation (Fig. 4.8), suggesting that these mesoderm-like cells are capable of giving rise to gonad lineages. NR5A1 showed low induction and GADD45G was not expressed (Fig. 4.8), indicating that either these genes are expressed downstream of the other bipotential gonad markers or additional factors may be necessary. From these results I concluded that an intermediate duration of CHIR (4 days) is optimal, as we then get high level of expression across the most markers. This may suggest that the gonad arises from populations in the anterior and posterior IM.

To assess the robustness of these results I applied the same conditions to the female iPSC line, CRL1502.3, qRT-PCR results are shown in Appendix 10. Similarly, 4 days of CHIR treatment led to induction across bipotential gonad markers, except LHX9 and GADD45G. The absence of LHX9 expression may indicate that this replicates patterning that is too early in differentiation to see the expression of this gene. Alternatively it may be due to this iPSC line proliferating at a much slower rate that the previous line (half that of the PCS_201_010 line) which may affect the timing of stages during differentiation.

145

Figure 4.8. Longer CHIR duration promotes gonad marker induction. qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad after 7 days of monolayer differentiation with 3, 4 or 5 days of CHIR treatment. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in APEL media on Matrigel treated with 4 μM CHIR for 3, 4 or 5 days followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml) treatment.

I then assessed the optimal CHIR concentration required to induce the bipotential gonad. The kidney differentiation protocol (188) uses 8 μM CHIR in the first 3-4 days of differentiation. Preliminary experiments performed by Dr. Minoru Takasato (Riken, Japan) showed induction of gonad markers NR5A1, GATA4 and NR0B1 when hESCs (C32 line) were treated with lower concentrations of CHIR (4 or 5 μM) (see Appendix 11). I therefore differentiated male iPSCs (PCS_201_010) with 3, 4 or 5 μM CHIR for 4 days and assessed induction of gonad markers after 7 or 12 days of differentiation (Fig. 4.9). Initially 4 μM CHIR was chosen based on optimisation in APEL media (Appendices 12 and 13). However, APEL media stopped being produced therefore I re-optimised CHIR concentration in E6 media (Fig. 4.9). At day 7 I saw high induction of GATA4, EMX2 and HOXC5, as well as low-level induction of GADD45G, SOX9, STAR and HSD3B1 (Fig. 4.9). Interestingly LHX9 and WT1 are only expressed by day 7 in response to a higher concentration of CHIR (Fig. 4.9). By 12 days of differentiation these markers have increased further and I begin to see expression of LHX9 and GADD45G (Fig. 4.9). This is different to the results observed in APEL media differentiations and suggests that media components affect the efficiency or timing of differentiation. However, a lower concentration of CHIR (3 μM) was optimal for induction of EMX2, GADD45G, HSD3B1 and STAR, suggesting that there is a population of cells in the anterior

146 IM that are important for induction of gonad (and in particular Leydig) lineages. I also observed a lack of NR5A1 induction at day 12 (Fig. 4.9), even though this is an essential gene for gonad development (Table 4.2). Clearly, I have not yet induced a gonad gene expression profile in these cells. Expression of three Hox genes indicated that I am inducing HOXC5+ lineages specifically over HOXD9 or HOXC6 (Fig. 4.9), suggesting that this Hox gene may be more important in human gonad development. By day 12 I also saw evidence for the activation of the testis pathway in these cells, with a 20-50-fold increase in AMH expression and with 3 μM CHIR conditions I observed induction of Leydig cell markers (HSD3B1 and STAR) (Fig. 4.9). Further characterisation via immunofluorescence staining would be necessary to validate these changes in expression. Based on these results I used 3 μM CHIR for 4 days to induce gonad lineages in the PCS_201_010 iPSC line with E6 media.

147 LHX9 NR5A1 30000 iPSC

s s 4

C

C

S 3 uM CHIR

S

P

i 20000 P

i

o 4 uM CHIR

o

t

t

e 3

e

v 10000

i v 5 uM CHIR

i

t

t

a

l

a

l e NT2D1

e

r

r

n 2

20 n

o

i

o

i

s

s s 15

s

e

e

r

r

p

p 1 x 10

x

e

e

e

5 e

n

n

e

e

G

0 G 0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N

Day 7 Day 12 Day 7 Day 12

GADD45G WT1

s s 30 200

C

C

S

S

P

P

i

i

150

o

o

t

t

e

e

v

v

i 100

i

20 t

t

a

a

l

l

e

e

r

r

50

n n 10

o

o

i

i

s

s

s

s

10 e

e

r

r

p

p

x 5

x

e

e

e

e

n

n

e

e

G G 0 0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N Day 7 Day 12 Day 7 Day 12

GATA4 EMX2

s 4000 s 800

C

C

S

S

P

P

i

i

3000

o

o

t

t

600

e

e

v

v

i 2000

i

t

t

a

a

l

l

e

e

r

r

1000 400

n 500 n

o

o

i

i

s 400 s

s

s

e

e

r 300 r

p p 200

x

x

e 200 e

e

e

n 100 n

e

e

G 0 G 0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N Day 7 Day 12 Day 7 Day 12

SOX9 AMH

s 400 s

C C 2000

S

S

P 300 P 1000

i i

o o

t t

200

e e

v v

i i

t t 60

a 100 a

l l

e e

r r

n 10 n

o o

i i 40 s 8 s

s s

e

e r 6 r

p p

x x 20

e 4 e

e e

n 2 n

e

e

G 0 G 0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N Day 7 Day 12 Day 7 Day 12

148 HSD3B1 STAR

s 100 s 8

C

C

S

S

P

P

i

i

o 80 o

t

t

6

e

e

v

v

i

i

t

t

a

a

l 60 l

e

e

r

r

4

n

n

o

o

i 40 i

s

s

s

s

e

e

r

r

p p 2

x 20 x

e

e

e

e

n

n

e

e

G 0 G 0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N Day 7 Day 12 Day 7 Day 12

DHH FOXL2

s 4 s 1.5

C

C

S

S

P

P

i

i

o

o

t

t

3

e

e

v

v

i i 1.0

t

t

a

a

l

l

e

e

r

r

2

n

n

o

o

i

i

s

s

s s 0.5

e

e

r

r

p 1 p

x

x

e

e

e

e

n

n

e

e

G 0 G 0.0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N Day 7 Day 12 Day 7 Day 12

HOXD9 HOXC5

s s 2.5 2500

C C

S

S

P P

i

i

o o 2.0 2000

t

t

e e

v v

i

i

t

t

a

a

l 1.5 l 1500

e e

r r

n n

o o i 1.0 i 1000

s s

s s

e e

r r

p p

x 0.5 x 500

e e

e

e

n n

e e

G 0.0 G 0 C 1 C 1 M M M M M M D M M M M M M D S u u u u u u 2 S u u u u u u 2 iP 3 4 5 3 4 5 T iP 3 4 5 3 4 5 T N N Day 7 Day 12 Day 7 Day 12

HOXC6

s 5

C

S

P

i

o 4 iPSC

t

e

v 3 uM CHIR

i

t

a

l 3

e 4 uM CHIR

r

n

o 5 uM CHIR i 2

s

s

e NT2D1

r

p

x 1

e

e

n

e

G 0 C 1 M M M M M M D S u u u u u u 2 iP 3 4 5 3 4 5 T N Day 7 Day 12

Figure 4.9. 3 μM CHIR promotes induction of markers for bipotential gonad and testis. qRT- PCR data showing relative gene expression levels of markers of the gonad after 7 and 12 days of

149 monolayer differentiation with 3, 4 or 5 μM CHIR. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. The NT2D1 cell line was used as a positive control. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3, 4 or 5 μM followed by FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml) treatment.

4.3.4.2 Testing the lateral-medial axis of the mesoderm for gonad marker induction The mesoderm of the body axis can be divided into three regions along the lateral-medial axis: the paraxial mesoderm (PM), intermediate mesoderm (IM) and the lateral plate mesoderm (LPM) (Fig. 4.7B). The human embryonic kidney arises from the IM and in the protocol described in Takasato et al. (2015) iPSCs are differentiated towards IM lineages via activation of FGF signalling between days 4 and 7 of differentiation. As mentioned above, mouse studies have suggested an IM origin for the gonad (8, 276), while more recent chicken gonad lineage tracing showed that GPCs arise from the LPM (300).

To test where along this lateral-medial axis the induction of bipotential gonad markers is highest I simulated a morphogen gradient with BMP4/FGF9 signalling. Human recombinant BMP4 was added at two concentrations in addition to FGF9 and gene expression was assessed by qRT-PCR after 7 days of differentiation (Fig. 4.10). This highlighted an interesting pattern whereby shifting towards the LPM (marked by increased FOXF1 expression) saw upregulation of markers GADD45G, ZFPM2, GATA4, NR0B1, HSD3B2 and SOX9 (Fig. 4.10). While in IM-patterned cells (FGF9 only, marked by increased LHX1 expression) I observed upregulation of NR5A1, WT1 and EMX2 (Fig. 4.10). Interestingly, WT1 and EMX2 are also expressed in the fetal kidney (Fig. 4.4A), so by patterning for IM (FGF9 only treatment) I may be generating renal-like lineages. This pattern of marker gene expression was also observed when these differentiation conditions were applied to the female iPSC line (Appendix 14). High concentration of BMP4 (50 ng/ml) induced GADD45G, yet also appeared to inhibit the expression of two genes essential to the bipotential gonad, WT1 and EMX2 (Fig. 4.10). Based on these results I chose an intermediate concentration of BMP4 (10 ng/ml) for gonad differentiation. These data may be explained by populations of cells within both the IM and LPM regions contributing to the bipotential gonad. We note that the separation between these three regions is not distinct. Indeed, they represent a continuum even in the mouse with genes such as Osr1 expressed in both IM and LPM to different levels.

150 6000 iPSC FGF9

s

C

S 4000 FGF9 + 5 BMP4

P

i

o FGF9 + 50 BMP4

t

e

v

i

t 600

a

l

e

r 400

n

o

i

s 200

s

e

r

p

x 10

e

e 8

n e 6

G 4 2 0 1 1 9 1 G 1 4 2 1 2 2 H 9 X F X A 5 T A M B X B M X H X H 5 4 W T P 0 M 3 O L O L R D A F R E D A S F N D G Z N S A H G Lineage marker IM LPM Bipotential gonad Testis

Figure 4.10. Shifting differentiation towards the lateral plate mesoderm favours induction of gonadal lineages. qRT-PCR data showing relative gene expression levels of markers of the intermediate mesoderm (IM), lateral plate mesoderm (LPM) and gonad after 7 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in APEL2 media (with 4% PFHM-II) on Matrigel treated with 4 days of CHIR at 4 μM followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml), with or without addition of BMP4 (at 5 or 50 ng/ml).

4.3.5 Investigating the role of Hedgehog and Retinoic acid signalling in early gonad differentiation In subsequent optimisation of the gonad differentiation protocol I drew on knowledge of signalling pathways involved in development of the bipotential gonad and testis, as activation or repression of specific pathways may improve differentiation outcomes. Table 4.3 summarises some of the known enhancers and inhibitors of gonad development, from studies on mouse, rat, chick and human.

151 Table 4.3 Growth factors involved in gonad development.

Enhancers of bipotential gonad development

BMP4 (300)

FGF9 (301) (302)

IGF1R and INSR (50)

RA (303)

RSPO1 and WNT4 (47)

SHH (300)

Enhancers of testis development

8-Br-cAMP – Leydig cell differentiation (194)

Activin A – Leydig and testis cord formation, Sertoli cell proliferation (304) (305)

CNBP – Sertoli cell localisation and testis cord formation (306)

ESR1, ESR2 – Sertoli cell proliferation and differentiation (307)

FGF9 – Sertoli cell differentiation (305)

FGF2 and FGFR2 – Sertoli cell proliferation and testicular niche signalling (308) (301)

Forskolin – Leydig cell differentiation (194)

FSH – Sertoli cell differentiation (109)

GDNF – testicular niche signalling (308) (309)

INSR and IGF1R – Sertoli cell proliferation - (307)

Laminin and collagen – basement membrane formation (310)

LIF – testicular niche signalling (308)

PGD2 – Sertoli cell differentiation (311)

PI3K pathway – testis cord formation (312)

RA – Sertoli cell differentiation - (307)

TGF-Beta – Sertoli and Peritubular myoid cell differentiation (305, 313)

TR – Sertoli cell proliferation (307)

Inhibitors of testis development

152 RA (106, 314)

Canonical WNT signalling (306, 315)

BMP4: Bone morphogenetic protein 4; CNBP: Cellular nucleic acid-binding protein; ESR1, ESR2: Estrogen receptors 1 and 2; FGF2/9: Fibroblast growth factor 2/9; FGFR2: Fibroblast growth factor receptor 2; FSH: Follicle stimulating hormone; GDNF: Glial cell line-derived neurotrophic factor; IGF1R: Type I Insulin-like growth factor receptor; INSR: Insulin receptor; LIF: Leukemia inhibitory factor; PGD2: Prostaglandin D2; RA; Retinoic acid; RSPO1: R-spondin-1; SHH: Sonic Hedgehog; TR: Thyroid hormone receptor; WNT4: wingless-type MMTV integration site family member 4.

It is known that in both chick and mouse embryos, Sonic Hedgehog (Shh) is expressed more highly in the endoderm next to gonadal precursor cells than the area giving rise to the mesonephros (316, 317). Yoshino et al. (2016) showed that SHH contributes to patterning of the LPM along the dorsal-ventral axis, when Hedgehog signalling was inhibited no genital ridge formed at the site of the presumptive gonadal area. They also show that BMP4 signalling is activated downstream of SHH and this pathway is also essential for formation of the chick gonad. I have already shown evidence for the lateral plate mesoderm contributing to human gonadal precursor cells, however I next wanted to test whether Hedgehog pathway activation may also be important in the initiation of human gonadogenesis. To activate Hedgehog signalling during differentiation I provided the Smoothened agonist (SAG) alongside FGF9 from days 4-7 of differentiation when mesodermal lineages are being specified. After 7 days of differentiation I saw little change in expression of LHX9, NR5A1 and GATA4 but saw a reduction in markers WT1, ZFPM2, NR0B1, EMX2 and HSD3B2 (Fig. 4.11), with most of these showing a dosage-dependent negative response to SAG treatment. Furthermore, I have previously shown that GADD45G is upregulated in response to BMP signalling (Fig. 4.10), but as no expression is observed here (Fig. 4.11) it may mean that SAG treatment does not lead to necessary levels of BMP signalling in vitro. In summary, these results suggest that Hedgehog signalling is not essential in initiation of human gonadogenesis.

153

Figure 4.11. Hedgehog signalling is dispensable in induction of gonadal lineages. qRT-PCR data showing relative gene expression levels of markers of the gonad after 7 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in APEL2 media (with 4% PFHM-II) on Matrigel treated with 4 days of CHIR at 4 μM followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml), with or without addition of SAG (at 200, 400 or 1000 nM).

Retinoic acid (RA) signalling is important in segmentation of the mesoderm (318) and the outgrowth of the mouse gonad (303) (Table 4.3). I tested the effects of treating cells with RA (0.1 μM) during days 4-7 of differentiation (Fig. 4.12). After 7 days of differentiation I saw an enhanced induction of gonad markers WT1, GATA4, NR0B1 and HSD3B2 yet a reduction in markers NR5A1, EMX2, AMH and SOX9 (Fig. 4.12). WT1 and NR0B1 in particular show a dramatic increase in expression in response to RA (300-fold compared to 40-fold and 400-fold compared to 90-fold for WT1 and NR0B1 respectively) (Fig. 4.12), suggesting that RA regulates these two genes either directly or indirectly. Previous reports in zebrafish embryos identified a 299 bp enhancer region 4 kb upstream of wt1a that is required for wt1a expression in the IM (319). Interestingly, this enhancer contains a highly conserved RA response element and wt1a mesoderm expression is dependent on RA binding at this enhancer site. Here I show evidence that this transcriptional regulation is important during human mesoderm development also.

RA is also a known promoter of ovarian development, inducing transdifferentiation of Sertoli cells to granulosa-like cells in the absence of Dmrt1 (106). Hence, RA plays distinct roles at

154 different times during gonad differentiation. Fig. 4.12 supports this ovary-promoting role, with a decrease in expression of SOX9 and AMH compared to cells not exposed to RA. Taking this together with the high induction observed across the majority of gonad markers in the absence of RA, I did not use RA in subsequent differentiations.

Figure 4.12. Retinoic acid regulates WT1 and NR0B1 during early gonadogenesis. qRT-PCR data showing relative gene expression levels of markers of the gonad after 7 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in APEL2 media (with 4% PFHM-II) on Matrigel treated with 4 days of CHIR at 4 μM followed by FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml), with or without RA (0.1 μM).

So far I have shown optimisation of the first 7 days of differentiation to induce gonadal precursor cells, from this I have also learned about signalling pathways important in the early development of the human gonad. These experiments not only inform us on the direction to take differentiation but give us clues into signalling pathways that may be perturbed in human DSDs.

155 4.3.6 A note on iPSC differentiation media The experiments outlined here have been performed in three different differentiation media: APEL, APEL2 and E6. The published kidney organoid differentiation protocol was performed in APEL; therefore initial optimisation used this medium until it was no longer in production (2016). APEL2 medium was then used, however to generate results similar to that seen in APEL this required supplementation with 4% protein-free hybridioma (PFHM-II). The use of PFHM-II in APEL2 medium resulted in significant variation between differentiation experiments; in the case of kidney organoids this medium reduced the efficiency of the protocol significantly. I also observed changes in gene expression, for example LHX9 was no longer expressed as early as day 7 (see Fig. 4.9 compared Fig. 4.8). Appendix 15 shows a comparison in gonad marker induction between three differentiation media, APEL and APEL2 (made by Stem Cell Technologies) and APEL EN (same constituents as APEL but made in house by Dr. Elizabeth Ng), highlighting the wide variability in results. For this reason I recently switched to E6 medium, this is a more minimal and stable medium, made from a DMEM-F12 base with two extra additives (see Methods 4.2.4). With this medium I have seen some changes in timing of gonad marker expression, however I observe less variability between differentiations, therefore this has become the medium of choice for iPSC differentiation in this project.

4.3.7 Induction of bipotential gonad lineages from iPSCs During the first 7 days of differentiation I see upregulation of some key markers (EMX2, WT1, GATA4) but no expression of some of the most gonad-specific markers, LHX9 and NR5A1. I therefore wanted to see what would happen if I took these monolayer differentiations and continue culture. I used the optimised 7-day protocol and then continued FGF9 treatment from days 7-12, as I knew this is an inducer of testis development (Table 4.3). In Figure 4.13 I analysed cells after 12-15 days of culture via qRT-PCR and immunofluorescence analysis.

By 12-15 days of culture I observed LHX9 induction as well as an increase in GADD45G (Fig. 4.13B). For several markers (WT1, NR5A1, ZFPM2, EMX2 and SOX9) I saw that their expression peaks around day 7 and then reduces (Fig. 4.13B), I also observed this at the protein level, where WT1 and SOX9 expression was highest at day 7 (Fig. 4.13C). This is not unusual in the context of gonad differentiation, as in mouse RNA seq analysis (Fig. 4.5A,B) I saw that many of these markers peak early in gonad differentiation (around E11.5). However, given that these cells are

156 still mesoderm-like lineages at day 7 I would expect expression to be maximal sometime after day 7.

During mouse male gonad differentiation, bipotential gonad markers are upregulated early then decline as markers of the testis begin to be highly expressed (Figs. 4.5A, B). I did not observe high induction of either SOX9 or AMH after 12-15 days (Fig. 4.13B, Ci-iii); therefore I wondered whether FGF9 treatment from days 7-12 has an inhibitory effect on the testis pathway. Recent differentiation protocols for human pancreatic tissue showed how differentiating iPSC-derived cells retain fate stability after initial growth factor treatment followed by culture in a simple media without growth factors (Prof. Doug Melton, unpublished). This is based on the idea that iPSCs can be directed down a particular lineage and once the necessary regulatory circuits are in place these cells will self-direct their further maturation. To test this hypothesis I differentiated iPSCs in monolayers using the optimised 7-day differentiation protocol followed by culture with or without FGF9 treatment (Fig. 4.14).

157

Figure 4.13. Differentiation of gonad-like lineages from iPSCs. (A) Schematic of the differentiation protocol from iPSCs. (B) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 7-15 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. (C) Immunofluorescence staining of differentiated cells after 7-15 days of monolayer culture. Staining with primary antibodies for Sertoli cell markers SOX9 (red) and WT1 (green), DAPI (blue) stains cell nuclei. Scale bars represent 100 μM. PCS_201_010 iPSCs grown in APEL2 media (with 4% PFHM-II) on Matrigel treated with 4 days of CHIR at 4 μM, then 3 days of FGF9

158 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml), followed by 5 days with FGF9 (200 ng/ml) and Heparin (1 μg/ml).

In Fig. 4.14B we can observe how the expression profile of gonad markers changes over the course of differentiation. Induction of all bipotential gonad markers is higher in cells not treated with FGF9 after day 7 (Fig. 4.14B), suggesting that these cells can differentiate to bipotential gonad lineages without growth factor instruction after day 7. Markers GADD45G, GATA4, NR0B1 and SOX9 show peak expression between 7 and 12 days of differentiation (Fig. 4.14B), their expression profiles closely recapitulating that seen in the developing mouse gonad where peak expression occurs at E11.5 (Figs. 4.5A, B). Over this time point we also see intermediate and lateral plate mesoderm markers begin to decrease, LHX1 and FOXF1 respectively (Fig. 4.14B, no growth factor condition). This suggests that sometime between days 7-12 these cells have committed to a bipotential gonad fate and the testis pathway then becomes activated. This is further supported by an expression peak of AMH at day 15 (no growth factor condition) (Fig. 4.14B), which we know also peaks later in mouse testis E12.5 (Fig. 4.5B). We still lack induction of NR5A1 (Fig. 4.14B). This may be because NR5A1 is only expressed in a subset of cells meaning qRT-PCR would not pick up expression changes. Alternatively, NR5A1 may be expressed highly in iPSCs/mesoderm; this would mask relevant changes in its expression. Indeed, Figure 4.14C shows NR5A1 expression is high at day 3 of iPSC differentiation to human kidney organoids, suggesting that we may already have sufficient levels of NR5A1 in our cell populations.

