Bone Homeostasis As a Multifaceted Puzzle: from WNT Ligands to Metal-Ion Transporters

Faculteit Farmaceutische, Biomedische en Diergeneeskundige Wetenschappen

Bone Homeostasis as a Multifaceted Puzzle: From WNT Ligands to Metal-ion Transporters

Bothomeostase als een Veelzijdige Puzzel: Van WNT Liganden tot Metaal-ion Transporters

Proefschrift voorgelegd tot het behalen van de graad van Doctor in de Biomedische Wetenschappen aan de Universiteit Antwerpen te verdedigen door

Gretl HENDRICKX

Bone Homeostasis as a Multifaceted Puzzle: From WNT Ligands to Metal-ion Transporters Doctoral dissertation, University of Antwerp, Belgium

ISBN: 978-94-6299-582-6 Author: G. Hendrickx Cover Design: Lien Campo, concept by G. Hendrickx Printing: Ridderprint BV

Curiosity-Driven

MEMBERS OF THE JURY

Prof. Dr. Wim Van Hul, promotor Center of Medical Genetics, University of Antwerp and University Hospital of Antwerp, Belgium

Dr. Eveline Boudin, co-promotor Center of Medical Genetics, University of Antwerp and University Hospital of Antwerp, Belgium

Prof. Dr. Vincent Timmerman, chairman VIB Department of Molecular Genetics, University of Antwerp, Belgium

Prof. Dr. Marc Cruts VIB Department of Molecular Genetics, University of Antwerp, Belgium

Prof. Dr. Bart Loeys Center of Medical Genetics, University of Antwerp and University Hospital of Antwerp, Belgium

Prof. Dr. Geert Carmeliet Department of Cellular and Molecular Medicine, KU Leuven, Belgium

Prof. Dr. Martine Cohen-Solal Department of Rheumatology, University Paris-Diderot, France

TABLE OF CONTENTS

Table of Contents

I. Summary/Samenvatting 9

II. Introduction 15 Chapter 1 17 A Look Behind The Scenes: The Risk and Pathogenesis of Primary Osteoporosis Gretl Hendrickx, Eveline Boudin and Wim Van Hul Nature Reviews Rheumatology 2015; 11;462–474

Chapter 2 51 An Introduction to WNT Signaling

III. Objectives 59

IV. Results 65

Part 1 – Osteoporosis 67

Chapter 3 69 Variation in the Kozak Sequence of WNT16 Results in an Increased Translation and is Associated with Osteoporosis Related Parameters Gretl Hendrickx, Eveline Boudin, Igor Fijałkowski, Torben Leo Nielsen, Marianne Andersen, Kim Brixen and Wim Van Hul Bone 2014; 59:57-65

Chapter 4 91 WNT16 Requires Gα Subunits as Intracellular Partners for Both its Canonical and Non-Canonical WNT Signaling Activity in Osteoblasts Gretl Hendrickx, Eveline Boudin and Wim Van Hul Work in progress

Table of Contents

Chapter 5 109 Genetic Screening of WNT4 and WNT5B in Two Populations with Deviating Bone Mineral Densities Gretl Hendrickx, Eveline Boudin, Ellen Steenackers, Torben Leo Nielsen, Marianne Andersen, Kim Brixen and Wim Van Hul Calcified Tissue International 2017; 100:244–249

Part 2 – Hyperostosis Cranialis Interna 123

Chapter 6 125

Conditional Mouse Models Support the Role of SLC39A14 (ZIP14) in Hyperostosis Cranialis Interna and in Bone Homeostasis

Gretl Hendrickx, Vere M. Borra, Ellen Steenackers, Timur A. Yorgan, Christophe Hermans, Eveline Boudin, Jérôme J. Waterval, Ineke D.C. Jansen, Tolunay Beker Aydemir, Niels Kamerling, Geert J. Behets, Christine Plumeyer, Patrick C. D’Haese, Björn Busse, Vincent Everts, Martin Lammens, Geert Mortier, Robert J. Cousins, Thorsten Schinke, Robert J. Stokroos, Johannes J. Manni and Wim Van Hul Submitted

V. General Discussion 165

VI. List of Abbreviations 179

VII. Curriculum Vitae 185

VIII. Dankwoord 193

SUMMARY / SAMENVATTING

Summary/Samenvatting

Summary

Bone homeostasis is a process characterized by the constant and lifelong remodeling of bone tissue to sustain bone mineral density (BMD) and quality. Here, mainly genetic factors contribute to the determination of BMD, which is why genetic variation also contributes to the development of skeletal pathologies characterized by a too low or high BMD, encompassing a multifaceted spectrum. Osteoporosis is the most prevalent skeletal disorder, characterized by a low BMD and an increased fracture risk and with a complex genetic background. Just as well, there are bone disorders with deviating BMD values caused by only one mutation with a large impact on the skeleton. Unlike osteoporosis, these monogenic bone disorders are mostly rare. Studying BMD and bone homeostasis by means of different genetic skeletal disorders was already proven to be an efficient strategy. The aim of this thesis was therefore to further elucidate the regulation of BMD by performing genetic and functional research on two different bone disorders; osteoporosis and Hyperostosis Cranialis Interna.

Osteoporosis was investigated in the first part of this thesis through the study of WNT4, WNT5B and WNT16, three genes associated with BMD in previous genetic studies and are all part of the WNT signaling pathway. First, a candidate gene association study of WNT16 identified additional associations with BMD at distinct skeletal sites. Further genetic screening of WNT16 in two populations with deviating BMD values detected a functional variant in the Kozak consensus sequence of WNT16, resulting in a higher translation efficiency and being associated with a higher BMD. A follow-up functional study investigating the mechanism of action of WNT16 on the canonical and non-canonical WNT signaling pathways demonstrated differential actions of WNT16 in different cell types and a possible orchestrating role for Gα subunits herein. In contrast to WNT16, genetic screening of WNT4 and WNT5B did not reveal interesting variants for further analysis.

In the second part of this thesis, we performed research of Hyperostosis Cranialis Interna (HCI). HCI is a monogenic skeletal disorder characterized bone overgrowth at the calvarium and skull base, resulting in cranial nerve entrapment and associated symptoms. By performing whole- exome sequencing, we identified a mutation (p.L441R) in the SLC39A14 (ZIP14) gene, encoding a zinc transporter. In vitro studies indicated that the subcellular localization of p.L441R ZIP14 is disturbed with consequent defects in cellular zinc homeostasis. Two conditional knock-in mouse models were generated, overexpressing p.L438R Zip14 in osteoblasts and osteoclasts. Remarkably, the calvaria of these mice were unaffected, whereas the rest of the skeleton was affected by Zip14L438R in vivo. Both mouse models, and mainly the osteoblast-specific knock-in

Summary/Samenvatting

mice, exhibit an increased cortical bone volume. Trabecular bone volume was differentially modulated by Zip14L438R, i.e. decreased (osteoblast knock-in) or unaffected (osteoclast knock-in). We therefore hypothesize that Zip14L438R, similar to estrogen and opposite to parathyroid hormone, modulates bone remodeling differently in the cortical and trabecular compartment. Of note, both knock-in models exhibit an increased cortical thickness, owing to an increased endosteal bone formation, which is the underlying cause of intracranial osteosclerosis in patients with HCI. We therefore believe we identified the disease causing gene for Hyperostosis Cranialis Interna and that this rare, unique bone disorder is the result of appropriate compensatory mechanisms in the appendicular skeleton and vertebral column and/or a calvarium-specific lack of such mechanism in HCI patients.

In conclusion, by investigating two different bone disorders, we contributed to the genetic knowledge on BMD. Functional studies towards the mechanisms of action of WNT16 and ZIP14 additionally led to novel insights in the multifaceted process of bone homeostasis.

Summary/Samenvatting

Samenvatting

Bothomeostase is een proces dat gekenmerkt wordt door constante en levenslange remodellering van botweefsel om de botmineraaldensiteit (BMD) en –kwaliteit te onderhouden. Bij het bepalen van de BMD spelen voornamelijk genetische factoren een rol, waardoor genetische variatie ook bijdraagt tot de ontwikkeling van skeletaandoeningen die gekenmerkt worden door een abnormaal lage of hoge BMD en tezamen een veelzijdig spectrum aan botaandoeningen vormen. Osteoporose is de vaakst voorkomende botaandoening die gekenmerkt wordt door een verlaagde BMD, een verhoogd fractuurrisico en een complexe genetische achtergrond heeft. Anderzijds zijn er ook botaandoeningen met afwijkende BMD waarden die veroorzaakt worden door slechts één mutatie met een grote impact op het skelet. In tegenstelling tot osteoporose zijn deze monogene aandoeningen meestal zeldzaam. BMD en bothomeostase bestuderen door middel van onderzoek naar de genetische oorzaak van verschillende botaandoeningen, is in het verleden al een succesvolle strategie gebleken. Het doel van deze thesis was daarom om de regulatie van BMD beter te begrijpen door genetisch en functioneel onderzoek te verrichten op twee verschillende botaandoeningen; osteoporose en Hyperostosis Cranialis Interna.