Between days 15 and 18 I saw a second peak in EMX2 expression (no growth factor condition) (Fig. 4.14B). As this is not observed in mouse (Fig. 4.5) this may imply that another non-testis cell type is differentiating. Overall this shows us that FGF9 treatment after day 7 inhibits development into gonad/testis lineages. With no growth factors after day 7 cells develop towards a bipotential gonad fate between days 7 and 12. We then see some testis pathway activation with upregulation of Sertoli cell markers SOX9 and AMH, although their low expression levels suggest that only a subset of cells are committing to a testis fate. Furthermore, the absence of Leydig cell marker STAR (Fig. 4.14B) indicates that conditions can be optimised further to support testis differentiation. Importantly NR0B1 expression decreases markedly from day 7-12 (Fig. 4.14B), this is also observed in mouse testis (Fig. 4.5). This downregulation of NR0B1 at the onset of testis differentiation is important as NR0B1 overexpression in human testis development can result in dysgenetic testes and male to female sex-reversal (89). Based

159 on this I next investigated other signalling factors that may promote induction of multiple testis lineages.

160

161

162

Figure 4.14. Effects of FGF9 treatment on induction of testis lineages. (A) Schematic of the differentiation protocol from iPSCs. (B) qRT-PCR data showing relative gene expression levels of markers of pluripotency (OCT4), intermediate mesoderm (LHX1), lateral plate mesoderm (FOXF1), bipotential gonad (LHX9, NR5A1, GADD45G, WT1, GATA4, ZFPM2, NR0B1, EMX2, HSD3B2) and testis (SOX9, AMH, STAR) after 7-18 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in APEL2 media (with 4% PFHM-II) on Matrigel treated with 4 days of CHIR at 4 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml), followed by culture with or without FGF9 (200 ng/ml) and Heparin (1 μg/ml) to 18 days. (C)

163 Graph represents average expression of NR5A1 from days 0 to 18 of human kidney organoid differentiation from iPSCs. Data was generated from bulk RNA-seq analysis by Dr. Minoru Takasato (Riken, Japan).

4.3.8 Testis pathway activation in iPSC-derived cells I have shown that gonad precursor (or bipotential gonad) cell lineages are induced from iPSCs around day 10-12 of differentiation. The next aim was to identify what conditions are necessary to induce the testis pathway in these cells. For this we postulated that a set of testis growth factors could be added around the time these cells show a bipotential gonad profile (~day 10). Growth factors are expensive and we are do not know all of the factors necessary to support development of a testicular niche; therefore I initially investigated the use of a conditioned medium. Recent studies have successfully supported differentiation of germ cells by growing them in a conditioned medium generated from growth of primary testis cells (320). We do not have access to human testis tissue but a commonly used substitute is the human NT2D1 cell line. This is derived from a metastasis of a lung carcinoma yet it shows some similarity to testicular cells (278). I generated a conditioned medium from NT2D1 cells (Methods 4.2.4) and added this to differentiating cells from day 10 of culture (Fig. 4.15A).

NT2D1 conditioned medium had a positive effect on induction of markers WT1 and GATA4, however I saw a negative impact on induction of LHX9, NR5A1, EMX2, SOX9 and AMH (Fig. 4.15B). Despite its common use as a Sertoli-like cell line, I have shown that NT2D1 cells are not an accurate representation of testis cells as they lack expression of key markers SOX9, GADD45G and DHH (see Fig. 4.9). Together, this shows that NT2D1 conditioned medium cannot support differentiation to testis lineages.

164

165 Figure 4.15. NT2D1 condition media does not support induction of testis lineages. (A) Schematic of the differentiation protocol from iPSCs. (B) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 7-17 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 4 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml), followed by culture with or without NT2D1 conditioned medium from day 10.

166 For subsequent differentiations I tested different regulators of testis differentiation (see Table 4.3) for their ability to induce testis lineages. I focused on regulators of Sertoli cells, as this is the first cell type to differentiate in the mammalian testis and they direct differentiation of Leydig and other testis lineages (281). In the mammalian testis, Sry is the first testis specific gene to be expressed and it activates Sox9. Once Sox9 is expressed it maintains its expression via auto- regulation and upregulation from factors FGF9 (77) and Prostaglandin D2 (PGD2) (78) (311).

Furthermore, PGD2 and FGF9 can both sex reverse XX mouse gonads cultured ex vivo (305, 321), highlighting their potential as inducers of human testis lineages in vitro.

Combinations of FGF9 and PGD2 were applied to iPSC-derived cells at day 10 of differentiation

(Fig. 4.16A). Induction of Sertoli markers SOX9 and AMH was highest in response to PGD2 (at 500 ng/ml) (Fig. 4.16B); this was also reflected with immunofluorescence where SOX9 showed highest expression in response to a high dose of PGD2 (Fig. 4.16Dx-xii). This suggests that PGD2 acts as a positive inducer of testis lineages and leads to enhanced induction of Sertoli-like cells in monolayer culture. I saw some low level induction of Leydig marker STAR (Fig. 4.16B) with low dose PGD2, however by immunofluorescence I did not see evidence of Leydig cells (background staining only for HSD3B1) (Fig. 4.16Cvii). In bright field imaging branching networks of cells were observed, particularly in the E6 only and high dose PGD2 conditions (Fig. 4.16Ci, iv, see arrows). The immunofluorescence markers used here did not allow us to characterise these structures, they may represent vasculature or epithelial membranes.

Interestingly I saw again that FGF9 treatment has an inhibitory effect on gonad induction, with lower NR5A1, GADD45G and AMH seen in FGF9-treated cells compared to conditions with no growth factors (E6 only) (Fig. 4.16B), and this was not rescued even when FGF9 was used in combination with PGD2. This suggests that testis differentiation is sensitive to changes in FGF9 dosage, possibly similar to NR0B1 and NR5A1 genes, which have a dosage-dependent effect on gonad development (reviewed in (322)). For this reason I did not continue to use FGF9 treatment after day 7 of differentiation. However I cannot rule out a testis-inducing capacity if used at lower concentrations. From this I showed that high dose of PGD2 promotes differentiation of Sertoli-like lineages, likely due to its upregulation of SOX9, allowing the testis pathway to be more strongly activated.

167

168 s SOX9 AMH

s

C

C

S 15 S 150

P

P

i

i

o

o

t

t

100

e

e

v

v

i

i

t 10 t a 50

a

l

l

e

e

r

r

n n 15

o

o

i

i

s 5 s

s s 10

e

e

r

r

p

p

x

x 5

e

e

e 0 e n 0

n

e

e C ly 9 w h w h 1 C ly 9 w h w h 1

G S n F lo ig lo ig D G S n F o ig o ig D P o G h h 2 P o G l l 2 i F 2 9 T i F 2 h 9 h T 6 D 2 F 9 N 6 2 F 9 N E D F E D D F G G G G G P G /F F P G /F G P 2 / P 2 /F D 2 D 2 G D G D P G G P P P

FGF9 s STAR

s

C

C

S S 20 8

P

P

i

i

o

o

t

t

e

e

v 6 v 15

i

i

t

t

a

a

l

l

e

e

r

r

10 4

n

n

o

o

i

i

s

s

s

s

e

e

r 2 r 5

p

p

x

x

e

e

e e 0

0 n

n

e

e C ly 9 w h w h 1 C ly 9 w h w h 1

G G S n F o ig o ig D S n F o ig o ig D G l l 2 P o G l h l h 2 iP o F 2 h 9 h T i F 2 9 T 6 2 F 9 6 D 2 F 9 N E D D F N E D F G G G G G P G /F G P G /F P 2 /F P 2 /F D 2 D 2 G D G D G P G P P P

FOXL2

s

C

S 8

P

i

o

t

e

v 6

i

t

a

l

e

r 4

n

o

i

s

s

e

r 2

p

x

e

e 0

n

e C ly 9 w h w h 1 G S F o g o g D n G l i l i 2 iP o F 2 h 9 h T 6 D 2 F 9 N E G D G F P G F G P / /F 2 2 D D G G P P

169

170 Figure 4.16. Prostaglandin D2 induces testis-specific markers. (A) Schematic of the differentiation protocol from iPSCs. (B) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 18 days of monolayer differentiation. Each

171 sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. The NT2D1 cell line was used as a positive control. (C) Bright field images of differentiated cells after 18 days of monolayer culture. Arrows indicate branching network formation. Scale bars indicate 250 μM. (D) Immunofluorescence staining of differentiated cells after 18 days of monolayer culture. Staining with primary antibodies for Sertoli cell marker SOX9 (red) and Leydig cell marker HSD3B1 (green), DAPI (blue) stains cell nuclei. Scale bars in i and ii represent 200 μM, scale bars in iii-v, vii-viii, x-xi, xiii-xiv and xvi-xvii represent 100 μM, scale bars in vi, ix, xii, xv and xviii represent 50 μM. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). Cells were grown in E6 media from days 7-10 then treated with or without combinations of FGF9 (200 ng/ml + 1 μg/ml Heparin) and PGD2 (low conc. = 100 ng/ml, high conc. = 500 ng/ml).

I next postulated whether I could induce Sertoli and Leydig cells simultaneously by adding a known regulator in combination with PGD2. DHH, secreted from Sertoli cells, is a paracrine trigger for the differentiation of fetal Leydig cells (98), thus I added Hedgehog agonist, SAG, alongside PGD2 at day 10 of differentiation (Fig. 4.17A). Analysis of differentiated cells by qRT- PCR showed that expression of WT1 was low (Fig. 4.17B) and LHX9 lost completely (not shown) by day 12 (Fig. 4.17B). Markers of the bipotential gonad GADD45G and EMX2 were higher when cells were not treated with SAG (Fig. 4.17B). I saw a small increase in SOX9 induction with SAG treatment at days 12 and 18, however this did not correspond with an increase in AMH production (Fig. 4.17B and Figs. 4.17Cii, vi, x). Importantly I observed no increased expression of Leydig markers when Hedgehog signalling was induced using SAG, with STAR expression highest in the PGD2 only treatment (Fig. 4.17B). In immunofluorescence staining of day 18 differentiated cells I observed that the most abundant cell type co-stained for WT1 and SOX9 (Sertoli-like) (Fig. 4.17Ci, vi, xi), however WT1 appeared to be lowly expressed, as in qRT-PCR (Fig. 4.17B). I also observed a cell type that was positive for Leydig cell marker STAR (Fig. 4.17Ciii, viii, xiii). These were most frequent in conditions with PGD2 alone or with low concentration SAG. As these cultures were differentiated on plastic 24 well plates, I could not use confocal imaging at higher magnification to establish intracellular staining patterns. Furthermore, I consistently observed branching networks with AMH+ cells between then (Fig. 4.17Di, v, ix). These branching structures may be positive for basement membrane marker Laminin (Fig. 4.17Diii, vii), however this would need to be performed at higher resolution. Based on these results I showed that addition of SAG confers no major benefit to the induction of testis lineages.

172

173

174

Figure 4.17. Activation of Hedgehog signalling via SAG inhibits testis differentiation. (A) Schematic of the differentiation protocol from iPSCs. (B) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 7-18 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. The NT2D1 cell line was used as a positive control. (C) Immunofluorescence staining of differentiated cells after 18 days of monolayer culture. Staining with primary antibodies for Sertoli cell markers WT1 (red) and SOX9 (yellow), Leydig cell marker

175 STAR (green), DAPI (blue) stains cell nuclei. Scale bars in i-v represent 100 μM, scale bars in vi- xv represent 50 μM. (D) Immunofluorescence staining of differentiated cells with primary antibodies for Sertoli cell marker AMH (green) and basement membrane marker Laminin (red). Scale bars represent 100 μM. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml).

Cells were grown in E6 media from days 7-10 then treated PGD2 (500 ng/ml) with or without SAG (low conc. = 200 nM, high conc. = 1000 nM).

I next repeated the optimal conditions identified in Figure 4.16 and Figure 4.17 (high dose PGD2) but grew cells on chamber slides rather than plastic 24-well plates so that I could investigate cellular staining with higher resolution. Fig. 4.18A shows gene expression after 18 days of differentiation. At day 18 cells are still showing high expression of markers of the bipotential gonad LHX9, GADD45G, WT1, GATA4, ZFPM2 and EMX2 (Fig. 4.18A). The HOXC5 gene continues to be upregulated over HOXD9 and HOXC6 (Fig. 4.18A), suggesting that I may be inducing a population of testis interstitial cells. I did not see induction of NR5A1 (Fig. 4.18A), however at the protein l saw that NR5A1 expressed in a subset of cells (Fig. 4.8Dii). SOX9 expression peaks at day 7 and is lowly expressed by 18 days, (Fig. 4.18A) consistent with previous data showing that this gene’s expression peaks between days 7 and 12 (Fig. 4.14B). AMH shows a 6-fold increase in expression in the presence and absence of PGD2 (Fig. 4.18A), suggesting that PGD2 has little effect on Sertoli cell induction. In immunofluorescence staining of monolayers I saw co-staining of SOX9 and WT1 (Fig. 4.18Ciii, ix, iv, x), characteristic of pre-Sertoli cells. In the central region of the culture well I noticed that many of the SOX9/AMH positive cells have a smaller, circular appearance (Fig. 4.18Bv,vi), I think these cells may be in the process of cell death possibly due to high cell density in this region. To avoid this I could perform a cell split during differentiation or seed cells at lower density prior to differentiation. I do not see a significant change in gene expression of Leydig cell markers STAR and HSD3B1 despite Leydig cell-inhibitor NR0B1 being downregulated. STAR appeared to be present in a few individual cells in immunofluorescence staining (Fig. 4.18Cvi,vii), however closer analysis revealed that this is non- specific staining (Fig. 4.18Cv, viii, xi, xiv). I again observed formation of branching networks, which appear to be positive for vasculature marker PECAM (Fig. 4.18Diii). Analysis for markers of testis vascularisation (e.g. Vascular endothelial growth factor (VEGF) (323, 324)) by qRT-PCR could confirm this finding. Alternatively, differentiation could be performed in a vasculature reporter iPSC line, such as the Sox17-mCherry line (generated by Prof. Melissa Little).

176 Overall gene expression and immunofluorescence analysis shows that these cells are more bipotential gonad-like than testis. Screening for markers of closely related tissues adrenal (ARHGAP36 and SULT2A1), kidney (NPHS2 and PAX2) and cartilage (ACTG1 and ANXA2) did not suggest that these cells are differentiating to non-gonadal fates (see Appendix 16). Yet low expression of SOX9/AMH indicates that Sertoli-like cells are not being induced with high efficiency and these conditions are not supportive of testicular development. Future experiments will investigate other supplements that may be important in the embryonic testicular niche.

177

178

179

180

181 Figure 4.18. Prostaglandin D2 has minimal effect on testis pathway induction. (A) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 18 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. The NT2D1 cell line was used as a positive control. (B) Immunofluorescence staining of differentiated cells after 18 days of monolayer culture. Staining with primary antibodies for Sertoli cell markers SOX9 (green) and AMH (red), DAPI (blue) stains cell nuclei. Scale bars in i-iv represent 50 μM. Scale bars in v-viii represent 20 μM. (C) Immunofluorescence staining of differentiated cells with primary antibodies for Sertoli cell markers WT1 (red), SOX9 (yellow) and Leydig cell marker STAR (green). Red square indicates non-specific staining. Scale bars represent 50 μM. (D) Immunofluorescence staining of differentiated cells (PGD2 condition) with primary antibodies for Sertoli and interstitial cell markers GATA4 (red), NR5A1 (yellow) and vasculature marker PECAM (green). Scale bars represent 50 μM. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). Cells were grown in E6 media from days 7-10 then treated with or without PGD2 (500 ng/ml).

182 4.3.9 Differentiation in 3D One of the key aims of this project was to differentiate cells to multiple testis lineages but also to generate testis organoids by culturing these cells in 3D. Culturing cells in 3D may improve communication between neighbouring cells and therefore promote induction of different testis lineages, as well as the formation of testis-like structures. There are an increasing number of methods to generate organoids and I tested two. The first was air-liquid interface, where cells are aggregated at a defined cell number after 7 days of differentiation and cultured on Transwell filters.

In the first instance, I repeated the monolayer differentiation conditions from Fig. 4.18 and compared these to cells cultured as organoids from day 7 of differentiation (Fig. 4.19A). The development of these organoids was not supported in the presence or absence of PGD2, with necrotic cell death developing in the center of the organoids over time (Fig. 4.19Ciii, iv). This suggests that additional supplementation is required or that smaller organoids with greater exposure to media would be more optimal. In organoids treated with PGD2 or no growth factors I saw a small but significant increase in SOX9 (p=0.023 and p=0.014 respectively) (Fig. 4.19C) compared to monolayer cultures. Surprisingly this did not translate to an increase in AMH expression in organoids (Fig. 4.19C), suggesting that this increase in SOX9 expression is not indicative of a functional change (testis pathway activation) in these cells. This is further reinforced by an absence of bipotential gonad/testis markers NR5A1, WT1 and FGF9 in organoids (Fig. 4.19B, C).

183 B LHX9 NR5A1

s

s

C

C

S S 25 8

P

P

i

i

o

o

t

t

20

e

e

v

v 6

i

i

t

t

a

a

l l 15

e

e

r

r

4

n

n

o

o

i i 10

s

s

s

s

e

e

r r 2

p p 5

x

x

e

e

e e 0 0

n

n

e

e C 7 6 2 6 2 C 7 6 2 6 2

G G S y E D E D S y E D E D a r a r iP e G id G iP e G id G d y P o P d y P o P a r n d a r n d l e a i l e a i o y g o o y g o n la r n n la r n o o o a o o o a m n 8 rg m n 8 rg o 1 o 1 o 8 d o 8 d 1 m 8 1 m 8 d 8 1 d 8 1 1 d 1 d d d

GADD45G WT1

s s

C

C

S 10 S 150

P P

i i

o o

t

t

8 100

e

e

v v

i i

t t 50

a a l 6 l

e e

r

r

n n 10

o o

i 4 i

s

s

s

s

e e

r r 5

p 2 p

x x

e

e

e 0 e 0

n

n

e e C 7 6 2 6 2 C 7 6 2 6 2 G S y E D E D G S y E D E D a r a r iP e G id G iP e G id G d y P o P d y P o P a r n d a r n d l e a i l e a i o y g o o y g o n la r n n la r n o o o a o o o a m n 8 rg m n 8 rg o 1 o o 1 o 8 d 8 d 1 m 8 1 m 8 d 8 1 d 8 1 1 d 1 d d d

GATA4 EMX2

s s

C C

S 4000 S 5000

P P

i i

o o

t t 4000

e e

v 3000 v

i i

t t

a a

l l 3000

e e

r r 2000 n n 100

o o

i i

s s

s s

e e

r 1000 r 50

p p

x x

e e

e 0 e 0

n n

e e C 7 6 2 6 2 C 7 6 2 6 2 G S y E D E D G S y E D E D a r a r iP e G id G iP e G id G d y P o P d y P o P a r n d a r n d l e a i l e a i o y g o o y g o n la r n n la r n o o o a o o o a m n 8 rg m n 8 rg o 1 o o 1 o 8 d 8 d 1 m 8 1 m 8 d 8 1 d 8 1 1 d 1 d d d

184 C SOX9 AMH

s

s

p = 0.023 C

C

S S 6 4

P

P

i

i

p = 0.014

o

o

t

t

e

e

v 3

v

i

i

t

t

4 a

a

l

l

e

e

r

r

2

n

n

o

o

i

i

s

s

2 s

s

e

e

r 1

r

p

p

x

x

e

e

e e 0 0 n

n

e e C 7 6 2 6 2 C 7 6 2 6 2 G E E G S y E D E D S y D D a r P a r G d G iP d e G id G i d e P i P y P o P y r o a r n d la e n id l e a i o y a o o y g o n a rg n n la r n o l o a o o o a o g m n 8 rg m n 8 r o 1 o 8 o 1 o 8 d 1 m d 1 m 8 d 8 d 8 1 8 1 1 d 1 d d d

FGF9 HSD3B1

s

s

C

C

S 20 S 150

P P

i i

o o

t

t

e e

v 15 v

i

i t t 100

a a

l

l

e e

r

r 10

n n

o o

i i

s s 50

s s

e e

r 5 r

p p

x

x

e e

e 0 e 0

n n

e e C 7 6 2 6 2 C 7 6 2 6 2 G S y E D E D G S y E D E D a r a r iP e G id G iP e G id G d y P o P d y P o P a r n d a r n d l e a i l e a i o y g o o y g o n la r n n la r n o o o a o o o a m n 8 rg m n 8 rg o 1 o o 1 o 8 d 8 d 1 m 8 1 m 8 d 8 1 d 8 1 1 d 1 d d d

STAR FOXL2

s

s

C

C

S 8 S 1.5

P

P

i

i

o

o

t

t

e

e

v 6

v

i

i

t t 1.0

a

a

l

l

e

e

r

r

4

n

n

o

o

i

i

s s 0.5

s

s

e

e

r

2 r

p

p

x

x

e

e

e 0 e 0.0

n

n

e

e C 7 6 2 6 2 C 7 6 2 6

G S y E D E D G S y E D E a r a r iP e G id G iP e G id d y P o P d y P o a r n d a r n l e a i l e a o y g o o y g n la r n n la r o o o a o o o m n 8 rg m n 8 o 1 o o 1 8 d 8 d 1 m 8 1 m d 8 1 d 8 1 d 1 d d

185

Figure 4.19. Air-liquid interface organoid culture does not improve induction of testis markers. (A) Schematic of the differentiation protocol from iPSCs. (B, C) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 18 days of monolayer or organoid differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. An unpaired t-test with Welch’s correction was applied to obtain p-values. (D) Brightfield images of differentiated cells after 18 days in monolayer or organoid culture. Error bars indicate 250 μM. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). At day 7 half of the cells were dissociated and reaggregated into 200,000 cell organoids then cultured on Transwell filters. Cells were grown in E6 media from days 7-10 then treated with or without PGD2 (500 ng/ml).

A more recent organoid culture system is the swirler approach ((325); Kumar et al. 2018, unpublished), where cells are dissociated to clusters (EDTA treatment) rather than single cells (trypsin treatment) and incubated in a constantly swirling media so they self-form organoids. This results in a greater number of organoids that are smaller in size, meaning they have better exposure to media, O2 and other nutrients. Using this approach, organoids were created at day

10 of differentiation and combinations of growth factors PGD2, FGF9 and GDNF were applied

(Fig. 4.20A). PGD2 has shown some evidence for inducing Sertoli cell markers (Fig. 4.16-18), FGF9

186 was used at a lower concentration to aid in proliferation of organoids (Dr. Santhosh Kumar, personal communication, Murdoch Children’s Research Institute) and GDNF is known to be involved in maturation of testicular cell types (308, 309) so was applied from day 13 of differentiation (Fig. 4.20A).