Osteoporose werd in het eerste deel van deze thesis onderzocht door de studie van WNT4, WNT5B en WNT16, drie genen die in voorafgaande genetische studies reeds werden geassocieerd met BMD en deel uitmaken van de WNT signaalweg. Vooreerst werden in een WNT16 kandidaatgen associatiestudie nog bijkomende associaties met BMD gevonden op verschillende plaatsen in het skelet. In een daaropvolgende genetische screening van WNT16 in twee populaties met afwijkende BMD waarden, detecteerden we een functionele variant in de Kozak consensus sequentie van WNT16. Deze variant verhoogt de translatie-efficiëntie en is geassocieerd met een verhoogde BMD. Een follow-up functionele studie naar het werkingsmechanisme van WNT16 op de canonieke en niet-canonieke WNT signaalweg toonde bovendien differentiële acties aan van WNT16 in verschillende celtypes en een mogelijke orkestrerende rol van Gα subunits hierin. In tegenstelling tot WNT16 werden bij de genetische screening van WNT4 en WNT5B geen interessante genetische varianten gevonden voor verder onderzoek.

In het tweede deel van deze thesis werd onderzoek verricht naar Hyperostosis Cranialis Interna (HCI). HCI is een monogene skeletaandoening die gekenmerkt wordt door een overgroei aan bot ter hoogte van het calvarium en de schedelbasis, wat resulteert in inklemming van craniale zenuwen en bijhorende symptomen. Door exoomsequenering uit te voeren hebben we een mutatie (p.L441R) gedetecteerd in het SLC39A14 (ZIP14) gen, dat codeert voor een

Summary/Samenvatting

zinktransporter. In vitro studies toonden aan dat de cellulaire lokalisatie van p.L441R ZIP14 verstoord is, wat leidt tot defecten in de cellulaire zinkhomeostase. Twee conditionele knock-in muismodellen werden ontwikkeld, met overexpressie van p.L438R Zip14 in osteoblasten en osteoclasten. Opmerkelijk was dat de calvaria van deze muizen onaangetast waren, terwijl de rest van het skelet wel aangetast was door Zip14L438R in vivo. Beide muismodellen, en voornamelijk het osteoblast-specifieke knock-in model, vertonen een toegenomen corticale botmassa. De trabeculaire botmassa werd differentieel gemoduleerd door Zip14L438R, i.e. gedaald (osteoblast knock-in) of niet aangetast (osteoclast knock-in). We vermoeden daarom dat Zip14L438R, net zoals oestrogeen en tegengesteld aan parathyroïd hormoon, botremodellering differentieel moduleert in het corticale en trabeculair botcompartement. Belangrijk hierbij is dat beide knock-in modellen een verhoogde corticale botmassa vertoonden die te wijten was aan een verhoogde endosteale botvorming, wat ook de onderliggende oorzaak is van de intracraniale osteosclerose in patiënten met HCI. We zijn er daarom van overtuigd dat we het ziekteveroorzakende gen voor Hyperostosis Cranialis Interna geïdentificeerd hebben en dat deze zeldzame, unieke botaandoening het resultaat is van onvoldoende compensatoire mechanismen in de schedel van deze patiënten, die wel aanwezig zijn in de rest van het skelet.

In conclusie, door twee verschillende botaandoeningen te onderzoeken, hebben we bijgedragen tot de genetische kennis omtrent BMD. Functionele studies naar de werkingsmechanismen van WNT16 en ZIP14 hebben bovendien geleid tot nieuwe inzichten in bothomeostase als veelzijdig proces.

INTRODUCTION

Chapter 1

A Look Behind the Scenes: the Risk and Pathogenesis of Primary Osteoporosis

Gretl Hendrickx1, Eveline Boudin1 and Wim Van Hul1

1Center of Medical Genetics, University of Antwerp, Belgium

This chapter is based on the review article published in: Nature Reviews Rheumatology 2015; 11;462–474

Introduction – Chapter 1

Abstract

Osteoporosis is a common disorder, affecting hundreds of millions of people worldwide, and characterized by decreased bone mineral density and increased fracture risk. Known nonheritable risk factors for primary osteoporosis include advanced age, sex-steroid deficiency and increased oxidative stress. Age is a nonmodifiable risk factor, but the influence of a person’s lifestyle (diet and physical activity) on their bone structure and density is modifiable to some extent. Heritable factors influencing bone fragility can be monogenic or polygenic. Osteogenesis imperfecta, juvenile osteoporosis and syndromes of decreased bone density are discussed as examples of monogenic disorders associated with bone fragility. So far, the factors associated with polygenic osteoporosis have been investigated mainly in genome-wide association studies. However, epigenetic mechanisms also contribute to the heritability of polygenic osteoporosis. Identification of these heritable and nonheritable risk factors has already led to the discovery of therapeutic targets for osteoporosis, which emphasizes the importance of research into the pathogenetic mechanisms of osteoporosis. Accordingly, this article discusses the many heritable and nonheritable factors that contribute to the pathogenesis of primary osteoporosis. Although osteoporosis can also develop secondary to many other diseases or their treatment, a discussion of the factors that contribute only to secondary osteoporosis is beyond the scope of this Review.