In qRT-PCR analysis I saw that organoids cultured without growth factors show higher expression of bipotential gonad markers GATA4, EMX2, GADD45G and LHX9 (Fig. 4.20B), indicating that these organoids are less mature. The Sertoli cell marker SOX9 showed a similar level of induction across all treatments (Fig. 4.20B). This was also evident in immunofluorescence where it showed widespread expression (Fig. 4.20Dii, vi, x). Co-staining with an additional Sertoli cell marker, WT1 (Fig. 4.20Di, v, ix), failed to reinforce their Sertoli cell fate, as this antibody did not show strong nuclear expression, as previously shown in Sertoli lineages (Fig. 4.6M, O). I observed a 20-fold induction of AMH in growth factor-free and PGD2/FGF9 conditions and this increased to a 60- fold induction when GDNF was also included (Fig. 4.20B). This suggests that GDNF promotes maturation of Sertoli cells from the pre-Sertoli stage. Assessment of gene expression patterns before and after GDNF treatment (day 13) could help determine the role of this growth factor. Co-staining of organoids with Laminin/AMH (Fig. 4.20E) revealed that organoids cultured with growth factors (Conditions 2 and 3) formed more complex tubule-like structures (Fig. 4.20E,v- vii), with a Laminin+ layer of cells partitioning clusters of AMH+ cells. Overall, this technique shows promise for the generation of testis organoids and has the benefit of allowing much higher throughput analysis than the air liquid interface method. Future experiments will aim to optimise these conditions further and look at the effects of GDNF on maturation of Sertoli-like cells.

187

188

189

Figure 4.20 Swirler culture organoids show testicular characteristics after 17 days of differentiation. (A) Schematic of the differentiation protocol from iPSCs. (B) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 17 days of organoid differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. (C) Bright field images of differentiated cells after 17 days in organoid culture. Scale bars indicate 250 μM. (D) Immunofluorescence staining of day

190 18 differentiated cells with primary antibodies for Sertoli cell markers WT1 (green), SOX9 (red), DAPI (blue) stains cell nuclei. Scale bars represent 50 μM. (E) Immunofluorescence with primary antibodies for Sertoli cell marker AMH (green) and basement membrane marker Laminin (red). Scale bars in i-vii represent 20 μM. Scale bars in ix-xii represent 50 μM. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). Cells were grown in E6 media from days 7-10 then dissociated with EDTA to form organoids. Organoids were cultured in swirler media with or without PGD2 (500 ng/ml), FGF9 (100 ng/ml), Heparin (1 μg/ml) and GDNF (10 ng/ml).

4.3.10 Differentiation in a SOX9 reporter iPSC line Here I have shown evidence for the differentiation of Sertoli-like cells from iPSCs and these appear to be one of the most abundant cell populations differentiating (Fig. 4.18C, 4.20D). To further characterise this cell lineage a reporter iPSC line specific to Sertoli cell markers will be used. A group at MCRI (Yudha Patria, Assoc. Prof. Shireen Lamande and Prof. John Bateman) generated a SOX9 reporter iPSC (see Methods 4.2.9), as SOX9 is also an important marker in their cell type of interest, chondrocytes. This reporter line contains a TdTomato flag within the SOX9 locus (exon 3) that does not disrupt SOX9 activity but drives Tomato fluorescence with as low as a 5-fold induction of SOX9.

To see whether this iPSC reporter could be used in characterisation of testis lineages I applied our differentiation protocol to this line. I saw highly similar patterns of gonad marker induction (Fig. 4.21A compared to Fig. 4.9, 3 μM condition) in this line, except for GADD45G, which is still only induced at low level in the SOX9 reporter line. By day 12 there is a 20-fold induction of SOX9 (Fig. 4.21A) and this is reflected by high Tomato fluorescence across the majority of cells in monolayer (Fig. 4.21B). FACs analysis will be used in future to gauge what proportion of these cells are differentiating towards SOX9+ Sertoli-like lineages. This indicates that this is a viable iPSC line for differentiating testis cells and will be used for characterisation of Sertoli-like lineages.

191

Figure 4.21 Differentiation to gonad-like lineages in a SOX9 reporter iPSC line. (A) qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad and testis after 7-12 days of organoid differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. The NT2D1 cell line was used as a positive control. (B) Immunofluorescence imaging of day 7-12 differentiated cells, TdTomato fluorescence (red) marks SOX9-positive cells, DAPI (blue) stains cell nuclei. Scale bars represent 50 μM. SOX9-TdTT-CRE#13 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). Cells were grown in E6 media from days 7-12. This experiment was performed in collaboration with Gorjana Robevska.

192 4.4 Discussion The development of iPSC models for various human tissues has been fundamental for making patient-specific disease models. This development is not yet applicable to the field of DSD, as no one has yet differentiated human gonad lineages from iPSCs. With the advent of fast and affordable MPS, we now have improved ability to identify candidate variants in undiagnosed DSD patients but these variants need to be functionally validated. At present, researchers test variant pathogenicity in sub-optimal cell lines and mouse models that are unable to recapitulate human gonad development (as demonstrated in Chapter 3). In this chapter I have described the development of a novel protocol for the induction of gonad lineages from human iPS cells, with the aim being to generate an in vitro model for DSDs.

To induce gonad lineages from iPSCs I drew initially from a protocol for the differentiation of human kidney tissue and developed this further based on current knowledge of gonad development. Differentiated gonad-like cells were characterised by analysis of gene expression and immunofluorescence staining, this has not only guided the protocol but also given us insight into the regulation of human gonadogenesis.

This was unique from previous Sertoli cell induction protocols (discussed in Chapter 1) in that I took a step-wise directed differentiation approach. Using this system I was able to more closely model gonad development in vitro and can dissect out how genes are regulated during this process. For example, I showed evidence that WT1 is regulated by Retinoic acid during early mesoderm differentiation, previously shown only in zebrafish development (319). I also showed that the human gonad may arise, at least in part, from cell populations in the intermediate (highly expressing NR5A1, WT1, EMX2) and lateral plate (highly expressing GADD45G, ZFPM2, GATA4, NR0B1, HSD3B2) mesoderm (Fig. 4.10).

This step-wise approach also allowed us to study the intermediate populations, the bipotential gonad, pre-Sertoli cell and Sertoli cell. For example, I saw that differentiating cells transition from bipotential gonad to pre-Sertoli cell between days 7-12 of differentiation (Fig. 4.14B). This was shown by temporal changes in gene expression of key markers (GATA4, GADD45G, SOX9 and AMH), closely recapitulating that seen in the developing mouse gonad (Fig. 4.5A,B). This is

193 important in the context of DSD as we have phenotypes where different stages of gonad development are affected. For example streak gonads result from the failure of bipotential gonads to develop, while 46,XY gonadal dysgenesis results from abnormal Sertoli cell differentiation.

The optimised differentiation protocol I have arrived at is shown in Figure 4.22. This differentiation protocol is still a work in progress. Ambitiously I aimed to generate multiple testis lineages that would organise into testis-like structures in vitro. Part of this aim has been satisfied, I can induce bipotential gonad lineages after 10 days of culture (Fig. 4.22), followed by induction of Sertoli-like cells at days 12-15 using a relatively minimal growth factor mix. I showed here that the highest induction of Sertoli markers in E6 media was at day 12 of differentiation (see Fig. 4.15, E6 only condition). This suggests that between days 12-18 conditions become sub-optimal for testis differentiation, highlighted in the cell death observed in Fig. 4.18Bv, vi. This is likely due to high cell density in monolayers; future optimisation may include plating cells at lower cell density or, preferably, moving to the swirler organoid culture system.

194

Figure 4.22. Schematic of the differentiation protocol for generating male gonad lineages from human iPSCs. This protocol is based on a directed differentiation approach with sequential changing of growth factor conditions. Human iPSCs are prepared for differentiation in culture with E8 medium. Differentiation starts in E6 medium supplemented with CHIR 99021 (3 μM), followed by E6 supplemented with FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). All growth factors are withdrawn between days 7-10, at which stage bipotential gonad lineages are induced. Cells are then cultured in E6 medium supplemented with PGD2 (500 ng/ml), Sertoli cell induction occurs from day 12-15.

I also aimed to generate multiple testis lineages, yet even in experiments with high induction of Sertoli markers (Fig. 4.15, E6 only condition) I have seen minimal expression of Leydig markers. Fetal Leydig (and adrenal) precursors become positive for NR5A1 before differentiating towards a Leydig cell fate. Here I have consistently shown low induction of NR5A1, the monitoring of this early marker could be used to guide Leydig cell induction. To induce multiple testis lineages we will investigate several alterations. Firstly, we could improve the immediate monolayer culture environment by media supplementation (e.g. glucose or glucocorticoids) and differentiating iPSCs on a more supportive extracellular matrix such as Laminin. Secondly, we could test addition of growth factors known to enhance non-Sertoli testis cell populations (e.g. Activin A and Forskolin, Table 4.3). Similarly, we know that maintenance of the testicular pathway relies on active suppression of the ovarian pathway; we could therefore test the effects of repressing

195 female signalling pathways like canonical WNT signalling. Finally, 3D cell culture would enable better interaction of Sertoli-like cells with other cell types and we know that Sertoli cell signalling directs differentiation of other testis cell types, e.g. Leydig cells via paracrine signal DHH (98). This would also be necessary to develop a more useful model for DSD, so we can establish how DSD variants affect gonad organisation in vitro.

Furthermore, we would like the ability to perform functional analysis with this model. Two of the key functions of the fetal testis (often disrupted in DSDs) are the support and maturation of germ cells (by Sertoli cells) and the production of Testosterone (by Leydig cells). We would like to optimise the testis differentiation protocol in the SOX9-TdTT-CRE#13 iPSC line so that we can further characterise the function of the SOX9+ Sertoli-like cells. We would dissociate our monolayer cultures (at ~day 10-12), then reaggregate with mouse germ cells (isolated from Oct4-GFP tagged mice) or human germ-like cells (differentiated from iPSCs using recently developed protocols (203, 207)), culturing these over an extended period as swirler organoids. This way we could establish if our iPSC-derived Sertoli-like cells can support germ cell lineages and form testis cord-like structures. A testosterone concentration assay could be used if we can successfully differentiate Leydig-like cells in future.

To develop a more complete model of DSD, this system would benefit from an ovarian differentiation protocol being established and characterised in future. This model could then be used to validate variants from a wider range of DSD phenotypes. I have shown here that I can reach a bipotential gonad expression profile in both male and female iPSC lines. This sets up the starting point for differentiation to ovarian lineages.

Although not presented in this thesis, our current differentiation protocol has been applied to an iPSC line harbouring a novel DSD variant. We identified a heterozygous mutation in the SART3 (Squamous cell carcinoma antigen recognized by T-cells 3) gene in an individual with 46,XY gonadal dysgenesis and intellectual disability (Ayers et al., unpublished). This is a novel DSD gene with an unknown role in the developing gonad; therefore the patient variant was introduced into the PCS_201_010 iPSC line using CRISPR. I saw a significant difference in induction of gonad markers (including NR5A1, WT1, EMX2 and AMH) compared to the isogenic parental iPSC line, showing that this protocol already has promise for validating DSD variants. We now need to

196 establish patient-derived iPSCs and gene-corrected isogenic controls to confirm these findings in the context of the patient’s genetic background. However, it is important to note that this model would not represent a high-throughput approach to variant validation given the cost and time required to make a variant iPSC line at present. The DSD field also ultimately needs an accurate and high throughput approach to functional analysis of gene variants.

Beyond the DSD field, this system could also be of benefit for germ cell/fertility research. Until recently, human germ cell differentiation from iPSCs was halted at the primordial germ cell (PGC) stage (206, 326). Maturation into germ cells failed because no one could recapitulate the global epigenetic remodeling that occurs as PGCs migrate from the hindgut to the presumptive gonad. This was recently overcome by the generation of xenogenic reconstituted ovaries, where human iPSC-derived PGCs were co-cultured with mouse ovarian somatic cells (207). This led to the differentiation of human oogonia-like cells within putative follicles. Notably, these oogonia- like cells exhibit the epigenetic reprogramming characteristic of in vitro oogenesis. Similar studies performed with mouse testicular somatic cells indicate that this process can also give rise to human iPSC-derived spermatogonia (Saitou, unpublished). At present this protocol is unsuitable for therapeutics given its use of animal cells and inefficiency (70 day culture protocol). Similarly, Shlush et al. (2017) recently generated adult Sertoli-like and haploid spermatid-like cells from human MSCs via growth factor treatment and co-culture with mouse Sertoli cells, simulating a testicular niche environment (203). Drawing on these findings, it appears that a complex multicellular structure is required to support the differentiation of human germ cells and the formation of gonadal-like structures in vitro. Furthermore, based on these results it may be possible to co-culture iPSC-derived PGC-like cells (206, 326) with our iPSC-derived testicular somatic cells and recapitulate these results in a xeno-free system. Collaborating with this field will help us to build a more complete model of the human gonad in future.

In summary, I have shown that I can generate bipotential gonad- and Sertoli-like lineages from iPSCs using a directed differentiation approach. This represents a significant step towards the generation of human testis tissue from pluripotent stem cells. Further refinement of this model would be beneficial if this is to become a testing tool for DSD variants, including the induction of additional testis lineages and characterisation of testis-like structures when cultured in 3D.

197

198 Chapter 5: Conclusions and Future Directions

DSDs affect 1.7% of live births (1), often presenting complex clinical scenarios requiring a diverse team of healthcare professionals. Reaching a genetic diagnosis in these cases is beneficial from both a medical management and psychosocial perspective. Prof. Andrew Sinclair has established one of the largest DSD cohorts internationally; by analysing these patients using MPS we have been able to yield a diagnostic rate of ~38% (2). The overarching aim of my PhD project was to improve our current diagnostic pipeline for DSD. I did this by focusing on two key problems that we face in understanding the genetic basis of DSD. The first is the low diagnostic rate in one subgroup of DSD, 46,XX (ovo)testicular DSD, where just 18% of patients receive a genetic diagnosis (2) (all 18% were SRY translocations). To address this, I studied a cohort of 34 individuals with SRY-negative 46,XX (ovo)testicular (or closely related) DSDs using varied genetic analyses and found diagnoses in nine patients (Chapter 2). In cases where diagnoses were not made, I identified candidate variants and validated these in cellular or animal models (Chapters 2 and 3). The second problem we encounter is the lack of a suitable in vitro model in which to study DSD genes. Advances in MPS mean that we have an increasing number of candidate DSD genes, yet we rely on sub-optimal cell lines and mouse models, neither of which can accurately model human gonad development. To advance this area of DSD research, I developed a protocol to differentiate human iPSCs to fetal gonad lineages (Chapter 4).

5.1 Approaching diagnoses in a 46,XX DSD cohort The first aim of this project was to identify molecular diagnoses in a cohort of 34 patients with 46,XX DSD. To do this I initially applied two genetic analyses to the entire cohort, MLPA and the targeted DSD screen (Chapter 2). I then performed a pilot study to test the efficacy of WGS in five cases from the cohort (Chapter 2).

Following the MLPA analysis, I identified a molecular diagnosis in five patients; these were duplications in the enhancer region of SOX9. This highlighted the high contribution of CNVs to these phenotypes, namely that SOX9 enhancer duplications contribute to approximately 21%

199 (5/24 cases) of SRY-negative 46,XX (ovo)testicular DSDs. The key limitation of this technology however was the limited regional information it provided. To get meaningful information (e.g. exact genomic coordinates) on these findings we had to perform follow-up CGH arrays. The CGH arrays available at the start of this project did not have sufficient coverage around this SOX9 regulatory region, which covers 2 Mb and lies in a gene desert; therefore the MLPA analysis was effective for screening a large number of samples for CNVs in regions of interest. Subsequent development of a DSD targeted CGH array will replace the MLPA for CNV analysis in our lab.

Following this, the targeted DSD screen identified a further three diagnoses and one likely diagnosis in the NR5A1 gene (Chapter 3, discussed in section 5.3). In the remaining patients, the vast number of variants of unknown significance provided a platform to explore candidate genes. Thus, the second aim of this 46,XX DSD study was to characterise gene variants functionally and identify the role they play in patient phenotypes. Each of the two candidate gene variants that have been functionally validated (LGR5 and RXFP2) were shown to be benign variants, while three further genes (EMX2, FOXL2 and WNT9A) are still being investigated. These results suggest that my process of identifying likely pathogenic variants can be improved. Tellingly, one of the key similarities of these likely benign variants (EMX2, LGR5 and RXFP2) was that they were all from singleton patients. If we had parental DNA and established that these were maternally or de novo inherited variants this would have significantly altered our weighting on their pathogenicity. Furthermore, there has recently been an expansion in the number and accuracy of tools that can be used to consider variant pathogenicity in silico before going into the wet lab. One example is the gnomAD database, which now provides constraint scores for genes based on large population studies as well as a break down of homozygous and heterozygous states of variants in XX versus XY individuals. Furthermore, using tools like the GTEX database can allow us to look at splicing and transcript-specific expression in our human tissue of interest. Future analysis of MPS data in our lab will use the seqr platform (Broad Institute), which incorporates each of these variant curation tools in one program. In addition, our focus for follow up variants will be on those with highly damaging protein effects and in individuals where inheritance pattern can be established.

I also applied WGS to a subset of the 46,XX (ovo)testicular DSD cohort. The rationale behind this approach was that it would enable us to identify CNVs, structural rearrangements and SNVs simultaneously. This study generated a large number of candidate variants, such as a duplication

200 in a potential regulatory region for the ovarian WNT4 gene. However the filtering and curation of these variants proved challenging without having parental data. Follow-up exome sequencing of available parental samples has helped to narrow down this candidate list. Despite this, the study has created a valuable data resource that can be re-analysed over time as new genes and bioinformatic techniques come to light.

The aim for this cohort, identifying molecular diagnoses, was attained for 9 of 34 patients (26%). This is a vast improvement on our original study of this cohort (2) where we found no diagnoses in SRY-negative 46,XX (ovo)testicular DSD patients. The consistently lower diagnostic rate observed for these DSD phenotypes lies in a number of factors, including fewer known diagnostic genes for these phenotypes and the high contribution of CNVs, which are not picked up by many MPS strategies. The other limiting factor here was the large number of singleton cases, which makes the identification of strong candidate variants challenging. An offshoot of this aim however, was how this process has informed us on the most suitable genetic testing approach for these phenotypes (discussed in 5.2). We have also now established a large database of MPS data for future research. The second aim, to functionally validate candidate 46,XX DSD gene variants, was also achieved. I was able to prove or disprove the pathogenicity of five variants (EMX2, LGR5, NR5A1 p.Arg92Trp & p.Ala260Val, RXFP2) using a range of cellular and animal modelling approaches specific to each gene function.

To date, there are no published cohort-wide studies investigating the genetic basis of SRY- negative 46,XX (ovo)testicular DSDs. Cohorts of up to 69 patients (143) exist, yet these studies have been presented in the context of diagnostic or novel gene findings in several patients, such as NR5A1 (146-148, 255, 256) or NR2F2 (143) respectively. This study therefore represents a novel contribution to this field and shows that combining technologies can result in a diagnostic rate of 26%.

5.2 Future directions for undiagnosed and new cases The outcomes of each genetic analysis used in this cohort study have been extremely valuable, informing the best screening approach for 46,XX (ovo)testicular DSDs. We saw that the targeted DSD gene panel has much lower diagnostic yield in these 46,XX DSD phenotypes (26%) than

201 other phenotypes, e.g. 46,XY Androgen insensitivity syndrome (60%) (2). With the lowered cost of MPS, many clinical genetics services have ceased using gene panels, replacing these with clinical exomes. For example, the Victorian Clinical Genetics Services now primarily use whole exome capture. Their initial analysis is phenotype-driven, searching a gene list derived from the patient’s specific phenotype. They then search the wider ‘Mendeliome’ (Online Mendelian Inheritance in Man disease-associated genes) while filtering out the ‘Incidentalome’ (327); if this too is negative they can explore the whole exome data for novel genes.

During this study we observed that CNVs appear to contribute to these phenotypes at a higher rate than other DSDs. Therefore ideally we need an approach that can capture both CNVs and SNVs simultaneously. For new cases received in future, we will likely run these on the DSD targeted CGH array, followed by WES if findings are negative from the array. Each of these techniques require small amounts of DNA and allow us to examine all of the known diagnostic genes/regions, while also generating data that can be mined for candidates in the absence of a diagnosis.

Future directions for the 22 undiagnosed 46,XX (ovo)testicular DSD patients will include application of the DSD targeted CGH array. This has been performed on just 23% (8/34) of cases so far. Furthermore, a recent collaboration with the Broad Institute means that many of these undiagnosed cases will undergo WGS. We are wary of the difficulty of analysing singleton whole genomes; therefore it will be useful to get parental samples sequenced as well (at least to exome coverage). Additionally, having a larger cohort of WGS datasets for these rare phenotypes may help analysis. The real value here will be in collaborating with other DSD researchers who have similar cohorts and sharing candidate genes in DSD databases to aid in the identification of novel genes.

As I saw during variant validation (Chapter 2 and discussed in 5.2), many of the in silico tools cannot accurately predict pathogenicity. For example, assessment of evolutionary sequence conservation often does little to support or disprove pathogenicity given that sex determination is so diverse among vertebrates. Functional data can provide strong evidence to implicate a variant as pathogenic or benign. Until now, we have applied functional assays to each VUS individually, often spending a long time optimising assays specific to a gene or protein, only for

202 it to be benign. With the rate at which VUSs are being identified now this process is less practical and very time-consuming. Databases like gnomAD and ClinVar are helping with mapping of human variation, the inclusion of information such as allele frequencies, ethnicity, zygosity and phenotypes helps to forecast what impact each of these variants have. Furthermore, these sequencing databases continue to give us higher resolution, allowing us to match detailed human phenotypic information with individual nucleotide variation. Bringing these fields together, we now need a database of functional data to aid variant interpretation. One approach gaining in popularity is the use of multiplexed assays for variant effect (MAVEs), which can assess the functional effect of hundreds to thousands of variants in a single assay (reviewed in (328, 329)). This will also help to create a more standardised approach to variant analysis and therefore help with reproducibility.