Introduction – Chapter 1

I. Introduction

Osteoporosis is a skeletal disorder affecting hundreds of millions of people, and is characterized by increased bone fragility and susceptibility to fracture. Osteoporotic fractures predominantly occur at the level of the forearm, hip and lumbar spine, and are associated with substantial morbidity and mortality. Hip fractures, for example, are associated with mortality of up to 36% in the first year after fracture.1,2

Moreover, osteoporosis is one of the defining health problems of an ageing society. Currently, one in three women and one in five men aged >50 years will experience an osteoporotic fracture.3,4 The hospitalization, follow-up and treatment of patients with osteoporosis already present a considerable financial burden, and in the future, these health-care costs are expected to increase further. The costs to health services of osteoporosis in the European Union were estimated at €31.7 billion in the year 2000, which is projected to increase to €76.7 billion in 2050 as a result of demographic changes in the European population.5 Osteoporosis is, therefore, an important target of research aimed at improving its prevention and treatment, as well as increasing the cost-effectiveness of care for patients affected by this common skeletal disorder.

Figure 1 | Overview of BMD values during life, indicating the importance of peak bone mass and the subsequent rate of decline in BMD in the development of primary osteoporosis. Peak bone mass is reached between the ages of 20 and 30 years. The lower the peak BMD value, the higher the risk of bone fragility later in life. Furthermore, age- related factors—such as menopause in women, lifestyle influences and the genetic background of the individual—will also determine a person’s risk of osteoporotic fracture. Abbreviation: BMD, bone mineral density.

Introduction – Chapter 1

Osteoporotic fracture is the consequence of a decrease in bone mineral density (BMD) in combination with deterioration of the bone microarchitecture. The risk of osteoporotic fracture is strongly dependent on peak bone mass, which is attained at around the age of 25 years (Figure 1). After peak bone mass is reached, BMD remains more or less stable until the age of 55 years, when age-related bone loss starts (Figure 1). The lower an individual’s peak bone mass, the higher their risk of having fragile bones later in life.6

After bones have completed their longitudinal growth, almost 10% of the bone is remodelled each year.5 Bone remodelling is a continuous physiological process whereby old bone is removed by osteoclasts and new bone is formed by osteoblasts. In this way, microdamage can be repaired, which maintains bone strength and mineral homeostasis. Having a stable bone mass requires that the activity of osteoclasts and osteoblasts must be finely balanced. Therefore, the bone remodelling process is strictly regulated by numerous signalling pathways. Osteocytes, which are terminally differentiated osteoblasts located in the mineralized bone tissue, are important coordinators of bone remodelling. Osteocytes can detect microcracks, mechanical strain and changes in the hormonal environment of the bone; they also initiate bone remodelling in response to these stimuli when necessary.7,8 Although the process of bone remodelling is strictly regulated, gradual bone loss still occurs as a result of hormonal changes and ageing. The rate of bone loss throughout life is dependent on a person’s lifestyle and genetic background. The risk of osteoporosis and fragility-related fractures is also influenced by bone quality, which is defined by various parameters including properties of the proteinaceous matrix and the mineral phase. For example, the degree of collagen crosslinking and post-translational modifications, as well as the collagen fibril diameter, all have an effect on bone strength.9 Noncollagenous proteins (such as proteoglycans, alkaline phosphatase, osteonectin and periostin) also have diverse effects on bone composition and mineralization.10,11 Finally, bone strength is influenced by the geometric architecture of the bone, which is also subject to genetic and nongenetic influences.12

Primary osteoporosis is a classic example of a common multifactorial disorder for which an individualized approach to prevention, diagnosis and treatment would help to improve patients’ outcomes. Improved understanding of all the hereditary and nonhereditary factors that might contribute to the development of fragile bones is consequently essential to treat patients with osteoporosis effectively. However, the complexity of this skeletal disorder presents a challenge for the medical community, since osteoporosis can also occur secondary to other diseases or their treatment. A discussion of the many factors that can predispose to secondary osteoporosis (such as glucocorticoid, antiepileptic and other medication use, impaired kidney function, chronic

Introduction – Chapter 1 obstructive pulmonary disease, bowel disease, malabsorption and hyperthyroidism) is beyond the scope of this Review. However, the reader should bear in mind that many of these factors have an increased prevalence in ageing individuals, and should be considered in global assessments of osteoporosis risk.

This Review summarizes the main heritable and nonheritable factors known to have a role in the pathogenesis of primary osteoporosis. We focus on the genetic and epigenetic basis of polygenic osteoporosis, as well as that of monogenic conditions associated with decreased bone mass or increased fracture risk.