Although a significant number of undiagnosed cases (22/34) remain; this study has effectively eliminated the known causes of 46,XX DSD in these cases. In the remaining cohort we are likely to find novel diagnoses. These may lie in splicing regions, unknown regulatory regions of genes, or entirely novel DSD genes.

5.3 NR5A1 variants NR5A1 has a complex activating and repressive role in gonad development, acting during bipotential gonad and ovarian/testicular differentiation. Two variants were found in the NR5A1 gene in four 46,XX (ovo)testicular DSD patients. Cellular assays showed that these variants disrupt the NR5A1/β-catenin complex, causing dysregulation of WNT signalling and NR0B1 activity (Chapter 3). This analysis of NR5A1 variants implicated a novel NR5A1 variant (p.Ala260Val) in 46,XX (ovo)testicular DSD and gave us mechanistic insights into how these variants can switch gonadal development from an ovarian to testicular fate. Interestingly, our results characterising the NR5A1 p.Arg92Trp variant contrasted with two previously published assays. This lack of reproducibility highlights the need for standardised cellular models that more closely resemble the human embryonic gonad.

Given the variable expressivity/incomplete penetrance in individuals with pathogenic NR5A1 variants, I assessed the MPS data of these four patients for genetic variants that may contribute

203 to oligogenic inheritance. This yielded some interesting candidates, specifically the POR variant found in a 46,XX ovotesticular DSD patient (DSD21), which has been found in combination with NR5A1 in a case of 46,XX T-DSD previously (148). However this study was limited in that we had access to only genes on the targeted DSD screen. A more unbiased approach could use exome sequencing data and compare both affected and unaffected 46,XX and 46,XY individuals harbouring these NR5A1 variants. One such study (330) reviewed two 46,XY DSD patients with the same NR5A1 variant but each presenting with mild hypospadias or gonadal dysgenesis. Exome analysis revealed an additional MAP3K1 variant in the patient with 46,XY gonadal dysgenesis, believed to contribute to the phenotypic variability. Investigating oligogenic inheritance in our 46,XX patients would make an interesting case study, particularly as we know this phenomenon likely contributes to other DSD genes, such as Andrgogen Receptor (2).

5.4 Developing a stem-cell based model for DSD The cohort and variant characterisation projects each highlighted the need for improved disease models in DSD research. In the next project I developed a protocol for the differentiation of iPSCs to human testis lineages. Despite significant progress in the in vitro differentiation of pluripotent cells to a range of human tissues, the generation of testicular lineages is a comparatively understudied area. The most promising advance in the differentiation of somatic testis lineages has been work by Buganim et al. (2012), where overexpression of transcription factors Nr5a1, Wt1, Dmrt1, Gata4 and Sox9 enabled reprogramming of MEFs to testis-like cells, exhibiting expression of Sertoli, Leydig and germ cell markers (200). A more recent study applied this approach to human, generating adult Sertoli-like cells from fibroblasts by overexpression of GATA4 and NR5A1 (201). An alternate approach has been work by Shlush et al. (2017), who used growth factors (including Testosterone, FSH and GDNF) and co-culture with mouse Sertoli cells over a 5-week differentiation to mimic the conditions of spermatogenesis and generate human adult Sertoli- and germ-like cells (203). While these systems show promise for generating testicular cell types, I wanted to create a more complete model of the human embryonic testis by employing a step-wise directed differentiation approach. Directed differentiation of human iPSCs to gonadal lineages has not been done before; therefore I used a differentiation protocol for kidney as a starting point. This kidney protocol was adapted for the gonad and conditions optimised further based on knowledge of mammalian gonadogenesis.

204 The first key outcome from this study was the induction of bipotential gonad lineages after 10- 12 days of iPSC monolayer differentiation. This was shown by upregulation of bipotential gonad- specific markers including LHX9, EMX2, WT1 and GATA4. Narrowing down the conditions necessary for bipotential gonad induction required testing many growth factors and compounds; this has given us insight into regulatory relationships and important signalling pathways in human gonad development. One example was the addition of human recombinant BMP4; our findings indicated that the human gonad arises from a more lateral part of the mesoderm. This challenges a widely held assumption about the origins of the human gonad. Further evidence for the role of the lateral plate mesoderm could be investigated using in vivo lineage tracing in mouse.

The second key outcome of the iPS cell analysis was the induction of Sertoli-like cells at day 12- 15 of monolayer differentiation, marked by expression of SOX9, WT1 and AMH. These cells make up the majority within these monolayers, indicating that our current protocol generates the Sertoli cell type with high efficiency. Sertoli cells are the first testicular cell to differentiate; they go on to direct differentiation of other testis lineages as well as cord formation. This fundamental role means that having an in vitro model for this cell type alone would be extremely valuable for studying 46,XY DSDs. To utilise these iPSC-derived Sertoli-like cells as a DSD model we would need to establish several additional factors. Firstly, our current gene expression analysis (qRT-PCR) represents a limited set of markers for Sertoli cells and lacks an appropriate control. Therefore this iPSC-derived population should be analysed by either bulk or single cell RNA-seq, followed by comparison to publicly available datasets from human fetal testis. This will allow us to establish how well our populations represent human fetal Sertoli cells. Furthermore, for these to be classified as human Sertoli-like cells we need to show that they are indeed functioning, this will be tested via co-culture with mouse or human germ cells.

The aim of this project was, ambitiously, to differentiate iPSCs to human fetal testis lineages and generate testicular organoids. Whilst we have not reached the final target of developing an organoid system, I have made significant advances towards achieving this. I successfully adapted a kidney differentiation protocol to generate bipotential gonad lineages and induced Sertoli cells from these. The biggest challenge has been the induction of non-Sertoli testis lineages, i.e. Leydig cells. I believe that shifting to 3D cultures will provide the best environment to allow for the formation of other testis lineages, discussed further in 5.5.

205 5.5 Future directions in iPSC modelling of DSD To generate a model that is a better representation of human fetal testis, we would like to develop organoids comprising the key fetal testis lineages as well as recognisable structures. In our current monolayer culture at day 18 I saw high cell density and cells showing morphological signs of stress. This impacts their ability to mature as Sertoli cells, to signal to other cells and does not promote formation of structures. We believe that the physiological conditions in swirler organoid culture (increased surface area exposure to media and O2) will help to resolve these issues.

A limitation in our study was the relatively small number of conditions that could be tested at once, due to time frame and cost in this culture format. Recent work optimising differentiation in other tissues has used microbioreactor arrays (331), a high-throughput approach with a miniaturised culture format (>8000 cell chambers per chip). Using this approach we could speed up protocol optimisation by testing many conditions concurrently.

Two key features contributing to the fetal testis, but lacking using our current approach, are the germ cells and vasculature. We know that the fetal testis develops normally and maintains endocrine function in the absence of germ cells (332-334), yet to form a more complete model for DSD it would be useful to have this cell type present in testis organoids. Human germ cells have recently been differentiated from iPSCs (207), yet this protocol relies on co-culture with mouse gonad cells. Together these highlight a pressing need for the germ cell field and our somatic cell field to collaborate. Ultimately we could use this published protocol (207) to derive human germ cells from iPSCs and co-culture these with our differentiated testis-like cells in swirler organoids.

Another structure we are unlikely to differentiate in this protocol is the mesonephros. Endothelial cells migrate from the mesonephros to the adjacent testis, where they contribute to vasculature and direct testis cord organisation (101, 335). The addition of human vascular cells at the stage of organoid aggregation may improve the formation of testicular structures as well as support long-term culture.

206 5.6 Conclusion The overall findings from this study have significant implications for DSD research. At the clinical level I provided several 46,XX (ovo)testicular DSD patients and their families with genetic diagnoses that may help in their clinical management. Furthermore, I identified which genetic analysis approach is most suitable for these rare phenotypes. In addition, the development of a protocol for iPSC differentiation to gonad lineages marks a significant advance towards applying human-specific disease modelling to DSD.

207 References

1. Blackless M, Charuvastra A, Derryck A. How sexually dimorphic are we? Review and synthesis. AMERICAN JOURNAL OF HUMAN BIOLOGY. 2000; 12(2):151-166.

2. Eggers S, Sadedin S, van den Bergen JA, Robevska G, Ohnesorg T, Hewitt J, et al. Disorders of sex development: insights from targeted gene sequencing of a large international patient cohort. GENOME BIOLOGY. 2016;17(1):243.

3. Capel B. The battle of the sexes. MECHANISMS OF DEVELOPMENT. 2000;92(1):89-103.

4. Schmahl J, Eicher EM, Washburn LL, Capel B. Sry induces cell proliferation in the mouse gonad. DEVELOPMENT. 2000;127(1):65-73.

5. Hu Y-C, Okumura LM, Page DC. Gata4 is required for formation of the genital ridge in mice. PLOS GENET. 2013;9(7):e1003629.

6. Kusaka M, Katoh-Fukui Y, Ogawa H, Miyabayashi K, Baba T, Shima Y, et al. Abnormal Epithelial Cell Polarity and Ectopic Epidermal Growth Factor Receptor (EGFR) Expression Induced in Emx2 KO Embryonic Gonads. ENDOCRINOLOGY. 2010;151(12):5893-904.

7. Francavilla S, Cordeschi G, Properzi G, Concordia N, Cappa F, Pozzi V. Ultrastructure of Fetal Human Gonad before Sexual-Differentiation and during Early Testicular and Ovarian Development. J SUBMICR CYTOL PATH. 1990;22(3):389-400.

8. Karl J, Capel B. Sertoli Cells of the Mouse Testis Originate from the Coelomic Epithelium. DEVELOPMENTAL BIOLOGY. 1998;203(2):323-33.

9. Martineau J, Nordqvist K, Tilmann C, Lovell-Badge R, Capel B. Male-specific cell migration into the developing gonad. CURRENT BIOLOGY. 1997;7(12):958-68.

10. Witschi E. Migration of the germ cells of human embryos from the yolk sac to the primitive gonadal folds: CONTRIB EMBRYOL; 1948.

11. De Felici M. The Formation and Migration of Primordial Germ Cells in Mouse and Man. RESULTS PROBL CELL D. 2016;58:23-46.

12. Smith C, Mackay S. Morphological development and fate of the mouse mesonephros. JOURNAL OF ANATOMY. 1991;174:171-84.

13. McLaren A, Ferguson-Smith MA, Adinolfi M. Somatic and Germ-Cell Sex in Mammals [and Discussion]. PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY OF LONDON B: BIOLOGICAL SCIENCES. 1988;322(1208):3-9.

14. McLaren A, Mittwoch U, Josso N. Germ Cells and Germ Cell Sex [and Discussion]. PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY OF LONDON B: BIOLOGICAL SCIENCES. 1995;350(1333):229-33.

15. McLaren A, Southee D. Entry of Mouse Embryonic Germ Cells into Meiosis. DEVELOPMENTAL BIOLOGY. 1997;187(1):107-13.

208 16. Hanley NA, Ball SG, Clement-Jones M, Hagan DM, Strachan T, Lindsay S, et al. Expression of steroidogenic factor 1 and Wilms' tumour 1 during early human gonadal development and sex determination. MECHANISMS OF DEVELOPMENT. 1999;87(1-2):175-80.

17. Luo X, Ikeda Y, Parker KL. A cell-specific nuclear receptor is essential for adrenal and gonadal development and sexual differentiation. CELL. 1994;77(4):481-90.

18. Sadovsky Y, Crawford PA, Woodson KG, Polish JA, Clements MA, Tourtellotte LM, et al. Mice deficient in the orphan receptor steroidogenic factor 1 lack adrenal glands and gonads but express P450 side-chain-cleavage enzyme in the placenta and have normal embryonic serum levels of corticosteroids. PROC NATL ACAD SCI U S A. 1995;92(24):10939-43.

19. Val P, Martinez-Barbera JP, Swain A. Adrenal development is initiated by Cited2 and Wt1 through modulation of Sf-1 dosage. DEVELOPMENT. 2007;134(12):2349-58.

20. Hasegawa T, Fukami M, Sato N, Katsumata N, Sasaki G, Fukutani K, et al. Testicular dysgenesis without adrenal insufficiency in a 46,XY patient with a heterozygous inactive mutation of steroidogenic factor-1. J CLIN ENDOCR METAB. 2004;89(12):5930-5.

21. Mallet D, Bretones P, Michel-Calemard L, Dijoud F, David M, Morel Y. Gonadal dysgenesis without adrenal insufficiency in a 46,XY patient heterozygous for the nonsense C16X mutation: A case of SF1 haploinsufficiency. J CLIN ENDOCR METAB. 2004;89(10):4829-32.

22. Svensson EC, Huggins GS, Dardik FB, Polk CE, Leiden JM. A functionally conserved N- terminal domain of the friend of GATA-2 (FOG-2) protein represses GATA4-dependent transcription. JOURNAL OF BIOLOGICAL CHEMISTRY. 2000;275(27):20762-9.

23. Svensson EC, Tufts RL, Polk CE, Leiden JM. Molecular cloning of FOG-2: A modulator of transcription factor GATA-4 in cardiomyocytes. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. 1999;96(3):956-61.

24. Lu JR, McKinsey TA, Xu HT, Wang DZ, Richardson JA, Olson EN. FOG-2, a heart- and brain- enriched cofactor for GATA transcription factors. MOLECULAR AND CELLULAR BIOLOGY. 1999;19(6):4495-502.

25. Tevosian SG, Albrecht KH, Crispino JD, Fujiwara Y, Eicher EM, Orkin SH. Gonadal differentiation, sex determination and normal Sry expression in mice require direct interaction between transcription partners GATA4 and FOG2. DEVELOPMENT. 2002;129(19):4627-34.

26. Birk OS, Casiano DE, Wassif CA, Cogliati T, Zhao L, Zhao Y, et al. The LIM homeobox gene Lhx9 is essential for mouse gonad formation. NATURE. 2000;403(6772):909-13.

27. Tsang TE, Shawlot W, Kinder SJ, Kobayashi A, Kwan KM, Schughart K, et al. Lim1 Activity Is Required for Intermediate Mesoderm Differentiation in the Mouse Embryo. DEVELOPMENTAL BIOLOGY. 2000;223(1):77-90.

28. Shawlot W, Behringer RR. Requirement for LIml in head-organizer function. NATURE. 1995;374(6521):425-30.

29. Fujimoto Y, Tanaka SS, Yamaguchi YL, Kobayashi H, Kuroki S, Tachibana M, et al. Homeoproteins Six1 and Six4 Regulate Male Sex Determination and Mouse Gonadal Development. DEVELOPMENTAL CELL. 2013;26(4):416-30.

209 30. Katoh-Fukui Y, Miyabayashi K, Komatsu T, Owaki A, Baba T, Shima Y, et al. Cbx2, a Polycomb Group Gene, Is Required for SryGene Expression in Mice. ENDOCRINOLOGY. 2012;153(2):913-24.

31. Katoh-Fukui Y, Tsuchiya R, Shiroishi T, Nakahara Y, Hashimoto N, Noguchi K, et al. Male- to-female sex reversal in M33 mutant mice. NATURE. 1998;393(6686):688-92.

32. Klattig J, Sierig R, Kruspe D, Makki MS, Englert C. WT1-mediated gene regulation in early urogenital ridge development. SEXUAL DEVELOPMENT. 2007;1(4):238-54.

33. Gao F, Maiti S, Alam N, et al. The Wilms tumor gene, Wt1, is required for Sox9 expression and maintenance of tubular architecture in the developing testis. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES. 2006;103(32):11987-92.

34. Hammes A, Guo J-K, Lutsch G, Leheste J-R, Landrock D, Ziegler U, et al. Two Splice Variants of the Wilms' Tumor 1 Gene Have Distinct Functions during Sex Determination and Nephron Formation. CELL. 2001;106(3):319-29.

35. Kreidberg JA, Sariola H, Loring JM, Maeda M, Pelletier J, Housman D, et al. WT-1 is required for early kidney development. CELL. 1993;74(4):679-91.

36. Miyamoto N, Yoshida M, Kuratani S, Matsuo I, Aizawa S. Defects of urogenital development in mice lacking Emx2. DEVELOPMENT. 1997;124(9):1653-64.

37. Chassot A-A, Bradford ST, Auguste A, et al. WNT4 and RSPO1 together are required for cell proliferation in the early mouse gonad. DEVELOPMENT. 2012;139(23):4461-72.

38. Chassot A-A, Ranc F, Grégoire EP, et al. Activation of -catenin signaling by Rspo1 controls differentiation of the mammalian ovary. HUMAN MOLECULAR GENETICS. 2008;17(9):1264-77.

39. Parma P, Radi O, Vidal V, Chaboissier M-C, Dellambra E, Valentini S, et al. R-spondin1 is essential in sex determination, skin differentiation and malignancy. NATURE GENETICS. 2006;38(11):1304-9.

40. Vainio S, McMahon AP, Heikkilä M, Vainio S, Kispert A, Heikkilä M, et al. Female development in mammals is regulated by Wnt-4 signalling. NATURE. 1999;397(6718):405-9.

41. Liu C-F, Parker K, Yao HH-C. WNT4/β-Catenin Pathway Maintains Female Germ Cell Survival by Inhibiting Activin βB in the Mouse Fetal Ovary. PLOS ONE. 2010;5(4):e10382.

42. Maatouk DM, DiNapoli L, Alvers A, et al. Stabilization of -catenin in XY gonads causes male-to-female sex-reversal. HUMAN MOLECULAR GENETICS. 2008;17(19):2949-55.

43. Smith CA, Shoemaker CM, Roeszler KN, Queen J, Crews D, Sinclair AH. Cloning and expression of R-Spondin1 in different vertebrates suggests a conserved role in ovarian development. BMC DEVELOPMENTAL BIOLOGY. 2008;8(1):72.

44. Heikkilä M, Prunskaite R, Naillat F, Itäranta P, Vuoristo J, Leppäluoto J, et al. The Partial Female to Male Sex Reversal in Wnt-4-Deficient Females Involves Induced Expression of Testosterone Biosynthetic Genes and Testosterone Production, and Depends on Androgen Action. ENDOCRINOLOGY. 2005;146(9):4016-23.

210 45. Tomizuka K, Horikoshi K, Kitada R, et al. R-spondin1 plays an essential role in ovarian development through positively regulating Wnt-4 signaling. HUMAN MOLECULAR GENETICS. 2008;17(9):1278-91.

46. Chassot A-A, Grégoire EP, Magliano M, Lavery R, Chaboissier MC. Genetics of Ovarian Differentiation: Rspo1, a Major Player. SEXUAL DEVELOPMENT. 2008;2(4-5):219-27.

47. Chassot A-A, Bradford ST, Auguste A, et al. WNT4 and RSPO1 together are required for cell proliferation in the early mouse gonad. DEVELOPMENT. 2012;139(23):4461-72.

48. Manuylov NL, Smagulova F, Leach L, et al. Ovarian development in mice requires the GATA4-FOG2 transcription complex. DEVELOPMENT. 2008;135(22):3731-43.

49. Nef S, Verma-Kurvari S, Merenmies J, Vassalli J-D, Efstratiadis A, Accili D, et al. Testis determination requires insulin receptor family function in mice. NATURE. 2003;426(6964):291- 5.

50. Pitetti JL, Calvel P, Romero Y, Conne B, Truong V, Papaioannou MD, et al. Insulin and IGF1 receptors are essential for XX and XY gonadal differentiation and adrenal development in mice. PLOS GENET. 2013;9(1):e1003160.

51. Naillat F, Yan W, Karjalainen R, Liakhovitskaia A, Samoylenko A, Xu Q, et al. Identification of the genes regulated by Wnt-4, a critical signal for commitment of the ovary. EXPERIMENTAL CELL RESEARCH. 2015;332(2):163-78.

52. Boulanger L, Pannetier M, Gall L, Allais-Bonnet A, Elzaiat M, Le Bourhis D, et al. FOXL2 Is a Female Sex-Determining Gene in the Goat. CURRENT BIOLOGY. 2014;24(4):404-8.

53. Pailhoux E, Vigier B, Chaffaux S, et al. A 11.7-kb deletion triggers intersexuality and polledness in goats. NATURE GENETICS. 2001;29(4):453-8.

54. Li M-H, Yang H-H, Li M-R, Sun Y-L, Jiang X-L, Xie Q-P, et al. Antagonistic Roles of Dmrt1 and Foxl2 in Sex Differentiation via Estrogen Production in Tilapia as Demonstrated by TALENs. ENDOCRINOLOGY. 2013;154(12):4814-25.

55. Schmidt D, Ovitt CE, Anlag K, Fehsenfeld S, Gredsted L, Treier A-C, et al. The murine winged-helix transcription factor Foxl2 is required for granulosa cell differentiation and ovary maintenance. DEVELOPMENT. 2004;131(4):933-42.

56. Crisponi L, Deiana M, Loi A, et al. The putative forkhead transcription factor FOXL2 is mutated in blepharophimosis/ptosis/epicanthus inversus syndrome. NATURE GENETICS. 2001;27(2):159-66.

57. Verdin H, De Baere E. FOXL2 Impairment in Human Disease. HORMONE RESEARCH IN PÆDIATRICS. 2012;77(1):2-11.

58. Uhlenhaut NH, Jakob S, Anlag K, Eisenberger T, Sekido R, Kress J, et al. Somatic Sex Reprogramming of Adult Ovaries to Testes by FOXL2 Ablation. CELL. 2009;139(6):1130-42.

59. Fleming NI, Knower KC, Lazarus KA, Fuller PJ, Simpson ER, Clyne CD. Aromatase Is a Direct Target of FOXL2: C134W in Granulosa Cell Tumors via a Single Highly Conserved Binding Site in the Ovarian Specific Promoter. PLOS ONE. 2010;5(12):e14389.

211 60. Pannetier M, Fabre S, Batista F, Kocer A, Renault L, et al. FOXL2 activates P450 aromatase gene transcription: towards a better characterization of the early steps of mammalian ovarian development. JOURNAL OF MOLECULAR ENDOCRINOLOGY. 2006;36(3):399-413.

61. Wang D-S, Kobayashi T, Zhou L-Y, Paul-Prasanth B, Ijiri S, Sakai F, et al. Foxl2 Up- Regulates Aromatase Gene Transcription in a Female-Specific Manner by Binding to the Promoter as Well as Interacting with Ad4 Binding Protein/Steroidogenic Factor 1. MOLECULAR ENDOCRINOLOGY. 2007;21(3):712-25.

62. Takasawa K, Kashimada K, Pelosi E, et al. FOXL2 transcriptionally represses Sf1 expression by antagonizing WT1 during ovarian development in mice. THE FASEB JOURNAL. 2014;28(5):2020-8.

63. Liu C, Peng J, Matzuk MM, Yao HH-C. Lineage specification of ovarian theca cells requires multicellular interactions via oocyte and granulosa cells. NATURE COMMUNICATIONS. 2015;6:6934.

64. Knarston I, Ayers K, Sinclair A. Molecular mechanisms associated with 46,XX disorders of sex development. CLINICAL SCIENCE. 2016;130(6):421-32.

65. Berta P, Hawkins JR, Sinclair AH, et al. Genetic evidence equating SRY and the testis- determining factor. NATURE. 1990;348(6300):448-50.

66. Gubbay J, Collignon J, Koopman P, et al. A gene mapping to the sex-determining region of the mouse Y chromosome is a member of a novel family of embryonically expressed genes. NATURE. 1990;346(6281):245-50.

67. Sinclair AH, Berta P, Palmer MS, et al. A gene from the human sex-determining region encodes a protein with homology to a conserved DNA-binding motif. NATURE. 1990;346(6281):240-4.

68. Maier EM, Leitner C, Löhrs U, Kuhnle U. True hermaphroditism in an XY individual due to a familial point mutation of the SRY gene. JPEM. 2003;16(4):575-80.

69. Lovell-Badge R, Robertson E. XY female mice resulting from a heritable mutation in the primary testis-determining gene, Tdy. DEVELOPMENT. 1990;109(3):635-46.

70. Margarit E, Coll MD, Oliva R, Gómez D, Soler A, Ballesta F. SRY gene transferred to the long arm of the X chromosome in a Y-positive XX true hermaphrodite. AMERICAN JOURNAL OF MEDICAL GENETICS. 2000;90(1):25-8.

71. Sharp A, Kusz K, Jaruzelska J, Tapper W, Szarras-Czapnik M, Wolski J, et al. Variability of sexual phenotype in 46,XX(SRY+) patients: the influence of spreading X inactivation versus position effects. JOURNAL OF MEDICAL GENETICS. 2005;42(5):420-7.

72. Koopman P, Gubbay J, Vivian N, et al. Male development of chromosomally female mice transgenic for Sry. NATURE. 1991;351(6322):117-21.

73. De Santa Barbara P, Méjean C, Moniot B, Malclès MH, Berta P, Boizet-Bonhoure B. Steroidogenic factor-1 contributes to the cyclic-adenosine monophosphate down-regulation of human SRY gene expression. BIOLOGY OF REPRODUCTION. 2001;64(3):775-83.

212 74. Larney C, Bailey TL, Koopman P. Switching on sex: transcriptional regulation of the testis- determining gene Sry. DEVELOPMENT. 2014;141(11):2195-205.

75. Sekido R, Bar I, Narvaez V, Penny G, Lovell-Badge R. SOX9 is up-regulated by the transient expression of SRY specifically in Sertoli cell precursors. DEVELOPMENTAL BIOLOGY. 2004;274(2):271-9.

76. Sekido R, Lovell-Badge R. Sex determination involves synergistic action of SRY and SF1 on a specific Sox9 enhancer. NATURE. 2008;453(7197):930-4.

77. Kim Y, Kobayashi A, Sekido R, DiNapoli L, Brennan J, Chaboissier M-C, et al. Fgf9 and Wnt4 Act as Antagonistic Signals to Regulate Mammalian Sex Determination. PLOS BIOLOGY. 2006;4(6):e187.

78. Malki S, Nef S, Notarnicola C, Thevenet L, Gasca S, Méjean C, et al. Prostaglandin D2 induces nuclear import of the sex-determining factor SOX9 via its cAMP-PKA phosphorylation. THE EMBO JOURNAL. 2005;24(10):1798-809.

79. Foster JW, Dominguez-Steglich MA, Guioli S, Kwok C. Campomelic dysplasia and autosomal sex reversal caused by mutations in an SRY-related gene. NATURE. 1994.

80. Wagner T, Wirth J, Meyer J, Zabel B, Held M, Zimmer J, et al. Autosomal sex reversal and campomelic dysplasia are caused by mutations in and around the SRY-related gene SOX9. CELL. 1994;79(6):1111-20.

81. Cox JJ, Willatt L, Homfray T, Woods CG. A SOX9 duplication and familial 46,XX developmental testicular disorder. THE NEW ENGLAND JOURNAL OF MEDICINE. 2011;364(1):91- 3.

82. Huang B, Wang S, Ning Y, Lamb AN, Bartley J. Autosomal XX sex reversal caused by duplication of SOX9. AMERICAN JOURNAL OF MEDICAL GENETICS. 1999;87(4):349-53.

83. Arango NA, Lovell-Badge R, Behringer RR. Targeted mutagenesis of the endogenous mouse Mis gene promoter: In vivo definition of genetic pathways of vertebrate sexual development. CELL. 1999;99(4):409-19.

84. Wilson MJ, Jeyasuria P, Parker KL, Koopman P. The transcription factors steroidogenic factor-1 and SOX9 regulate expression of Vanin-1 during mouse testis development. JOURNAL OF BIOLOGICAL CHEMISTRY. 2005;280(7):5917-23.

85. Kashimada K, Svingen T, Feng CW, Pelosi E, Bagheri-Fam S, Harley VR, et al. Antagonistic regulation of Cyp26b1 by transcription factors SOX9/SF1 and FOXL2 during gonadal development in mice. FASEB J. 2011;25(10):3561-9.

86. Bradford ST, Hiramatsu R, Maddugoda MP, Bernard P, Chaboissier MC, Sinclair A, et al. The Cerebellin 4 Precursor Gene Is a Direct Target of SRY and SOX9 in Mice. BIOLOGY OF REPRODUCTION. 2009;80(6):1178-88.

87. Bhandari RK, Haque MM, Skinner MK. Global Genome Analysis of the Downstream Binding Targets of Testis Determining Factor SRY and SOX9. PLOS ONE. 2012;7(9).

213 88. Seminara SB, Achermann JC, Genel M, et al. X-Linked Adrenal Hypoplasia Congenita: A Mutation in DAX1Expands the Phenotypic Spectrum in Males and Females 1. THE JOURNAL OF CLINICAL ENDOCRINOLOGY AND METABOLISM. 1999;84(12):4501-9.

89. Bardoni B, Zanaria E, Guioli S, Floridia G, Worley KC, Tonini G, et al. A dosage sensitive locus at chromosome Xp21 is involved in male to female sex reversal. NATURE GENETICS. 1994;7(4):497-501.

90. Lovell-Badge R, Swain A, Narvaez V, Swain A, Burgoyne P, Narvaez V, et al. Dax1 antagonizes Sry action in mammalian sex determination. NATURE. 1998;391(6669):761-7.

91. Ludbrook LM, Bernard P, Bagheri-Fam S. Excess DAX1 leads to XY ovotesticular disorder of sex development (DSD) in mice by inhibiting steroidogenic factor-1 (SF1) activation of the testis enhancer of SRY-box-9 (Sox9). ENDOCRINOLOGY. 2012; 153(4):1948-58. 92. Bouma GJ, Washburn LL, Albrecht KH, et al. Correct dosage of Fog2 and Gata4 transcription factors is critical for fetal testis development in mice. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES. 2007;104(38):14994-9.

93. DeFalco T, Takahashi S, Capel B. Two distinct origins for Leydig cell progenitors in the fetal testis. DEVELOPMENTAL BIOLOGY. 2011;352(1):14-26.

94. Wu XF, Wan SQ, Lee MM. Key factors in the regulation of fetal and postnatal Leydig cell development. J CELL PHYSIOL. 2007;213(2):429-33.

95. Hoivik EA, Lewis AE, Aumo L, Bakke M. Molecular aspects of steroidogenic factor 1 (SF- 1). MOLECULAR AND CELLULAR ENDOCRINOLOGY. 2010;315(1-2):27-39.

96. Sugawara T, Holt JA, Kiriakidou M, Strauss JF. Steroidogenic factor 1-dependent promoter activity of the human steroidogenic acute regulatory protein (StAR) gene. BIOCHEMISTRY-US. 1996;35(28):9052-9.

97. Parker KL, Schimmer BP. Steroidogenic factor 1: a key determinant of endocrine development and function. ENDOCR REV. 1997;18(3):361-77.

98. Yao HH-C, Whoriskey W, Capel B. Desert Hedgehog/Patched 1 signaling specifies fetal Leydig cell fate in testis organogenesis. GENES & DEVELOPMent. 2002;16(11):1433-40.

99. Das DK, Sanghavi D, Gawde H, Idicula-Thomas S, Vasudevan L. Novel homozygous mutations in Desert Hedgehog gene in patients with 46,XY complete gonadal dysgenesis and prediction of its structural and functional implications by computational methods. EUROPEAN JOURNAL OF MEDICAL GENETICS. 2011;54(6):E529-E34.

100. Werner R, Merz H, Birnbaum W, Marshall L, Schroder T, Reiz B, et al. 46,XY Gonadal Dysgenesis due to a Homozygous Mutation in Desert Hedgehog (DHH) Identified by Exome Sequencing. J CLIN ENDOCR METAB. 2015;100(7):E1022-E9.

101. Combes AN, Wilhelm D, Davidson T, Dejana E, Harley V, Sinclair A, et al. Endothelial cell migration directs testis cord formation. DEVELOPMENTAL BIOLOGY. 2009;326(1):112-20.

102. Brennan J, Karl J, Capel B. Divergent vascular mechanisms downstream of Sry establish the arterial system in the XY gonad. DEV BIOL. 2002;244(2):418-28.

214 103. Yao HH-C, Aardema J, Holthusen K. Sexually dimorphic regulation of inhibin beta B in establishing gonadal vasculature in mice. BIOLOGY OF REPRODUCTION. 2006;74(5):978-83.

104. DeFalco T, Bhattacharya I, Williams AV, Sams DM, Capel B. Yolk-sac-derived macrophages regulate fetal testis vascularization and morphogenesis. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. 2014;111(23):E2384- 93.

105. Matson CK, Murphy MW, Sarver AL, Griswold MD, Bardwell VJ, Zarkower D. DMRT1 prevents female reprogramming in the postnatal mammalian testis. NATURE. 2011;476(7358):101-4.

106. Minkina A, Matson CK, Lindeman RE, Ghyselinck NB, Bardwell VJ, Zarkower D. DMRT1 Protects Male Gonadal Cells from Retinoid-Dependent Sexual Transdifferentiation. DEVELOPMENTAL CELL. 2014;29(5):511-20.

107. Bernard P, Sim H, Knower K, Vilain E, Harley V. Human SRY inhibits β-catenin-mediated transcription. THE INTERNATIONAL JOURNAL OF BIOCHEMISTRY & CELL BIOLOGY. 2008;40(12):2889-900.

108. Jameson SA, Lin Y-T, Capel B. Testis development requires the repression of Wnt4 by Fgf signaling. DEVELOPMENTAL BIOLOGY. 2012;370(1):24-32.

109. Piprek RP. Genetic mechanisms underlying male sex determination in mammals. JOURNAL OF APPLIED GENETICS. 2009;50(4):347-60.

110. Yao HH-C, Matzuk MM, Jorgez CJ, Menke DB, Page DC, Swain A, et al. Follistatin operates downstream ofWnt4 in mammalian ovary organogenesis. DEVELOPMENTAL DYNAMICS. 2004;230(2):210-5.

111. Lindeman RE, Gearhart MD, Minkina A, Krentz AD, Bardwell VJ, Zarkower D. Sexual Cell- Fate Reprogramming in the Ovary by DMRT1. CURRENT BIOLOGY. 2015;25(6):764-71.

112. Bowles J, Koopman P. Retinoic acid, meiosis and germ cell fate in mammals. DEVELOPMENT. 2007;134(19):3401-11.

113. Ottolenghi C, Pelosi E, Tran J, et al. Loss of Wnt4 and Foxl2 leads to female-to-male sex reversal extending to germ cells. HUMAN MOLECULAR GENETICS. 2007;16(23):2795-804.

114. Erdoğan S, Kara C, Uçaktürk A, Aydın M. Etiological Classification and Clinical Assessment of Children and Adolescents with Disorders of Sex Development - Original Article. JOURNAL OF CLINICAL RESEARCH IN PEDIATRIC ENDOCRINOLOGY. 2011;3(2):77-83.

115. Jaruratanasirikul S, Engchaun V. Management of children with disorders of sex development: 20-year experience in southern Thailand. WORLD JOURNAL OF PEDIATRICS. 2013;10(2):168-74.

116. Öçal G, Berberoğlu M, Sıklar Z, Aycan Z, Hacıhamdioglu B, Erdeve ŞS, et al. Clinical Review of 95 Patients with 46,XX Disorders of Sex Development Based on the New Chicago Classification. JOURNAL OF PEDIATRIC AND ADOLESCENT GYNECOLOGY. 2015;28(1):6-11.

117. Hughes IA, Houk C, Ahmed SF, Lee PA, Group LWPESESfPEC. Consensus statement on management of intersex disorders. JOURNAL OF PEDIATRIC UROLOGY. 2006;2(3):148-62.

215 118. Kousta E, Papathanasiou A, Skordis N. Sex determination and disorders of sex development according to the revised nomenclature and classification in 46,XX individuals. HORMONES. 2010;9(3):218-131.

119. Pagon RA, Adam MP, Ardinger HH, Wallace SE, Amemiya A, Bean LJ, et al. Nonsyndromic 46,XX Testicular Disorders of Sex Development. GENE REVIEWS. 1993.

120. López M, Torres L, Méndez JP, Cervantes A, Pérez-Palacios G, Erickson RP, et al. Clinical traits and molecular findings in 46,XX males. CLINICAL GENETICS. 1995;48(1):29-34.

121. McElreavey K, Vilain E, Abbas N. A regulatory cascade hypothesis for mammalian sex determination: SRY represses a negative regulator of male development. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. 1993; 90(8):3368-72.

122. Aksglaede L, Skakkebaek NE, Juul A. Abnormal Sex Chromosome Constitution and Longitudinal Growth: Serum Levels of Insulin-Like Growth Factor (IGF)-I, IGF Binding Protein-3, Luteinizing Hormone, and Testosterone in 109 Males with 47,XXY, 47,XYY, or Sex-Determining Region of the Y Chromosome (SRY)-Positive 46,XX Karyotypes. THE JOURNAL OF CLINICAL ENDOCRINOLOGY AND METABOLISM. 2008;93(1):169-76.

123. Vetro A, Ciccone R, Giorda R, Patricelli MG, Della Mina E, Forlino A, et al. XX males SRY negative: a confirmed cause of infertility. J MED GENET. 2011;48(10):710-2.

124. Kim GJ, Sock E, Buchberger A, et al. Copy number variation of two separate regulatory regions upstream of SOX9 causes isolated 46,XY or 46,XX disorder of sex development. JOURNAL OF MEDICAL GENETICS. 2015;52(4):240-7.

125. Vetro A, Dehghani MR, Kraoua L, Giorda R, Beri S, Cardarelli L, et al. Testis development in the absence of SRY: chromosomal rearrangements at SOX9 and SOX3. EUROPEAN JOURNAL OF HUMAN GENETICS. 2015;23(8):1025-32.

126. Fonseca ACS, Bonaldi A, Bertola DR, Kim CA, Otto PA, Vianna-Morgante AM. The clinical impact of chromosomal rearrangements with breakpoints upstream of the SOX9 gene: two novel de novo balanced translocations associated with acampomelic campomelic dysplasia. BMC MEDICAL GENETICS. 2013;14(1):50.

127. Hill-Harfe KL, Kaplan L, Stalker HJ, Zori RT, Pop R, Scherer G, et al. Fine Mapping of Chromosome 17 Translocation Breakpoints ⩾900 Kb Upstream of SOX9 in Acampomelic Campomelic Dysplasia and a Mild, Familial Skeletal Dysplasia. THE AMERICAN JOURNAL OF HUMAN GENETICS. 2005;76(4):663-71.

128. Velagaleti GVN, Bien-Willner GA, Northup JK, Lockhart LH, Hawkins JC, Jalal SM, et al. Position Effects Due to Chromosome Breakpoints that Map ∼900 Kb Upstream and ∼1.3 Mb Downstream of SOX9 in Two Patients with Campomelic Dysplasia. THE AMERICAN JOURNAL OF HUMAN GENETICS. 2005;76(4):652-62.

129. Kurth I, Klopocki E, Stricker S, van Oosterwijk J, Vanek S, Altmann J, et al. Duplications of noncoding elements 5' of SOX9 are associated with brachydactyly-anonychia. NAT GENET. 2009;41(8):862-3.

130. Sutton E, Hughes J, White S, Sekido R, Tan J, Arboleda V, et al. Identification of SOX3 as an XX male sex reversal gene in mice and humans. JOURNAL OF CLINICAL INVESTIGATION. 2011;121(1):328-41.

216 131. Nicholl RM, Grimsley L, Butler L, Palmer RW, Rees HC, Savage MO, et al. Trisomy 22 and intersex. ARCHIVES OF DISEASE IN CHILDHOOD FETAL AND NEONATAL EDITION. 1994;71(1):F57- 8.

132. Seeherunvong T, Perera EM, Bao Y, Benke PJ, Benigno A, Donahue RP, et al. 46,XX sex reversal with partial duplication of chromosome arm 22q. AMERICAN JOURNAL OF MEDICAL GENETICS. 2004;127A(2):149-51.

133. Aleck KA, Argueso L, Stone J, Hackel JG, Erickson RP. True hermaphroditism with partial duplication of chromosome 22 and without SRY. AMERICAN JOURNAL OF MEDICAL GENETICS. 1999;85(1):2-4.

134. Cantu JM, Hernandez A, Vaca G, Plascencia L, Martinez-Basalo C, Ibarra B, et al. Trisomy 22q12 leads to qter: “aneusomie de recombinaison” of a pericentric inversion. ANNALES DE GÉNÉTIQUE. 1981;24(1):37-40.

135. Chiang H-S, Wu Y-N, Wu C-C, Hwang J-L. Cytogenic and molecular analyses of 46,XX male syndrome with clinical comparison to other groups with testicular azoospermia of genetic origin. JOURNAL OF THE FORMOSAN MEDICAL ASSOCIATION. 2013;112(2):72-8.

136. Eozenou C, Fusee L, Elaidy A, Mcelreavey K, Mazen I, Bashamboo A. A Mutation in Wt1 (Wilms' Tumor Suppressor 1) Associated with 46,XX Testicular DSD. HORM RES PAEDIAT. 2017;88:203.

137. Gomes NL, de Paula LCP, Silva JM, Silva TE, Lerario AM, Nishi MY, et al. A 46,XX Testicular Disorder of Sex Development Caused by a Wilms' Tumour Factor-1 (WT1) Pathogenic Variant. CLIN GENET. 2018.

138. Kojima Y, Hayashi Y, Mizuno K, Sasaki S, Fukui Y, Koopman P, et al. Up-regulation of SOX9 in human sex-determining region on the Y chromosome (SRY)-negative XX males. CLINICAL ENDOCRINOLOGY. 2008;68(5):791-9.

139. Eid W, Opitz L, Biason-Lauber A. Genome-Wide Identification of CBX2 Targets: Insights in the Human Sex Development Network. MOLECULAR ENDOCRINOLOGY. 2015;29(2):247-57.

140. Kuroiwa A, Handa S, Nishiyama C, Chiba E, Yamada F, Abe S, et al. Additional copies of CBX2 in the genomes of males of mammals lacking SRY, the Amami spiny rat (Tokudaia osimensis) and the Tokunoshima spiny rat (Tokudaia tokunoshimensis). CHROMOSOME RESEARCH. 2011;19(5):635-44.

141. Naasse Y, Bakhchane A, Charoute H, Jennane F, Bignon-Topalovic J, Malki A, et al. A Novel Homozygous Missense Mutation in the FU-CRD2 Domain of the R-spondin1 Gene Associated with Familial 46,XX DSD. SEX DEV. 2017;11(5-6):269-74.

142. Mandel H, Shemer R, Borochowitz ZU, Okopnik M, Knopf C, Indelman M, et al. SERKAL Syndrome: An Autosomal-Recessive Disorder Caused by a Loss-of-Function Mutation in WNT4. THE AMERICAN JOURNAL OF HUMAN GENETICS. 2008;82(1):39-47.

143. Bashamboo A, Eozenou C, Jorgensen A, Bignon-Topalovic J, Siffroi JP, Hyon C, et al. Loss of Function of the Nuclear Receptor NR2F2, Encoding COUP-TF2, Causes Testis Development and Cardiac Defects in 46,XX Children. AM J HUM GENET. 2018;102(3):487-93.

217 144. Hsu W-H, Chen C-M, You L-R. COUP-TFII is required for morphogenesis of the neural crest-derived tympanic ring. SCIENTIFIC REPORTS. 2017;7(1):12386.

145. Li X, Large MJ, Creighton CJ, Lanz RB, Jeong JW, Young SL, et al. COUP-TFII regulates human endometrial stromal genes involved in inflammation. MOLECULAR ENDOCRINOLOGY. 2013;27(12):2041-54.

146. Baetens D, Stoop H, Peelman F, Todeschini A-L, Rosseel T, Coppieters F, et al. NR5A1 is a novel disease gene for 46,XX testicular and ovotesticular disorders of sex development. GENETICS IN MEDICINE. 2016.

147. Bashamboo A, Donohoue PA, Vilain E, Rojo S, Calvel P, Seneviratne SN, et al. A recurrent p.Arg92Trp variant in steroidogenic factor-1 (NR5A1) can act as a molecular switch in human sex development. HUMAN MOLECULAR GENETICS. 2016; 25(16):3446-3453..

148. Igarashi M, Takasawa K, Hakoda A, Kanno J, Takada S, Miyado M, et al. Identical NR5A1 Missense Mutations in Two Unrelated 46,XX Individuals with Testicular Tissues. HUMAN MUTATION. 2017; 38(1):39-42.

149. Combes AN, Spiller CM, Harley VR, Sinclair AH, Dunwoodie SL, Wilhelm D, et al. Gonadal defects in Cited2-mutant mice indicate a role for SF1 in both testis and ovary differentiation. THE INTERNATIONAL JOURNAL OF DEVELOPMENTAL BIOLOGY. 2010;54(4):683-9.