II. Nonheritable risk factors 1. Age-related changes

Both men and women experience a progressive decline in BMD, which starts as soon as peak bone mass is reached, and is initially independent of sex-steroid levels (Figure 1). Advancing age is, therefore, a strong risk factor for primary osteoporosis, which mainly affects people older than 50 years (Figure 1). Actually, ‘ageing’ represents the gradual or sudden disturbance of several mechanisms, which can also occur throughout life. Among the plethora of mechanisms that influence primary osteoporosis, oxidative stress, apoptotic mechanisms, sex-steroid deficiency and macroautophagy are the most important. Moreover, disturbance of these mechanisms contributes to the increased predisposition of ageing individuals to the development of secondary osteoporosis. Each of these mechanisms is discussed separately in this article, but can occur concomitantly in the ageing individual (Figure 2).

1.1 Oxidative stress

Oxidative stress is an age-associated mechanism linked to osteoporosis.13 Throughout life, normal intracellular metabolism generates, as by-products, reactive oxygen species (ROS) such as – – superoxide anions (O2 ), hydroxyl radicals (HO ) and hydrogen peroxide (H2O2). Oxidative stress arises when excess ROS accumulate, which occurs when levels of ROS-scavenging enzymes— such as superoxide dismutase (SOD), glutathione peroxidase (GPx), glutathione (GSH) and catalase (CAT)—are reduced, membrane-associated NADPH oxidase (NOX) activity is increased, or as a result of mitochondrial respiratory chain leakage (Figure 2). Subsequently, oxidative stress can damage macromolecules such as lipids in the cell membrane, nuclear and mitochondrial DNA and transcription factors.14 Increased oxidative stress levels and the accumulation of oxidative damage are associated with both ageing itself and age-related disorders. In mouse models of premature ageing due to oxidative damage, the mice have an osteoporosis-

Introduction – Chapter 1 like phenotype, which supports a deleterious role of oxidative stress in bone.15,16 Also, mice deficient in SOD [Cu-Zn] (Sod1–/–) have reduced BMD as well as higher levels of oxidative stress and lower osteoblast and osteoclast numbers than wild-type mice.17,18

Figure 2 | Ageing-associated changes that contribute to bone fragility. Although many mechanisms contribute to bone loss in elderly individuals, age-related bone loss is mainly dictated by increased oxidative stress resulting from increased ROS production and impaired antioxidant systems. In addition, apoptosis of bone cells increases, and macroautophagic mechanisms decrease in efficiency with advancing age. These factors, in turn, contribute to an unbalanced bone remodelling process favouring bone loss, and an increased risk of osteoporotic fracture. The thickness of lines indicating activation or inhibition represents the intensity of the effect. Abbreviations: BMD, bone mineral density; CAT, catalase; FOXO, Forkhead box O transcription factor; GPx, glutathione peroxidase; GSH, – – Shc glutathione; HO , hydroxyl radicals; H2O2, hydrogen peroxide; NOX, NADPH oxidase; O2 , superoxide; p66 , mitochondrial isoform of SHC-transforming protein 1; ROS, reactive oxygen species; SOD, superoxide dismutase.

The forkhead box O (FOXO) family of transcription factors is essential for both defence against oxidative stress and bone homeostasis (Figure 2). Accumulation of ROS leads to retention of FOXO transcription factors in the nucleus, which in turn activates transcription of target genes involved in DNA repair (Gadd45), ROS detoxification (Sod2, Cat), cell cycle arrest (CyclinG2, Cdkn1a, Cdkn1b) and apoptosis (Faslg, Bin1).13 Mice with combined deletion of Foxo1, Foxo3 and Foxo4 show elevated oxidative stress levels in bone, and loss of both cancellous and cortical bone. The bone loss was attributed to a reduction in bone formation resulting from decreased osteoblast numbers.19 Moreover, β catenin is an essential co-activator of FOXO transcription factors, implying that crosstalk occurs between FOXO transcription factors and Wnt–β-catenin

Introduction – Chapter 1 signalling, a crucial pathway in bone homeostasis. Oxidative stress increases with age, and FOXO transcription factors sequester more and more β-catenin at the expense of Wnt–β-catenin signalling activity in ageing individuals. Therefore, this mechanism could contribute to the decreased bone formation and loss of bone mass associated with both ageing and the development of osteoporosis.20–22

Finally, the p53–p66Shc signalling pathway is an important regulator of intracellular redox status, and is involved in age-related skeletal involution.23,24 ROS levels are elevated by p53, via increasing the abundance of p66Shc protein (Figure 2). This observation was confirmed in a p66Shc knockout (p66Shc–/–) mouse model, in which the mice showed a general decrease in levels of oxidative stress along with a lower intracellular ROS concentration in bone and a higher bone mass than wild-type mice.23–27 Also, compared to their wild-type littermates, the p66Shc–/– mice had a 30% increase in lifespan, which reinforces the notion that oxidative stress is strongly associated with ageing. Moreover, mice with elevated p53 activity display premature ageing and have osteoporosis,16 whereas p53-null mice have a high-bone-mass phenotype due to increased osteoblast numbers and an increased bone formation rate.16,28,29