150. De Santa Barbara P, Bonneaud N, Boizet B, Desclozeaux M, Moniot B, Sudbeck P, et al. Direct interaction of SRY-related protein SOX9 and steroidogenic factor 1 regulates transcription of the human anti-Mullerian hormone gene. MOL CELL BIOL. 1998;18(11):6653-65.

151. Vilain E. The Genetics of Ovotesticular Disorders of Sex Development. SPRINGER NEW YORK. 2011; p. 105-6.

152. Dutta D, Shivaprasad KS, Das RN, Ghosh S, Chatterjee U, Chowdhury S, et al. Ovotesticular disorder of sexual development due to 47,XYY/46,XY/45,X mixed gonadal dysgenesis in a phenotypic male presenting as cyclical haematuria: clinical presentation and assessment of long-term outcomes. ANDROLOGIA. 2012;46(2):191-3.

153. Benko S, Gordon CT, Mallet D, Sreenivasan R, Thauvin-Robinet C, Brendehaug A, et al. Disruption of a long distance regulatory region upstream of SOX9 in isolated disorders of sex development. JOURNAL OF MEDICAL GENETICS. 2011;48(12):825-30.

154. Ohnesorg T, van den Bergen JA, Belluoccio D, Shankara-Narayana N, Kean AM, Vasilaras A, et al. A duplication in a patient with 46,XX ovo-testicular disorder of sex development refines the SOX9testis-specific regulatory region to 24 kb. CLINICAL GENETICS. 2017;372(8):525.

155. Tomaselli S, Megiorni F, De Bernardo C, Felici A, Marrocco G, Maggiulli G, et al. Syndromic true hermaphroditism due to an R-spondin1 ( RSPO1) homozygous mutation. HUMAN MUTATION. 2008;29(2):220-6.

156. Biason-Lauber A, Konrad D, Navratil F, Schoenle EJ. A WNT4Mutation Associated with Müllerian-Duct Regression and Virilization in a 46,XX Woman. THE NEW ENGLAND JOURNAL OF MEDICINE. 2004;351(8):792-8.

157. Philibert P, Biason-Lauber A, Rouzier R, Pienkowski C, Paris F, Konrad D, et al. Identification and Functional Analysis of a New WNT4 Gene Mutation among 28 Adolescent Girls

218 with Primary Amenorrhea and Müllerian Duct Abnormalities: A French Collaborative Study. THE JOURNAL OF CLINICAL ENDOCRINOLOGY AND METABOLISM. 2008;93(3):895-900.

158. Mizuno K, Kojima Y, Kamisawa H, Moritoki Y, Nishio H, Nakane A, et al. Elucidation of Distinctive Genomic DNA Structures in Patients with 46,XX Testicular Disorders of Sex Development Using Genome Wide Analyses. THE JOURNAL OF UROLOGY. 2014;192(2):535-41.

159. de Lau W, Barker N, Low TY, Koo B-K, Li VSW, Teunissen H, et al. Lgr5 homologues associate with Wnt receptors and mediate R-spondin signalling. NATURE. 2011;476(7360):293- 7.

160. Koizumi M, Oyama K, Yamakami Y, et al. Lgr4 Controls Specialization of Female Gonads in Mice. BIOLOGY OF REPRODUCTION. 2015;93(4):90-.

161. Ropke A, Tewes AC, Gromoll J, Kliesch S, Wieacker P, Tuttelmann F. Comprehensive sequence analysis of the NR5A1 gene encoding steroidogenic factor 1 in a large group of infertile males. EUR J HUM GENET. 2013;21(9):1012-5.

162. Robevska G, van den Bergen JA, Ohnesorg T, Eggers S, Hanna C, Hersmus R, et al. Functional characterization of novel NR5A1 variants reveals multiple complex roles in disorders of sex development. HUM MUTAT. 2018;39(1):124-39.

163. Voican A, Bachelot A, Bouligand J, Francou B, Dulon J, Lombes M, et al. NR5A1 (SF-1) mutations are not a major cause of primary ovarian insufficiency. J CLIN ENDOCRINOL METAB. 2013;98(5):E1017-21.

164. Dangle P, Touzon MS, Reyes-Mugica M, Witchel SF, Rajkovic A, Schneck FX, et al. Female-to-male sex reversal associated with unique Xp21.2 deletion disrupting genomic regulatory architecture of the dosage-sensitive sex reversal region. J MED GENET. 2017;54(10):705-9.

165. Johansson HKL, Svingen T, Fowler PA, Vinggaard AM, Boberg J. Environmental influences on ovarian dysgenesis - developmental windows sensitive to chemical exposures. NATURE REVIEWS ENDOCRINOLOGY. 2017;13(7):400-14.

166. Bashamboo A, McElreavey K. Gene Mutations Associated with Anomalies of Human Gonad Formation. SEXUAL DEVELOPMENT. 2013;7(1-3):126-46.

167. Meyers CM, Boughman JA, Rivas M, Wilroy RS, Simpson JL. Gonadal (ovarian) dysgenesis in 46,XX individuals: Frequency of the autosomal recessive form. AMERICAN JOURNAL OF MEDICAL GENETICS. 1996;63(4):518-24.

168. Palmer JR, Hatch EE, Rao RS, Kaufman RH, Herbst AL, Noller KL, et al. Infertility among women exposed prenatally to diethylstilbestrol. AM J EPIDEMIOL. 2001;154(4):316-21.

169. Abel MH, Wootton AN, Wilkins V, et al. The Effect of a Null Mutation in the Follicle- Stimulating Hormone Receptor Gene on Mouse Reproduction 1. ENDOCRINOLOGY. 2000;141(5):1795-803.

170. Lalioti MD. Impact of follicle stimulating hormone receptor variants in fertility. CURRENT OPINION IN OBSTETRICS & GYNECOLOGY. 2011;23(3):158-67.

219 171. Aittomäki K, Lucena JL, Pakarinen P, Sistonen P, Tapanainen J, Gromoll J, et al. Mutation in the follicle-stimulating hormone receptor gene causes hereditary hypergonadotropic ovarian failure. CELL. 1995;82(6):959-68.

172. Di Pasquale E, Beck-Peccoz P, Persani L. Hypergonadotropic Ovarian Failure Associated with an Inherited Mutation of Human Bone Morphogenetic Protein-15 (BMP15) Gene. The AMERICAN JOURNAL OF HUMAN GENETICS. 2004;75(1):106-11.

173. Lang-Muritano M, Sproll P, Wyss S, Kolly A, Hurlimann R, Konrad D, et al. Early onset complete ovarian failure and lack of puberty in a woman with mutated estrogen receptor beta (ESR2). J CLIN ENDOCRINOL METAB. 2018;103(10):3748-3756.

174. Ko L, Cardona GR, Henrion-Caude A, et al. Identification and Characterization of a Tissue- Specific Coactivator, GT198, That Interacts with the DNA-Binding Domains of Nuclear Receptors. MOLECULAR AND CELLULAR BIOLOGY. 2002;22(1):357-69.

175. Zangen D, Kaufman Y, Zeligson S, Perlberg S, Fridman H, Kanaan M, et al. XX Ovarian Dysgenesis Is Caused by a PSMC3IP/HOP2 Mutation that Abolishes Coactivation of Estrogen- Driven Transcription. THE AMERICAN JOURNAL OF HUMAN GENETICS. 2011;89(4):572-9.

176. Rajkovic A, Pangas SA, Ballow D, Suzumori N, Matzuk MM. NOBOX Deficiency Disrupts Early Folliculogenesis and Oocyte-Specific Gene Expression. SCIENCE. 2004;305(5687):1157-9.

177. Bouilly J, Bachelot A, Broutin I, Touraine P, Binart N. Novel NOBOX loss-of-function mutations account for 6.2% of cases in a large primary ovarian insufficiency cohort. HUM MUTAT. 2011;32(10):1108-13.

178. Qin Y, Shi Y, Zhao Y, Carson SA, Simpson JL, Chen Z-J. Mutation analysis of NOBOX homeodomain in chinese women with premature ovarian failure. FERTILITY AND STERILITY. 2009;91(4):1507-9.

179. Zhao XX, Suzumori N, Yamaguchi M, Suzumori K. Mutational analysis of the homeobox region of the human NOBOX gene in Japanese women who exhibit premature ovarian failure. FERTILITY AND STERILITY. 2005;83(6):1843-4.

180. Weinberg-Shukron A, Renbaum P, Kalifa R, Zeligson S, Ben-Neriah Z, Dreifuss A, et al. A mutation in the nucleoporin-107 gene causes XX gonadal dysgenesis. JOURNAL OF CLINICAL INVESTIGATION. 2015;125(11):4295-304.

181. Cole F, Kauppi L, Lange J, Roig I, Wang R, Keeney S, et al. Homeostatic control of recombination is implemented progressively in mouse meiosis. NATURE CELL BIOLOGY. 2012;14(4):424-30.

182. Kim GJ, Sock E, Buchberger A, et al. Copy number variation of two separate regulatory regions upstream of SOX9 causes isolated 46,XY or 46,XX disorder of sex development. JOURNAL OF MEDICAL GENETICS. 2015;52(4):240-7.

183. McClelland KS, Yao HH. Leveraging Online Resources to Prioritize Candidate Genes for Functional Analyses: Using the Fetal Testis as a Test Case. SEX DEV. 2017;11(1):1-20.

184. Evans MJ, Kaufman MH. Establishment in culture of pluripotential cells from mouse embryos. NATURE. 1981;292(5819):154-6.

220 185. Martin GR. Isolation of a pluripotent cell line from early mouse embryos cultured in medium conditioned by teratocarcinoma stem cells. PROC NATL ACAD SCI U S A. 1981;78(12):7634-8.

186. Takahashi K, Yamanaka S. Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. CELL. 2006;126(4):663-76.

187. Lancaster MA, Knoblich JA. Generation of cerebral organoids from human pluripotent stem cells. NAT PROTOC. 2014;9(10):2329-40.

188. Takasato M, Er PX, Chiu HS, Little MH. Generation of kidney organoids from human pluripotent stem cells. NATURE PROTOCOLS. 2016;11(9):1681-92.

189. Yazawa T, Mizutani T, Yamada K, Kawata H, Sekiguchi T, Yoshino M, et al. Differentiation of adult stem cells derived from bone marrow stroma into Leydig or adrenocortical cells. ENDOCRINOLOGy. 2006;147(9):4104-11.

190. Gondo S, Okabe T, Tanaka T, Morinaga H, Nomura M, Takayanagi R, et al. Adipose tissue- derived and bone marrow-derived mesenchymal cells develop into different lineage of steroidogenic cells by forced expression of steroidogenic factor 1. ENDOCRINOLOGY. 2008;149(9):4717-25.

191. Jadhav U, Jameson JL. Steroidogenic Factor-1 (SF-1)-Driven Differentiation of Murine Embryonic Stem (ES) Cells into a Gonadal Lineage. ENDOCRINOLOGY. 2011;152(7):2870-82.

192. Crawford PA, Sadovsky Y, Milbrandt J. Nuclear receptor steroidogenic factor 1 directs embryonic stem cells toward the steroidogenic lineage. MOLECULAR AND CELLULAR BIOLOGY. 1997;17(7):3997-4006.

193. Sonoyama T, Sone M, Honda K, Taura D, Kojima K, Inuzuka M, et al. Differentiation of Human Embryonic Stem Cells and Human Induced Pluripotent Stem Cells into Steroid-Producing Cells. ENDOCRINE REVIEWS. 2014;35(3).

194. Yang Y, Su ZJ, Xu WT, Luo J, Liang R, Xiang Q, et al. Directed Mouse Embryonic Stem Cells into Leydig-Like Cells Rescue Testosterone-Deficient Male Rats In Vivo. STEM CELLS AND DEVELOPMENT. 2015;24(4):459-70.

195. Yang Y, Li Z, Wu X, Chen H, Xu W, Xiang Q, et al. Direct Reprogramming of Mouse Fibroblasts toward Leydig-like Cells by Defined Factors. STEM CELL REPORTS. 2017;8(1):39-53.

196. Bucay N, Yebra M, Cirulli V, Afrikanova I, Kaido T, Hayek A, et al. A Novel Approach for the Derivation of Putative Primordial Germ Cells and Sertoli Cells from Human Embryonic Stem Cells. STEM CELLS. 2009;27(1):68-77.

197. Dym M. The fine structure of the monkey (Macaca) Sertoli cell and its role in maintaining the blood-testis barrier. THE ANATOMICAL RECORD. 1973;175(4):639-56.

198. Myers M, Ebling FJP, Nwagwu M, Boulton R, Wadhwa K, Stewart J, et al. Atypical development of Sertoli cells and impairment of spermatogenesis in the hypogonadal (hpg) mouse. JOURNAL OF ANATOMY. 2005;207(6):797-811.

221 199. Rune GM, Pretzer D, De Souza P, Bollmann U, Merker HJ. Ultrastructure of adult and juvenile marmoset (Callithrix jacchus) Sertoli cells in vivo and in vitro. JOURNAL OF ANDROLOGY. 1992;13(6):560-70.

200. Buganim Y, Itskovich E, Hu Y-C, Cheng AW, Ganz K, Sarkar S, et al. Direct Reprogramming of Fibroblasts into Embryonic Sertoli-like Cells by Defined Factors. CELL STEM CELL. 2012;11(3):373-86.

201. Liang J, Wang N, He J, Du J, Guo Y, Li L, et al. Induction of Sertoli cells from human fibroblasts by NR5A1 and GATA4. BIORXIV. 2018.

202. Kjartansdóttir KR, Reda A, Panula S, Day K, Hultenby K, Söder O, et al. A Combination of Culture Conditions and Gene Expression Analysis Can Be Used to Investigate and Predict hES Cell Differentiation Potential towards Male Gonadal Cells. PLOS ONE. 2015;10(12):e0144029.

203. Shlush E, Maghen L, Swanson S, Kenigsberg S, Moskovtsev S, Barretto T, et al. In vitro generation of Sertoli-like and haploid spermatid-like cells from human umbilical cord perivascular cells. STEM CELL RESEARCH & THERAPY. 2017;8(1):37.

204. Hayashi K, Ohta H, Kurimoto K, Aramaki S, Saitou M. Reconstitution of the Mouse Germ Cell Specification Pathway in Culture by Pluripotent Stem Cells. CELL. 2011;146(4):519-32.

205. Hikabe O, Hamazaki N, Nagamatsu G, Obata Y, Hirao Y, Hamada N, et al. Reconstitution in vitro of the entire cycle of the mouse female germ line. NATURE. 2016;539(7628):299-303.

206. Irie N, Weinberger L, Tang WWC, Kobayashi T, Viukov S, Manor YS, et al. SOX17 Is a Critical Specifier of Human Primordial Germ Cell Fate. CELL. 2015;160(1-2):253-68.

207. Yamashiro C, Sasaki K, Yabuta Y, Kojima Y, Nakamura T, Okamoto I, et al. Generation of human oogonia from induced pluripotent stem cells in vitro. SCIENCE. 2018.

208. Sadedin SP, Dashnow H, James PA, Bahlo M, Bauer DC, Lonie A, et al. Cpipe: a shared variant detection pipeline designed for diagnostic settings. GENOME MED. 2015;7(1):68.

209. Scott DJ, Layfield S, Yan Y, Sudo S, Hsueh AJ, Tregear GW, et al. Characterization of novel splice variants of LGR7 and LGR8 reveals that receptor signaling is mediated by their unique low density lipoprotein class A modules. THE JOURNAL OF BIOLOGICAL CHEMISTRY. 2006;281(46):34942-54.

210. Gayevskiy V, Roscioli T, Dinger ME, Cowley MJ. Seave: a comprehensive web platform for storing and interrogating human genomic variation. BIOINFORMATICS. 2018:bty540-bty.

211. Venselaar H, Te Beek TA, Kuipers RK, Hekkelman ML, Vriend G. Protein structure analysis of mutations causing inheritable diseases. An e-Science approach with life scientist friendly interfaces. BMC BIOINFORMATICS. 2010;11:548.

212. Tannour-Louet M, Han S, Louet J-F, Zhang B, Romero K, Addai J, et al. Increased gene copy number of VAMP7 disrupts human male urogenital development through altered estrogen action. NATURE MEDICINE. 2014;20:715.

213. Croft B, Ohnesorg T, Sinclair AH. The Role of Copy Number Variants in Disorders of Sex Development. SEX DEV. 2018;12(1-3):19-29.

222 214. Shankara Narayana N, Kean AM, Ewans L, Ohnesorg T, Ayers KL, Watson G, et al. Painful ovulation in a 46,XX SRY -ve adult male with SOX9 duplication. ENDOCRINOLOGY, DIABETES & METABOLISM CASE REPORTs. 2017;2017.

215. Hyon C, Chantot-Bastaraud S, Harbuz R, Bhouri R, Perrot N, Peycelon M, et al. Refining the regulatory region upstream of SOX9 associated with 46,XX testicular disorders of Sex Development (DSD). AM J MED GENET A. 2015;167a(8):1851-8.

216. Xiao B, Ji X, Xing Y, Chen YW, Tao J. A rare case of 46, XX SRY-negative male with approximately 74-kb duplication in a region upstream of SOX9. EUR J MED GENET. 2013;56(12):695-8.

217. Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. GENET MED. 2015;17(5):405-24.

218. Beysen D, Moumne L, Veitia R, Peters H, Leroy BP, De Paepe A, et al. Missense mutations in the forkhead domain of FOXL2 lead to subcellular mislocalization, protein aggregation and impaired transactivation. HUM MOL GENET. 2008;17(13):2030-8.

219. Zhao L, Wang C, Lehman ML, He M, An J, Svingen T, et al. Transcriptomic analysis of mRNA expression and alternative splicing during mouse sex determination. MOL CELL ENDOCRINOL. 2018;478:84-96.

220. Jameson SA, Natarajan A, Cool J, DeFalco T, Maatouk DM, Mork L, et al. Temporal Transcriptional Profiling of Somatic and Germ Cells Reveals Biased Lineage Priming of Sexual Fate in the Fetal Mouse Gonad. PLOS GENET. 2012;8(3):e1002575.

221. Barker N, Tan S, Clevers H. Lgr proteins in epithelial stem cell biology. DEVELOPMENT. 2013;140(12):2484-94.

222. Vassart G, Pardo L, Costagliola S. A molecular dissection of the glycoprotein hormone receptors. TRENDS BIOCHEM SCI. 2004;29(3):119-26.

223. Kossack N, Troppmann B, Richter-Unruh A, Kleinau G, Gromoll J. Aberrant transcription of the LHCGR gene caused by a mutation in exon 6A leads to Leydig cell hypoplasia type II. MOL CELL ENDOCRINOL. 2013;366(1):59-67.

224. Rastetter RH, Bernard P, Palmer JS, Chassot A-A, Chen H, Western PS, et al. Marker genes identify three somatic cell types in the fetal mouse ovary. DEVELOPMENTAL BIOLOGY. 2014;394(2):242-52.

225. Carmon KS, Gong X, Lin Q, Thomas A, Liu Q. R-spondins function as ligands of the orphan receptors LGR4 and LGR5 to regulate Wnt/beta-catenin signaling. PROC NATL ACAD SCI U S A. 2011;108(28):11452-7.

226. Morita H, Mazerbourg S, Bouley DM, Luo C-W, Kawamura K, Kuwabara Y, et al. Neonatal lethality of LGR5 null mice is associated with ankyloglossia and gastrointestinal distension. MOLECULAR AND CELLULAR BIOLOGY. 2004;24(22):9736-43.

223 227. Snyder JC, Rochelle LK, Lyerly HK, Caron MG, Barak LS. Constitutive internalization of the leucine-rich G protein-coupled receptor-5 (LGR5) to the trans-Golgi network. THE JOURNAL OF BIOLOGICAL CHEMISTRY. 2013;288(15):10286-97.

228. Morgan RG, Molnar E, Jones RF, Collard TJ, Lane JD, Greenhough A, et al. Nutrient stress alters the glycosylation status of LGR5 resulting in reduced protein stability and membrane localisation in colorectal tumour cells: implications for targeting cancer stem cells. BR J CANCER. 2015;112(4):714-9.

229. Feng S, Ferlin A, Truong A, Bathgate R, Wade JD, Corbett S, et al. INSL3/RXFP2 signaling in testicular descent. ANNALS OF THE NEW YORK ACADEMY OF SCIENCES. 2009;1160(1):197- 204.

230. Gorlov IP. Mutations of the GREAT gene cause cryptorchidism. HUMAN MOLECULAR GENETICS. 2002;11(19):2309-18.

231. Johnston SE, McEwan JC, Pickering NK, Kijas JW, Beraldi D, Pilkington JG, et al. Genome- wide association mapping identifies the genetic basis of discrete and quantitative variation in sexual weaponry in a wild sheep population. MOLECULAR ECOLOGY. 2011;20(12):2555-66.

232. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, et al. Analysis of protein-coding genetic variation in 60,706 humans. NATURE. 2016;536(7616):285-91.

233. Cederroth CR, Pitetti J-L, Papaioannou MD, Nef S. Genetic programs that regulate testicular and ovarian development. MOLECULAR AND CELLULAR ENDOCRINOLOGY. 2007;265- 266:3-9.

234. Später D, Hill TP, Gruber M, Hartmann C. Role of canonical Wnt-signalling in joint formation. EUROPEAN CELLS & MATERIALS. 2006;12:71-80.

235. Später D, Hill TP, Sullivan RJ, Gruber M. Wnt9a signaling is required for joint integrity and regulation of Ihh during chondrogenesis. DEVELOPMENT. 2006;133(15):3039-49.

236. Waggoner DJ, Chow CK, Dowton SB, Watson MS. Partial monosomy of distal 10q: three new cases and a review. AM J MED GENET. 1999;86(1):1-5.

237. Piard J, Mignot B, Arbez-Gindre F, Aubert D, Morel Y, Roze V, et al. Severe sex differentiation disorder in a boy with a 3.8 Mb 10q25.3-q26.12 microdeletion encompassing EMX2. AM J MED GENET A. 2014;164A(10):2618-22.

238. Barsoum IB, Bingham NC, Parker KL, Jorgensen JS, Yao HH. Activation of the Hedgehog pathway in the mouse fetal ovary leads to ectopic appearance of fetal Leydig cells and female pseudohermaphroditism. DEV BIOL. 2009;329(1):96-103.

239. Haines B, Hughes J, Corbett M, Shaw M, Innes J, Patel L, et al. Interchromosomal insertional translocation at Xq26.3 alters SOX3 expression in an individual with XX male sex reversal. J CLIN ENDOCRINOL METAB. 2015;100(5):E815-20.

240. Niknafs N, Beleva-Guthrie V, Naiman DQ, Karchin R. SubClonal Hierarchy Inference from Somatic Mutations: Automatic Reconstruction of Cancer Evolutionary Trees from Multi-region Next Generation Sequencing. PLOS COMPUT BIOL. 2015;11(10):e1004416.

224 241. Mitchison HM, Schmidts M, Loges NT, Freshour J, Dritsoula A, Hirst RA, et al. Mutations in axonemal dynein assembly factor DNAAF3 cause primary ciliary dyskinesia. NAT GENET. 2012;44(4):381-9, S1-2.

242. Tsang M, Friesel R, Kudoh T, Dawid IB. Identification of Sef, a novel modulator of FGF signalling. NAT CELL BIOL. 2002;4(2):165-9.

243. Kovalenko D, Yang X, Nadeau RJ, Harkins LK, Friesel R. Sef inhibits fibroblast growth factor signaling by inhibiting FGFR1 tyrosine phosphorylation and subsequent ERK activation. THE JOURNAL OF BIOLOGICAL CHEMISTRY. 2003;278(16):14087-91.

244. Jordan BK, Mohammed M, Ching ST, Delot E, Chen XN, Dewing P, et al. Up-regulation of WNT-4 signaling and dosage-sensitive sex reversal in humans. AM J HUM GENET. 2001;68(5):1102-9.

245. Bray K, Gillette M, Young J, Loughran E, Hwang M, Sears JC, et al. Cdc42 overexpression induces hyperbranching in the developing mammary gland by enhancing cell migration. BREAST CANCER RES. 2013;15(5):R91.

246. Jordan BK, Shen JH, Olaso R, Ingraham HA, Vilain E. Wnt4 overexpression disrupts normal testicular vasculature and inhibits testosterone synthesis by repressing steroidogenic factor 1/beta-catenin synergy. PROC NATL ACAD SCI U S A. 2003;100(19):10866-71.

247. Kircher M, Witten DM, Jain P, O'Roak BJ, Cooper GM, Shendure J. A general framework for estimating the relative pathogenicity of human genetic variants. NAT GENET. 2014;46(3):310-5.

248. Petrovski S, Wang Q, Heinzen EL, Allen AS, Goldstein DB. Genic intolerance to functional variation and the interpretation of personal genomes. PLOS GENET. 2013;9(8):e1003709.

249. Jiang Y, Oliver P, Davies KE, Platt N. Identification and characterization of murine SCARA5, a novel class A scavenger receptor that is expressed by populations of epithelial cells. THE JOURNAL OF BIOLOGICAL CHEMISTRY. 2006;281(17):11834-45.

250. Karner CM, Merkel CE, Dodge M, Ma Z, Lu J, Chen C, et al. Tankyrase is necessary for canonical Wnt signaling during kidney development. DEV DYN. 2010;239(7):2014-23.

251. Morohashi KI, Omura T. Ad4BP/SF-1, a transcription factor essential for the transcription of steroidogenic cytochrome P450 genes and for the establishment of the reproductive function. THE FASEB JOURNAL. 1996;10(14):1569-77.

252. Val P, Lefrancois-Martinez AM, Veyssiere G, Martinez A. SF-1 a key player in the development and differentiation of steroidogenic tissues. NUCL RECEPT. 2003;1(1):8.

253. Lin L, Achermann JC. Steroidogenic factor-1 (SF-1, Ad4BP, NR5A1) and disorders of testis development. SEX DEV. 2008;2(4-5):200-9.

254. Ferraz-de-Souza B, Lin L, Achermann JC. Steroidogenic factor-1 (SF-1, NR5A1) and human disease. MOL CELL ENDOCRINOL. 2011;336(1-2):198-205.

255. Domenice S, Machado AZ, Ferreira FM, Ferraz-de-Souza B, Lerario AM, Lin L, et al. Wide spectrum of NR5A1-related phenotypes in 46,XY and 46,XX individuals. BIRTH DEFECTS RES C EMBRYO TODAY. 2016;108(4):309-20.

225 256. Takasawa K, Igarashi M, Ono M, Takemoto A, Takada S, Yamataka A, et al. Phenotypic Variation in 46,XX Disorders of Sex Development due to the NR5A1 p.R92W Variant: A Sibling Case Report and Literature Review. SEXUAL DEVELOPMENT. 2017;11(5-6):284-8.

257. Doghman M, Figueiredo BC, Volante M, Papotti M, Lalli E. Integrative analysis of SF-1 transcription factor dosage impact on genome-wide binding and gene expression regulation. NUCLEIC ACIDS RES. 2013;41(19):8896-907.

258. Roy A, Kucukural A, Zhang Y. I-TASSER: a unified platform for automated protein structure and function prediction. NAT PROTOC. 2010;5(4):725-38.

259. Zhang Y. I-TASSER server for protein 3D structure prediction. BMC BIOINFORMATICS. 2008;9:40.

260. Chan AO, But WM, Lee CY, Lam YY, Ng KL, Loung PY, et al. Aetiological bases of 46,XY disorders of sex development in the Hong Kong Chinese population. HONG KONG MED J. 2015;21(6):499-510.

261. Kohler B, Lin L, Ferraz-de-Souza B, Wieacker P, Heidemann P, Schroder V, et al. Five novel mutations in steroidogenic factor 1 (SF1, NR5A1) in 46,XY patients with severe underandrogenization but without adrenal insufficiency. HUM MUTAT. 2008;29(1):59-64.

262. Mizusaki H, Kawabe K, Mukai T, Ariyoshi E, Kasahara M, Yoshioka H, et al. Dax-1 (dosage- sensitive sex reversal-adrenal hypoplasia congenita critical region on the X chromosome, gene 1) gene transcription is regulated by wnt4 in the female developing gonad. MOLECULAR ENDOCRINOLOGy. 2003;17(4):507-19.

263. Camats N, Fernandez-Cancio M, Audi L, Schaller A, Fluck CE. Broad phenotypes in heterozygous NR5A1 46,XY patients with a disorder of sex development: an oligogenic origin? EUR J HUM GENET. 2018;26(9):1329-1338.

264. Vidal VP, Chaboissier MC, de Rooij DG, Schedl A. Sox9 induces testis development in XX transgenic mice. NAT GENET. 2001;28(3):216-7.

265. Miyado M, Inui M, Igarashi M, Katoh-Fukui Y, Takasawa K, Hakoda A, et al. The p.R92W variant of NR5A1/Nr5a1 induces testicular development of 46,XX gonads in humans, but not in mice: phenotypic comparison of human patients and mutation-induced mice. BIOLOGY OF SEX DIFFERENCES. 2016;7(1):56.

266. Swartz JM, Ciarlo R, Guo MH, Abrha A, Weaver B, Diamond DA, et al. A 46,XX Ovotesticular Disorder of Sex Development Likely Caused by a Steroidogenic Factor-1 (NR5A1) Variant. HORMONE RESEARCH IN PAEDIATRICS. 2016.

267. Guran T, Buonocore F, Saka N, Ozbek MN, Aycan Z, Bereket A, et al. Rare Causes of Primary Adrenal Insufficiency: Genetic and Clinical Characterization of a Large Nationwide Cohort. J CLIN ENDOCRINOL METAB. 2016;101(1):284-92.

268. Lin L, Philibert P, Ferraz-de-Souza B, Kelberman D, Homfray T, Albanese A, et al. Heterozygous missense mutations in steroidogenic factor 1 (SF1/Ad4BP, NR5A1) are associated with 46,XY disorders of sex development with normal adrenal function. J CLIN ENDOCRINOL METAB. 2007;92(3):991-9.

226 269. Yang L, Soonpaa MH, Adler ED, Roepke TK, Kattman SJ, Kennedy M, et al. Human cardiovascular progenitor cells develop from a KDR+ embryonic-stem-cell-derived population. NATURE. 2008;453(7194):524-8.

270. Wang D, Haviland DL, Burns AR, Zsigmond E, Wetsel RA. A pure population of lung alveolar epithelial type II cells derived from human embryonic stem cells. PROC NATL ACAD SCI U S A. 2007;104(11):4449-54.

271. Pagliuca FW, Millman JR, Gurtler M, Segel M, Van Dervort A, Ryu JH, et al. Generation of functional human pancreatic beta cells in vitro. CELL. 2014;159(2):428-39.

272. McCracken KW, Howell JC, Wells JM, Spence JR. Generating human intestinal tissue from pluripotent stem cells in vitro. NAT PROTOC. 2011;6(12):1920-8.

273. Takasato M, Er PX, Becroft M, Vanslambrouck JM, Stanley EG, Elefanty AG, et al. Directing human embryonic stem cell differentiation towards a renal lineage generates a self- organizing kidney. NATURE CELL BIOLOGY. 2014;16(1):118-26.

274. Gadue P, Huber TL, Paddison PJ, Keller GM. Wnt and TGF-beta signaling are required for the induction of an in vitro model of primitive streak formation using embryonic stem cells. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. 2006;103(45):16806-11.

275. Takasato M, Er PX, Chiu HS, Maier B, Baillie GJ, Ferguson C, et al. Kidney organoids from human iPS cells contain multiple lineages and model human nephrogenesis. NATURE. 2015;526(7574):564-8.

276. Sekido R, Lovell-Badge R. Mechanisms of gonadal morphogenesis are not conserved between chick and mouse. DEVELOPMENTAL BIOLOGY. 2007;302(1):132-42.

277. Kao T, Labonne T, Niclis JC, Chaurasia R, Lokmic Z, Qian E, et al. GAPTrap: A Simple Expression System for Pluripotent Stem Cells and Their Derivatives. STEM CELL REPORTS. 2016;7(3):518-26.

278. Knower KC, Sim H, McClive PJ, Bowles J, Koopman P, Sinclair AH, et al. Characterisation of urogenital ridge gene expression in the human embryonal carcinoma cell line NT2/D1. SEXUAL DEVELOPMENT. 2007;1(2):114-26.

279. Schumacher V, Gueler B, Looijenga LHJ, Becker JU, Amann K, Engers R, et al. Characteristics of testicular dysgenesis syndrome and decreased expression of SRY and SOX9 in Frasier syndrome. MOLECULAR REPRODUCTION AND DEVELOPMENT. 2008;75(9):1484-94.

280. Escalante-Alcalde D, Merchant-Larios H. Somatic and germ cell interactions during histogenetic aggregation of mouse fetal testes. EXPERIMENTAL CELL RESEARCH. 1992;198(1):150-8.

281. Svingen T, Koopman P. Building the mammalian testis: origins, differentiation, and assembly of the component cell populations. GENES & DEVELOPMENT. 2013;27(22):2409-26.

282. Del Valle I, Buonocore F, Duncan AJ, Lin L, Barenco M, Parnaik R, et al. A genomic atlas of human adrenal and gonad development. WELLCOME OPEN RES. 2017;2:25.

227 283. Gierl MS, Gruhn WH, von Seggern A, Maltry N, Niehrs C. GADD45G Functions in Male Sex Determination by Promoting p38 Signaling and Sry Expression. DEVELOPMENTAL CELL. 2012;23(5):1032-42.

284. Warr N, Carre G-A, Siggers P, Faleato JV, Brixey R, Pope M, et al. Gadd45γ and Map3k4 Interactions Regulate Mouse Testis Determination via p38 MAPK-Mediated Control of Sry Expression. DEVELOPMENTAL CELL. 2012;23(5):1020-31.

285. Meeks JJ, Crawford SE, Russell TA, Morohashi K, Weiss J, Jameson JL. Dax1 regulates testis cord organization during gonadal differentiation. DEVelopment. 2003;130(5):1029-36.

286. Behringer RR, Finegold MJ, Cate RL. Mullerian-inhibiting substance function during mammalian sexual development. CELL. 1994;79(3):415-25.

287. Bitgood MJ, Shen L, McMahon AP. Sertoli cell signaling by Desert hedgehog regulates the male germline. CURR BIOL. 1996;6(3):298-304.

288. Barrionuevo F, Bagheri-Fam S, Klattig J, Kist R, Taketo MM, Englert C, et al. Homozygous inactivation of Sox9 causes complete XY sex reversal in mice. BIOL REPROD. 2006;74(1):195-201.

289. Caron KM, Soo SC, Wetsel WC, Stocco DM, Clark BJ, Parker KL. Targeted disruption of the mouse gene encoding steroidogenic acute regulatory protein provides insights into congenital lipoid adrenal hyperplasia. PROC NATL ACAD SCI U S A. 1997;94(21):11540-5.

290. Lindstrom NO, Guo JJ, Kim AD, Tran T, Guo QY, Brandine GD, et al. Conserved and Divergent Features of Mesenchymal Progenitor Cell Types within the Cortical Nephrogenic Niche of the Human and Mouse Kidney. J AM SOC NEPHROL. 2018;29(3):806-24.

291. Combes AN, Phipson B, Zappia L, Lawlor K, Er PX, Oshlack A, et al. High throughput single cell RNA-seq of developing mouse kidney and human kidney organoids reveals a roadmap for recreating the kidney. BIORXIV. 2017.

292. Belle M, Godefroy D, Couly G, Malone SA, Collier F, Giacobini P, et al. Tridimensional Visualization and Analysis of Early Human Development. CELL. 2017;169(1):161-73.e12.

293. Dolle P, Duboule D. Two gene members of the murine HOX-5 complex show regional and cell-type specific expression in developing limbs and gonads. EMBO J. 1989;8(5):1507-15.

294. Gaunt SJ, Coletta PL, Pravtcheva D, Sharpe PT. Mouse Hox-3.4: homeobox sequence and embryonic expression patterns compared with other members of the Hox gene network. DEVELOPMENT. 1990;109(2):329-39.

295. Spirov AV, Borovsky M, Spirova OA. HOX Pro DB: the functional genomics of hox ensembles. NUCLEIC ACIDS RES. 2002;30(1):351-3.

296. Zubair M, Ishihara S, Oka S, Okumura K, Morohashi K. Two-step regulation of Ad4BP/SF- 1 gene transcription during fetal adrenal development: initiation by a Hox-Pbx1-Prep1 complex and maintenance via autoregulation by Ad4BP/SF-1. MOL CELL BIOL. 2006;26(11):4111-21.

297. Svingen T, Koopman P. Involvement of homeobox genes in mammalian sexual development. SEX DEV. 2007;1(1):12-23.

228 298. Takasato M, Little MH. A strategy for generating kidney organoids: Recapitulating the development in human pluripotent stem cells. DEVELOPMENTAL BIOLOGY. 2016;420(2):210- 220.

299. Takasato M, Little MH. The origin of the mammalian kidney: implications for recreating the kidney in vitro. DEVELOPMENT. 2015;142(11):1937-47.

300. Yoshino T, Murai H, Saito D. Hedgehog-BMP signalling establishes dorsoventral patterning in lateral plate mesoderm to trigger gonadogenesis in chicken embryos. NATURE COMMUNICATIONS. 2016;7:12561.

301. Schmahl J, Kim Y, Colvin JS, Ornitz DM, Capel B. Fgf9 induces proliferation and nuclear localization of FGFR2 in Sertoli precursors during male sex determination. DEVELOPMENT. 2004;131(15):3627-36.

302. Colvin JS, Green RP, Schmahl J, Capel B, Ornitz DM. Male-to-female sex reversal in mice lacking fibroblast growth factor 9. CELL. 2001;104(6):875-89.

303. Sandell LL, Sanderson BW, Moiseyev G, Johnson T, Mushegian A, Young K, et al. RDH10 is essential for synthesis of embryonic retinoic acid and is required for limb, craniofacial, and organ development. GENE DEV. 2007;21(9):1113-24.

304. Archambeault DR, Yao HH-C. Activin A, a product of fetal Leydig cells, is a unique paracrine regulator of Sertoli cell proliferation and fetal testis cord expansion. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. 2010;107(23):10526-31.

305. Gustin SE, Stringer JM, Hogg K, Sinclair AH, Western PS. FGF9, activin and TGFbeta promote testicular characteristics in an XX gonad organ culture model. REPRODUCTION. 2016;152(5):529-43.

306. Zheng B, Yu J, Guo Y, Gao T, Shen C, Zhang X, et al. Cellular nucleic acid-binding protein is vital to testis development and spermatogenesis in mice. REPRODUCTION. 2018;156(1):59- 69.

307. Lucas TF, Nascimento AR, Pisolato R, Pimenta MT, Lazari MF, Porto CS. Receptors and signaling pathways involved in proliferation and differentiation of Sertoli cells. SPERMATOGENESIS. 2014;4:e28138.

308. Shlush E, Maghen L, Swanson S, Kenigsberg S, Moskovtsev S, Barretto T, et al. In vitro generation of Sertoli-like and haploid spermatid-like cells from human umbilical cord perivascular cells. STEM CELL RESEARCH & THERAPY. 2017;8(1):37.

309. Davidoff MS, Middendorff R, Koeva Y, Pusch W, Jezek D, Muller D. Glial cell line-derived neurotrophic factor (GDNF) and its receptors GFRalpha-1 and GFRalpha-2 in the human testis. ITAL J ANAT EMBRYOl. 2001;106(2 Suppl 2):173-80.

310. Skinner MK, Tung PS, Fritz IB. Cooperativity between Sertoli cells and testicular peritubular cells in the production and deposition of extracellular matrix components. J CELL BIOL. 1985;100(6):1941-7.

229 311. Wilhelm D, Martinson F, Bradford S, Wilson MJ, Combes AN, Beverdam A, et al. Sertoli cell differentiation is induced both cell-autonomously and through prostaglandin signaling during mammalian sex determination. DEV BIOL. 2005;287(1):111-24.

312. Uzumcu M, Westfall SD, Dirks KA, Skinner MK. Embryonic testis cord formation and mesonephric cell migration requires the phosphotidylinositol 3-kinase signaling pathway. BIOL REPROD. 2002;67(6):1927-35.

313. Cupp AS, Kim G, Skinner MK. Expression and action of transforming growth factor beta (TGFbeta1, TGFbeta2, and TGFbeta3) during embryonic rat testis development. BIOL REPROD. 1999;60(6):1304-13.

314. Li H, Kim KH. Retinoic acid inhibits rat XY gonad development by blocking mesonephric cell migration and decreasing the number of gonocytes. BIOL REPROD. 2004;70(3):687-93.

315. Harris A, Siggers P, Corrochano S, Warr N, Sagar D, Grimes DT, et al. ZNRF3 functions in mammalian sex determination by inhibiting canonical WNT signaling. PROC NATL ACAD SCI U S A. 2018;115(21):5474-9.

316. Narita T, Ishii Y, Nohno T, Noji S, Yasugi S. Sonic hedgehog expression in developing chicken digestive organs is regulated by epithelial–mesenchymal interactions. DEVELOPMENT, GROWTH & DIFFERENTIATION. 1998;40(1):67-74.

317. Zhang XM, Ramalho-Santos M, McMahon AP. Smoothened Mutants Reveal Redundant Roles for Shh and Ihh Signaling Including Regulation of L/R Asymmetry by the Mouse Node. CELL. 2001;105(6):781-92.

318. Duester G. Retinoic acid synthesis and signaling during early organogenesis. CELL. 2008;134(6):921-31.

319. Bollig F, Perner B, Besenbeck B, Kothe S, Ebert C, Taudien S, et al. A highly conserved retinoic acid responsive element controls wt1a expression in the zebrafish pronephros. DEVELOPMENT. 2009;136(17):2883-92.

320. Geens M, Sermon KD, Van de Velde H, Tournaye H. Sertoli cell-conditioned medium induces germ cell differentiation in human embryonic stem cells. J ASSIST REPROD GENET. 2011;28(5):471-80.

321. Adams IR, McLaren A. Sexually dimorphic development of mouse primordial germ cells: switching from oogenesis to spermatogenesis. DEVELOPMENT. 2002;129(5):1155-64.

322. Suntharalingham JP, Buonocore F, Duncan AJ, Achermann JC. DAX-1 (NR0B1) and steroidogenic factor-1 (SF-1, NR5A1) in human disease. BEST PRACT RES CLIN ENDOCRINOL METAB. 2015;29(4):607-19.

323. Bott RC, Clopton DT, Cupp AS. A proposed role for VEGF isoforms in sex-specific vasculature development in the gonad. REPROD DOMEST ANIM. 2008;43 Suppl 2:310-6.

324. Cool J, DeFalco TJ, Capel B. Vascular-mesenchymal cross-talk through Vegf and Pdgf drives organ patterning. PROC NATL ACAD SCI U S A. 2011;108(1):167-72.

230 325. Przepiorski A, Sander V, Tran T, Hollywood JA, Sorrenson B, Shih JH, et al. A Simple Bioreactor-Based Method to Generate Kidney Organoids from Pluripotent Stem Cells. STEM CELL REPORTS. 2018;11(2):470-84.

326. Sasaki K, Yokobayashi S, Nakamura T, Okamoto I, Yabuta Y, Kurimoto K, et al. Robust In Vitro Induction of Human Germ Cell Fate from Pluripotent Stem Cells. CELL STEM CELL. 2015;17(2):178-94.

327. Kalia SS, Adelman K, Bale SJ, Chung WK, Eng C, Evans JP, et al. Recommendations for reporting of secondary findings in clinical exome and genome sequencing, 2016 update (ACMG SF v2.0): a policy statement of the American College of Medical Genetics and Genomics (vol 19, pg 249, 2016). GENET MED. 2017;19(4):484-5.

328. Gasperini M, Starita L, Shendure J. The power of multiplexed functional analysis of genetic variants. NATURE PROTOCOLS. 2016;11(10):1782-7.

329. Starita LM, Ahituv N, Dunham MJ, Kitzman JO, Roth FP, Seelig G, et al. Variant Interpretation: Functional Assays to the Rescue. AMERICAN JOURNAL OF HUMAN GENETICS. 2017;101(3):315-25.

330. Mazen I, Abdel-Hamid M, Mekkawy M, Bignon-Topalovic J, Boudjenah R, El Gammal M, et al. Identification of NR5A1 Mutations and Possible Digenic Inheritance in 46, XY Gonadal Dysgenesis. SEXUAL DEVELOPMENT. 2016;10(3):147-51.