1.2 Apoptosis

Apoptosis is an essential mechanism involved in all stages of an organism’s development, as well as in eliminating poorly functioning and possibly cancerous cells later in life. However, ageing has a detrimental effect on the regulation of apoptosis, through which it contributes to the development of disorders such as osteoporosis (Figure 2). The adverse effect of ageing on bone is mainly related to the death of osteocytes (the most abundant and longest-lived cell type in bone tissue), which substantially contributes to bone fragility in elderly individuals through the increased production of receptor activator of nuclear factor κB (RANK) ligand (RANKL, also known as TNF ligand superfamily member 11) by dying osteocytes.30 Moreover, in mice, targeted deletion of Bax and Bak1 (which encode the two proapoptotic members of the Bcl-2 protein family) in osteocytes and osteoblasts results in inhibition of apoptosis in these two cell types. Inhibition of osteocyte apoptosis resulted in increased trabecular bone mass, owing to prolongation of the lifespan of short-lived osteoblasts. Conversely, inhibition of osteoblast apoptosis led to increased local bone resorption, resulting in increased RANKL expression, an increase in porosity of the bone cortex, and numerous osteoclasts within the pores. This phenotype was attributed to exaggeration of the adverse (proapoptotic) effects of ageing on long- lived cortical osteocytes, in another study from the same group, which confirmed the influence of ageing on osteocyte apoptosis.31 Sex steroids, on the contrary, have an antiapoptotic effect on

Introduction – Chapter 1 osteocytes,32–34 consequently postmenopausal oestrogen deficiency increases osteocyte apoptosis. In ovariectomized rat and mouse models of oestrogen loss, osteocyte apoptosis was markedly increased.21,32

A phenomenon associated with osteocyte apoptosis is micropetrosis or hypermineralization of the osteocyte lacunae. Osteocyte apoptosis is followed by hypermineralization of the perilacunar bone and, ultimately, filling of the canaliculae with mineralized connective tissue.35 Aged bone shows an increase in hypermineralized occlusions,36 and bone from patients with osteoporosis specifically shows an increased percentage of hypermineralized lacunae (Figure 2).36,37 Overall, osteocyte apoptosis and micropetrosis cause a decrease in canalicular fluid flow and deterioration in the detection of microdamage by osteocytes, which has deleterious consequences for the bone’s structural integrity. In this way, hypermineralization of the lacunae contributes to bone brittleness and fragility.36,37

Osteoblasts also undergo apoptosis owing to the increased oxidative stress levels associated with ageing. Studies in p66Shc–/– mice showed that the p66Shc redox protein is required for translation of oxidative signals in osteoblasts into osteoblast apoptosis. As we have discussed, these mice have reduced ROS levels in bone and a high-bone-mass phenotype.27 Moreover, in the mouse model of simultaneous deletion of Foxo1, Foxo3 and Foxo4, bone loss results from both decreased bone formation and reduced osteoblast numbers due to increased osteoblast apoptosis.19 A possible mechanism through which p66Shc might regulate apoptosis is, therefore, inhibition of these FOXO transcription factors.

1.3 Sex-steroid deficiency

Sex steroids have an indispensable role in bone growth, the attainment of peak bone mass during the first three decades of life, and skeletal homeostasis during adulthood.38 Moreover, differences in androgen and oestrogen levels in boys and girls contribute to sex-related differences in bone size and strength. Periosteal bone formation is restrained by oestrogens and stimulated by androgens, resulting in bigger and stronger bones in male than in female individuals.39 Furthermore, these differences in early bone growth and peak bone mass help to explain why osteoporosis is more prevalent in women than in men.

Androgens and oestrogens also modulate bone homeostasis differently during adulthood. Mouse models of cell-specific deletion of ERα and AR (the oestrogen and androgen receptors, respectively), together with studies of hormone-replacement therapy in postmenopausal women, have identified dissimilar actions of both androgens and oestrogens on different bone cell types.38

Introduction – Chapter 1

The effects of oestrogens in women both promote resorption of trabecular bone (via direct effects on osteoclasts) and attenuate resorption of cortical bone (via indirect effects on osteoblast progenitors through a non-nucleus-initiated mechanism), resulting in an overall increase in bone turnover.40,41 Androgens in men, by contrast, exert protective effects on trabecular bone through osteoblasts and osteocytes, but not through osteoclasts.42–44 In addition to the antiapoptotic effects of oestrogens on osteocytes mentioned above,32–34 antiapoptotic effects on osteoblasts have been widely reported.32,34,45–48