331. Titmarsh DM, Glass NR, Mills RJ, Hidalgo A, Wolvetang EJ, Porrello ER, et al. Induction of Human iPSC-Derived Cardiomyocyte Proliferation Revealed by Combinatorial Screening in High Density Microbioreactor Arrays. SCIENTIFIC REPORTS. 2016;6.

332. Buehr M, McLaren A, Bartley A, Darling S. Proliferation and migration of primordial germ cells in We/We mouse embryos. DEV DYN. 1993;198(3):182-9.

333. Merchant H. Rat Gonadal and Ovarian Organogenesis with and without Germ-Cells - Ultrastructural Study. DEVELOPMENTAL BIOLOGY. 1975;44(1):1-21.

334. Pellas TC, Ramachandran B, Duncan M, Pan SS, Marone M, Chada K. Germ-Cell Deficient (Gcd), an Insertional Mutation Manifested as Infertility in Transgenic Mice. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. 1991;88(19):8787- 91.

335. Cool J, Carmona FD, Szucsik JC, Capel B. Peritubular myoid cells are not the migrating population required for testis cord formation in the XY gonad. SEX DEV. 2008;2(3):128-33.

336. Roost MS, van Iperen L, Ariyurek Y, Buermans HP, Arindrarto W, Devalla HD, et al. KeyGenes, a Tool to Probe Tissue Differentiation Using a Human Fetal Transcriptional Atlas. STEM CELL REPORTS. 2015;4(6):1112-24.

231

232 Appendices

Appendix 1 – Primers for patient variant analysis.

Name Sequence (5'  3') Species Experiment

hFOXL2_ex1_fwd CAGCGCCTGGAGCGGAGAG Human Sanger sequencing

hFOXL2_ex1_rev CTTGCCGGGCTGGAAGTGC Human Sanger sequencing

Mutagenesis LGR5_c.C8T_for GCGGGGACAGAGATGCCGGTTGCCG Human (QuickChange), cDNA

Mutagenesis LGR5_c.C8T_rev CGGCAACCGGCATCTCTGTCCCCGC Human (QuickChange), cDNA

Mutagenesis LGR5_c.C8T_for_v2 CCGAGCCGGGAGATGTCCATGGTGC Human (QuickChange), cDNA

Mutagenesis LGR5_c.C8T_rev_v2 GCACCATGGACATCTCCCGGCTCGG Human (QuickChange), cDNA

For sequencing LGR5.cDNAseq.f1 AAGACACGTACCCACAGA Human cloned cDNA

For sequencing LGR5.cDNAseq.f2 GCAACCCTTCTCTTATTAC Human cloned cDNA

For sequencing LGR5.cDNAseq.f3 TGCCTTATGCTTACCAGT Human cloned cDNA

For sequencing LGR5.cDNAseq.f4 TTGCTCAATTCCCTTTGCTT Human cloned cDNA

RXFP2.c.2090_2093del.f AGCTTCTGTGCTCTTCCAACA Human Sanger sequencing

RXFP2.c.2090_2093del.r GTGACAGGAGCCATTCTGCT Human Sanger sequencing

Sanger sequencing, hWNT9A_ex4_For TGCACAGCCCTCACACTG Human cloning

Sanger sequencing, hWNT9A_ex4_Rev GCAGGTGTAGACCCTTCACA Human cloning

233 Appendix 2 – CRISPR guide RNAs for mutant mouse generation.

Wnt9a

crWnt9a-5 ATAAGGTGGGGAGAGCCGTC TGG

crWnt9a-3 GGACCCTACCTGGTACTGCG TGG

Emx2

crRNA sequence GGGGTGTGGCGAATGCGAGG AGG (- strand) Repair oligo ACTCCAACCCGGACTTGGTGTTCGCCGAGGCGGTCTCGCACCCGCCCAACCCCGCCGTGC (- strand) CGGTGCACCCGGTGCCGCCGCCGCACGTCCTGGCCGCCCACCCCCTGCCGTCCTCGCATT CGCCACACCCCCTCTTCGCCTCGCAGCAGCGGGACCCGTCCACCTTCTACCCCTGGCTCAT CCACCGCTACCGATATCTG PAM (CCT; + strand) mutated to CGT

Appendix 3 – Gene lists used to filter WGS SNV data. The first list (Diagnostic genes) contains all diagnostic genes for DSD phenotypes (n = 64, annotated by Reproductive Development group). The second list (Candidate XX genes) contains all candidate genes for 46,XX (ovo)testicular DSDs based on mouse models and relevant signalling pathways (n = 75, annotated by myself). The third list (Candidate POI genes) contains candidate genes for POI (n = 94, annotated by Dr. Elena Tucker).

Diagnostic genes (n=64) Candidate XX genes (n=75) Candidate POI genes (n=94)

BMP15 AKR1C1 ABCB7

CBX2 APC AFF2

DHH ATP6AP2 AIFM1

DMRT1 AXIN1 AKT1

DMRT2 AXIN2 ALDH1A1

FOXL2 BMP2 ALDH1A2

GATA4 CSNK1A1 ALDH1A3

NR0B1 CTNNB1 ANKRD22

NR5A1 CYB5A ATM

MAP3K1 DHCR7 BBS9

RSPO1 DMRT3 BCORL1

234 SOX3 DVL1 BLM

SOX9 ELK1 BPESC1

SRY EMX1 BRSK1

TSPYL1 EMX2 C3orf38

WNT4 ESR2 CCBE1

WT1 FAM58A CDKN1B

ZFPM2 FBLN2 CENPI

AKR1C2 FGF2 CGGBP1

AKR1C4 FGF9 CHM

AMH FKBP4 CITED2

AMHR2 FST CNOT6

AR FZD1 CPEB1

ARX FZD2 CTNNA3

ATRX GADD45G CXCL12

CDKN1C GPC5 CYP26B1

CYB5A HHAT DACH2

CYP11A1 HSD17B1 DIAPH2

CYP11B1 HSD3B1 DNAH5

CYP17A1 IGF1R DNAJC8

CYP19A1 INSR DUSP22

CYP21A2 IRF2BPL EIF2B2

FGFR2 KDM3A EIF4B

HSD17B3 KISS1 EIF5B

HSD17B4 LEF1 EPB41L5

HSD3B2 LEPR ERAL1

LHCGR LGR4 ESR1

NR3C1 LGR5 EYA3

POR LHX1 FIGLA

235 SRD5A2 LHX4 FOXE1

STAR LHX9 FOXO1

BBS9 LRP5 FOXO3

CHD7 LRP6 FOXO4

FGF8 MACF1 GALT

FGFR1 MAP3K4 GDF9

FSHB MRPS22 GPR3

FSHR MSX1 HAAO

GNRH1 NOBOX HDX

GNRHR NOTUM HK3

HESX1 NR2F2 HNF1B

KAL1 NR5A2 HSD17B4

KISS1R NSMF INHBA

LEP PBX1 INHBB

LHX3 PDGFB LAMC1

PROK2 PSMC3IP LHX8

PROKR2 RUNX1 LMNA

PROP1 SEMA3A MAGT1

TAC3 SIX1 NAIP

WDR11 SIX4 NANOS3

ATF3 SOX10 NBN

HOXA13 SOX17 NGF

INSL3 SOX2 NUPR1

MAMLD1 SOX7 NXF2B

RXFP2 SOX8 NXF3

SOX8 NXF5

SRD5A1 PCDH19

TAX1BP3 PGRMC1

236 TBX2 PLP1

TCF21 PMAIP1

TLE3 PMM2

VNN1 POF1B

WNT2B POLG

WNT9A POU5F1

WWOX PRKX

ZNF280B PSMC3IP

PTPN4

RALB

RECQL4

RPA2

SKP2

SMPDL3B

STAMBPL1

STS

SYCE1

THRA

TSPAN7

UPRT

USP9X

UTP14A

VCX

WRN

XPNPEP2

ZFX

ZNF654

237 Appendix 4 – CRISPR targeting for Wnt9a mouse mutant. Schematic showing the genomic region of Wnt9a, the CRSIPR guide RNAs (crRNAs) were targeted to exons 2 and 3 of this gene.

238 Appendix 5 – Sanger sequencing of CRISPR knock-in Emx2 mutation.

239 Appendix 6 – Schism identifies a 41 kb deletion starting in the coding region of DNAAF3 in DSD04.

240 Appendix 7 – Mutagenesis and Cloning Primers for NR5A1 variant analysis.

Name Sequence (5' -> 3') Species Experiment

Site-directed NR5A1_c.C274T_ex4_For GCCCAAACTTGTTCCAGCCACCCCTCATACG Human mutagenesis

Site-directed Human NR5A1_c.C274T_ex4_Rev CGTATGAGGGGTGGCTGGAACAAGTTTGGGC mutagenesis

Site-directed NR5A1_c.C779T_p.A260V_For GGCCGAAGGCCACCGGCTGGTCG Human mutagenesis

Site-directed NR5A1_c.C779T_p.A260V_Rev CGACCAGCCGGTGGCCTTCGGCC Human mutagenesis

Cloning, hDAX1_XhoI_For TTTCTCGAGCCTATTGGATACTATTACCTGGG Human luciferase assay

Cloning, hDAX1_HindIII_Rev TTTAAGCTTGCATGTTGTAGAGGATGCTG Human luciferase assay

Appendix 8 – qRT-PCR primer sequences.

Primer Primer sequence Species

hACTG1.RT.Fow CCGAGCCGTGTTTCCTTCC Human

hACTG1.RT.Rev GCCATGCTCAATGGGGTACT Human

hANXA2.RT.Fow TCTACTGTTCACGAAATCCTGTG Human

hANXA2.RT.Rev AGTATAGGCTTTGACAGACCCAT Human

hARHGAP36.RT.Fow GTTGCTTCTGTCAATGTGGTCCG Human

hARHGAP36.RT.Rev GACTTCCACACGCGCTTAGCAA Human

hDAX1.RT.Fow CAAGGAGTACGCCTACCTCA Human

hDAX1.RT.Rev GCGTCATCCTGGTGTGTTC Human

hDHH.RT.Fow AACCCCGACATCATCTTCAA Human

hDHH.RT.Rev ACATGTTCATCACGGCAATG Human

hEMX2.RT.Fow CTCAGCCTCACGGAAACTCA Human

hEMX2.RT.Rev TTGCGAATCTGAGCCTTCTT Human

hEYA1.RT.Fow GTAGTGAATCCCCCAGTGGC Human

241 hEYA1.RT.Rev TGGTCGTGGGCTGAAACTAC Human hFGF9.RT.Fow GTGGACTCTACCTCGGGATG Human hFGF9.RT.Rev CCAGTTTTCTTCGAACTGTTCTC Human hFOXF1.RT.Fow CGTATCTGCACCAGAACAGC Human hFOXF1.RT.Rev GACAAACTCCTTTCGGTCACA Human hGAPDH.RT.Fow AGCCACATCGCTCAGACAC Human hGAPDH.RT.Rev GCCCAATACGACCAAATCC Human hGATA4.RT.Fow CTGTCATCTCACTACGGGCA Human hGATA4.RT.Rev GGGAGACGCATAGCCTTGT Human hHOXC5.RT.Fow CTAAGAGCAGTGGGGAGATCA Human hHOXC5.RT.Rev GTCATCCACGGGTAAATCTGTG Human hHOXC6.RT.Fow CCTTTTATTCGCCACAGGAGAA Human hHOXC6.RT.Rev TGCAGTTTGAGAGCATGTCTTT Human hHOXD9.RT.Fow GGACTCGCTTATAGGCCATGA Human hHOXD9.RT.Rev GCAAAACTACACGAGGCGAA Human hHSD3B1.RT.Fow CCTTCGGACCAGAATTGAGA Human hHSD3B1.RT.Rev ATACAGGCGGTGTGGATGAT Human hHSD3B2.RT.Fow CTTGGACAAGGCCTTCAGAC Human hHSD3B2.RT.Rev GGCTCATCCAGAATGTCTCC Human hLHX1.RT.Fow ATGCAACCTGACCGAGAAGT Human hLHX1.RT.Rev CAGGTCGCTAGGGGAGATG Human hMIXL1.RT.Fow GGTACCCCGACATCCACTT Human hMIXL1.RT.Rev GCCTGTTCTGGAACCATACCT Human hNPHS2.RT.Fow ACCAAATCCTCCGGCTTAGG Human hNPHS2.RT.Rev CAACCTTTACGCAGAACCAGA Human hNR5A1.RT.Fow CATCATCCTCTTCAGCCTGG Human hNR5A1.RT.Rev TGGCACAGGGTGTAGTCAAG Human hOCT4.RT.Fow AGCAAAACCCGGAGGAGT Human

242 hOCT4.RT.Rev CCACATCGGCCTGTGTATATC Human hPAX2.RT.Fow GCAACCCCGCCTTACTAAT Human hPAX2.RT.Rev AACTAGTGGCGGTCATAGGC Human hRPL21.RT.Fow TAAGCACTCTAAGAGCCGAGAT Human hRPL21.RT.Rev GCGCTTTAGTTGAACCCAGGTA Human hRPL29.RT.Fow CAGTCCCGAAAATGGCACAGA Human hRPL29.RT.Rev GGCTTTACGAGGGCCTTGATA Human hSIX1.RT.Fow GACTCCGGTTTTCGCCTTTG Human hSIX1.RT.Rev CACTTGCTCCTGCGTAAAGC Human hSOX17.RT.Fow ACGCCGAGTTGAGCAAGA Human hSOX17.RT.Rev TCTGCCTCCTCCACGAAG Human hSOX9.RT.Fow CCGAAAGCGGAGCTCGAAAC Human hSOX9.RT.Rev AGTTTCCGGGGTTGAAACTGG Human hSTAR.RT.Fow TAGCGACATTCAAGCTGTGC Human hSTAR.RT.Rev GTTCAGCTCCTGGCTGATG Human hSULT2A1.RT.Fow GGTTTGACCACATTCATGGCTGG Human hSULT2A1.RT.Rev CGGGTTCTAACGTCTTTCCCAG Human hT.RT.Fow AGGTACCCAACCCTGAGGA Human hT.RT.Rev GCAGGTGAGTTGTCAGAATAGGT Human hWT1.RT.Fow GAAATGGACAGAAGGGCAGA Human hWT1.RT.Rev GACACCGTGCGTGTGTATTC Human hWT1.RT1.Fow TGTCAGCGAAAGTTCTCCCG Human hWT1.RT1.Rev GCTGAAGGGCTTTTCACCTG Human hZFPM2.RT.Fow CTTGGCAAGGAGTGGAAGAC Human hZFPM2.RT.Rev TCTTCACCCTCAGAGATGGC Human

243 Appendix 9 – Markers for characterisation of gonad differentiation.

Gene Cell type marker Reference

OCT4 Pluripotency (273)

T Posterior primitive streak (273)

MIXL1 Posterior primitive streak (273)

SOX17 Anterior primitive streak (273)

LHX1 Intermediate mesoderm (273)

Intermediate PAX2 mesoderm/Kidney (273)

FOXF1 Lateral plate mesoderm (273)

NPHS2 Embryonic kidney (273)

SULT2A1 Embryonic adrenal (336)

ARHGAP36 Embryonic adrenal (336)

ACTG1 Chondrocyte progenitor Shireen Lamande, data unpublished

ANXA2 Chondrocyte progenitor Shireen Lamande, data unpublished

HOXC5 Hox gene (293-295)

HOXC6 Hox gene (293-295)

HOXD9 Hox gene (293-295)

244 Appendix 10 – 4 days of CHIR treatment promotes gonad marker induction in a female iPSC line. qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad after 7 days of monolayer differentiation with 4 days of CHIR treatment. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. CRL1502.3 iPSCs grown in APEL media on Matrigel treated with 4 μM CHIR for 4 days followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml) treatment.

1000

s 800 iPSC

C

S 4 day CHIR P 600

i

o

t 400

e

v

i t 200

a

l

e

r

n 10

o

i

s

s 8

e

r

p

x 6

e

e 4

n

e

G 2 0 9 1 G 1 4 2 1 2 2 2 9 X A 5 T A M X X B L X H 5 4 W T P A M 3 X O L R D A F D E D O S N D G Z S F A H G

245 Appendix 11 – Lower concentration of CHIR leads to gonad marker induction in the C32 hESC line. Figure provided by Dr. Minoru Takasato (Riken, Japan).

246 Appendix 12 – 4 μM CHIR promotes gonad marker induction in a male iPSC line. qRT-PCR data showing relative gene expression levels of markers of the bipotential gonad after 7 days of monolayer differentiation with 3, 4 or 5 μM CHIR. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown in APEL media on Matrigel treated with 4 days CHIR at 3, 4 or 5 μM followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml) treatment.

200

iPSC s 150

C

S 3uM CHIR

P

i 4uM CHIR o 100

t

e 5uM CHIR

v

i

t

a 50

l

e

r

n

o 20

i

s

s

e

r 15

p

x

e

e 10

n

e

G 5

0 9 1 G 1 4 2 1 2 2 X A 5 T A M B X B H 5 4 W T P 0 M 3 L R D A F R E D N D G Z N S A H G

Appendix 13 – 4 μM CHIR promotes gonad marker induction in the female iPSC line. qRT- PCR data showing relative gene expression levels of markers of the bipotential gonad after 7 days of monolayer differentiation with 3, 4 or 5 μM CHIR. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. CRL1502.3 iPSCs grown in APEL media on Matrigel treated with 4 days CHIR at 3, 4 or 5 μM followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml) treatment.

18000 iSPC 16000 3 uM CHIR

s

C 14000

S 4 uM CHIR

P

i 5 uM CHIR

o

t 12000

e 150

v

i

t

a

l

e 100

r

n

o

i

s 50

s

e

r

p

x 10

e

e

n

e

G 5

0 9 1 G 4 2 1 2 2 1 X A 5 A M B X B T H 5 4 T P 0 M 3 W L R D A F R E D N D G Z N S A H G

247 Appendix 14 – Shifting differentiation towards the lateral plate mesoderm favours induction of gonadal lineages in a female iPSC line. qRT-PCR data showing relative gene expression levels of markers of the gonad after 7 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. CRL1502.3 iPSCs grown in APEL2 media (with 4% PFHM-II) on Matrigel treated with 4 days of CHIR at 4 μM followed by FGF9 (200 ng/ml) and Heparin (1 μg/ml), with or without addition of BMP4 (at 5 or 50 ng/ml).

2000 iPSC FGF9 1500

s FGF9 + 5 BMP4

C

S FGF9 + 50 BMP4

P

i 1000

o

t

e

v

i 500

t

a

l

e

r

n

o

i 15

s

s

e

r

p

x 10

e

e

n

e

G 5

0 9 1 G 1 4 2 1 2 2 9 X A 5 T A M B X B X H 5 4 W T P 0 M 3 O L R D A F R E D S N D G Z N S A H G

Appendix 15 – APEL differentiation media comparison. qRT-PCR data showing relative gene expression levels of markers of the gonad after 7 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. PCS_201_010 iPSCs grown on Matrigel treated with 4 days of CHIR at 4 μM followed by FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml). Three differentiation media were tested: APEL media (Stem Cell Technologie), APEL media (made by Dr. Elizabeth Ng, MCRI) and APEL2 media with 4% PFHM-II (Stem Cell Technologies).

500 iPSC

s 400

C APEL SD day 7

S

P 300

i APEL EN day 7

o

t

200 e APEL2 SD day 7

v

i

t

a l 40

e

r

n

o

i

s 20

s

e

r

p 10

x

e

e

n

e 5

G

0 9 1 G 1 4 2 9 X A 5 T A X X H 5 4 W T M O L R D A E S N D G A G

248 Appendix 16 – Day 18 differentiated cells are negative for markers of closely related lineages adrenal, kidney and cartilage. qRT-PCR data showing relative gene expression levels of markers of the cartilage (ACTG1, ANXA2), kidney (NPHS2, PAX2) and adrenal gland (ARHGAP36, SULT2A1) after 18 days of monolayer differentiation. Each sample represents a biological triplicate (Mean ± S.E.M). Gene expression is quantified relative to iPSCs. The NT2D1 cell line was used as a positive control. PCS_201_010 iPSCs grown in E6 media on Matrigel treated with 4 days of CHIR at 3 μM, then 3 days of FGF9 (200 ng/ml), Heparin (1 μg/ml) and BMP4 (10 ng/ml).

Cells were grown in E6 media from days 7-10 then treated with or without PGD2 (500 ng/ml).

ACTG1 ANXA2 NPHS2

s

s 6 6 s 25

C

C

C

S

S

S

P

P

P

i

i

i

o

o o 20

t

t

t

e

e

e

v

v

v

i

i 4 4 i

t

t

t

a

a

a

l

l l 15

e

e

e

r

r

r

n

n

n

o

o

o

i

i

i

s 10

s

s

s

s 2 2 s

e

e e

r

r r

p

p

p

x x x 5

e

e

e

e

e

e

n

n n

e

e

e

G

G 0 0 G 0 s 7 6 2 1 s 7 6 2 1 s 7 6 2 1 C y E D D C y E D D C y E D D S a - G 2 S a - G 2 S a - G 2 T P 8 T T iP D 8 P i D P N iP D 8 P 1 - N 1 - 1 - N y 8 y 8 y 8 a 1 a 1 a 1 D y D y D y a a a D D D

PAX2 ARHGAP36 SULT2A1

s s s 150 5 4

C C

C

S S

S

P P

P

i i

i

100

o o o 4

t t

t

3

e e

e

v v

v

i i

i

t t t 50

a a

a

l l l 3

e e

e

r r

r

2

n n n 10

o o

o

i i i 2

s s s 8

s s

s

e e

e

r r r 6

p p p 1

x x x 1

e e e 4

e e

e

n n n 2

e e

e

G G G 0 0 0 s 7 6 2 1 s 7 6 2 1 s 7 6 2 1 C y E D D C y E D D C y E D D S a - G 2 S a - G 2 S a - G 2 T P D 8 T P D 8 T iP D 8 P i P N i P N 1 - N 1 - 1 - y 8 y 8 y 8 a 1 a 1 a 1 D y D y D y a a a D D D

249

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Knarston, Ingrid May

Title: Disorders of sex development: genetic analysis and development of a novel in vitro cell model

Date: 2018

Persistent Link: http://hdl.handle.net/11343/221695

File Description: PhD Thesis

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.