Overall, sex steroids are of major importance for maintaining the balance between bone formation and resorption and for slowing down the rate of bone turnover.49 In later life, the decline in sex steroid levels contributes to bone fragility and osteoporotic fractures in both men and women (Figure 1).14,50 In women, the effects of sex-steroid deficiency cause an accelerated decline in BMD in the early postmenopausal period owing to an imbalance between resorption and formation of bone, which is attributable to a prolonged lifespan of osteoclasts and decreased lifespan of osteoblasts and osteocytes.14,49 After this initial drop in BMD, however, the subsequent rates of bone loss are similar in men and women.38 In both sexes, age-related bone loss is characterized by decreased remodelling in the trabecular compartment and increased remodelling in the cortical compartment, which result in increased cortical porosity.51–53 These disparate effects in different bone types could be explained by the fact that oestrogens regulate trabecular and cortical bone through different cell types, as previously mentioned; these two bone types, therefore, respond differently to the age-related decline in oestrogen levels.38

Finally, sex steroids have antiosteoporotic effects through their ability to protect against excessive oxidative stress (Figure 2).14,54 In 5 month-old gonadectomized female and male mice, which show the same increases in ROS levels and p53 and p66Shc activity that are observed with advancing age, this increase in oxidative stress levels could be reversed by replacement of sex steroids (oestrogens or a non-aromatizable androgen) or by administration of the antioxidant N- acetylcysteine. Furthermore, the subsequent increase in osteoblast and osteocyte apoptosis could also be prevented by either N acetylcysteine treatment or hormone replacement.33 Similarly, in humans, the loss of oestrogens associated with menopause results in an increase in osteoblast and osteocyte apoptosis, which contributes to the development of postmenopausal osteoporosis.

1.4 Macroautophagy

In the past decade, knowledge of macroautophagy—an evolutionarily conserved process that recycles damaged organelles and misfolded proteins in order to prolong cell survival—has

Introduction – Chapter 1 substantially increased. Macroautophagy is an important mechanism not only for sustaining cell function, but also for responding to environmental or internal stressors, such as hypoxia or increased oxidative stress.55,56 A genome-wide association study (GWAS) identified a link between variants in macroautophagy-related genes and BMD.57 Also, macroautophagy is now known to decline in efficiency with age, and this change disturbs bone homeostasis.58 Further in vivo evidence for a role of macroautophagy in the development of osteoporosis has been gained in mice with deletion of ribosomal protein S6 kinase β1 (S6K1, which is encoded by Rps6kb1). Female mice lacking S6K1 are protected from age-related bone loss, despite a 20% increase in lifespan.59 This ribosomal protein limits excessive activation of macroautophagy under nutrient- deprived conditions, suggesting that Rps6kb1–/– mice are less prone than wild-type mice to the age-related decline in macroautophagy.59,60 Furthermore, observations in mice with targeted deletion of other macroautophagy-related genes (Atg7, Atg5 and Rb1cc1) in osteoclasts, osteoblasts and osteocytes have reinforced the notion that macroautophagy in all three of these cell types is essential for bone homeostasis.61–63

2. Lifestyle factors

The influence of a person’s lifestyle on their risk of osteoporosis should not be underestimated, and (unlike ageing) is modifiable to some extent. The composition of a person’s diet, frequency and type of physical activity, cigarette smoking and alcohol consumption can have considerable advantageous or deleterious effects on their susceptibility to osteoporotic fracture (Figure 3).

Some endogenous nutrients still need to be taken up through nutrition or the environment, such as vitamin D and calcium (Figure 3). Together with parathyroid hormone (PTH), vitamin D is part of an endogenous homeostatic mechanism that regulates calcium and phosphate levels, which are important determinants of the strength and mineralization of bone tissue. Insufficient supply of calcium, phosphate and vitamin D impairs bone formation and mineralization, as well as increasing the resorption of bone. In individuals with inadequate intake of these nutrients, bone mass and strength decreases and fracture risk increases. Furthermore, deficiency of these nutrients decreases muscular mass and strength, which impairs neuromuscular function and balance, thereby increasing the risk of falls and fractures. Therefore, sufficient intake of calcium and vitamin D, as well as sun exposure (which is essential to produce vitamin D), are strongly recommended, certainly for people at risk of osteoporosis.64–66

Introduction – Chapter 1

Figure 3 | Factors that influence the development of osteoporosis. Both heritable and nonheritable factors contribute to the development of primary osteoporosis, which is consequently considered a multifactorial disorder. The combination and impact of the different risk factors present in an ageing individual determines their overall risk of bone fragility and associated fractures.

As previously mentioned, the age-related increase in oxidative stress has detrimental effects on bone homeostasis. Inclusion of fruits and phytochemicals in a person’s diet might, therefore, aid in the defence against excessive ROS levels.67,68 Several studies have demonstrated a beneficial effect of frequent fruit intake on bone health.69–72 Antioxidants in tomatoes, plums, citrus fruits, berries, grapes and apples have demonstrated specific advantageous effects on bone health in human, animal and in vitro studies.68 Besides antioxidants, dietary protein also affects bone health. Protein is identified as being both detrimental and beneficial to bone mass, depending on several factors, including the amount of protein, its source, concomitant calcium intake and the acid– base balance of the diet.73 Overall, high-protein diets are associated with greater bone mass and fewer fractures than low-protein diets, as long as calcium intake is adequate.73,74 The acid– base balance is adversely influenced by substantial consumption of meat, eggs, fish and cereal (which increase the net acid load), whereas substantial consumption of fruit and vegetables results in a potential positive effect (increased net base load).73 Besides positively affecting the redox balance, increased consumption of fruit and vegetables generates a state of metabolic

Introduction – Chapter 1 alkalosis, which is favourable for bone formation and calcium balance, prevents bone loss, and reduces the risk of osteoporotic fractures.75,76

In the past decade, the importance of physical activity for preventing and ameliorating osteoporosis has become increasingly evident (Figure 3). Levels of oxidative stress are reduced by exercise.77 Physical activity also induces mechanical pressure and load on the entire musculoskeletal system. As mentioned previously, osteocytes form a tight and extensive communication network that can sense mechanical strain during physical exercise. Osteocytes react to these stimuli by emitting specific signals to cells at the bone surface,7 with the purpose of triggering bone remodelling.78,79 For example, osteocytes decrease their secretion of sclerostin in response to mechanical forces. Sclerostin is encoded by the SOST gene, mutations in which cause two monogenic sclerosing bone dysplasias (van Buchem disease and sclerosteosis).80,81 Sclerostin is an extracellular inhibitor of Wnt–β-catenin signalling.82,83 The Wnt signalling pathway is activated when Wnt ligands bind to the receptor complex formed by LRP5–LRP6 (LDL receptor-related proteins 5 and 6, respectively) co-receptors and Frizzled receptors. Activation of this pathway by Wnt ligands results in stabilization of β catenin and activates the transcription of several target genes involved in osteogenesis. Therefore, decreased secretion of sclerostin by osteocytes, as a reaction to increased mechanical load, results in heightened activation of the Wnt–β-catenin signalling pathway, resulting in increased bone formation as well as improvements in BMD and bone strength. Various genetic mutations in LRP5 can lead to monogenic bone disorders with phenotypes of either increased or decreased bone mass, and several in vitro and in vivo experiments have confirmed that Wnt signalling is one of the most important pathways in the regulation of osteoblast proliferation, differentiation and function.84,85 Consequently, this pathway is an important regulator of bone formation, and is considered to be an interesting target for osteoporosis treatment.

The effect of alcohol consumption on bone metabolism, and consequently on bone tissue, depends on the quantity consumed (Figure 3).86,87 Chronic, excessive intake of alcohol is associated with a decrease in BMD and bone mineral content, resulting in an increased susceptibility to fractures. These features are the result of both direct and indirect effects of alcohol on bone metabolism.87–89 Furthermore, excessive alcohol intake can result in inhibition of bone formation, as well as increased bone resorption, via several mechanisms.87 However, despite the adverse effects of high alcohol consumption on bone, other evidence suggests that moderate alcohol consumption is beneficial for bone.86,87

Introduction – Chapter 1

Several studies comparing smokers and nonsmokers showed a clear association between cigarette smoking and a dose-dependent deficit in BMD at several sites (Figure 3).90–92 The mechanisms underlying this bone loss are not yet clear, mainly because smoking affects many tissues and regulatory pathways, such that the effects seen on bone are probably a combination of many direct and indirect actions. Regardless of this uncertainty, smoking cessation is strongly recommended in individuals at risk of osteoporosis, as its deleterious effects on bone are at least partially reversible.92

III. Heritable risk factors

In polygenic osteoporosis, which is the most common type, the heritable component is defined by a large number of single nucleotide polymorphisms (SNPs), each with small effect sizes, and by epigenetic factors. However, osteoporosis can also be caused by a single mutation in one gene. Monogenic conditions associated with reduced bone mass have a low prevalence and are mostly detected in childhood or adolescence. Studying the genetic cause of these rare Mendelian disorders has been a valuable strategy resulting in the identification of genetic determinants that also contribute to multifactorial osteoporosis.

1. Monogenic conditions causing decreased BMD

The group of monogenic disorders linked to decreased BMD and increased fracture risk can be subdivided according to the specific genetic cause and the presence or absence of additional symptoms. Here, we discuss the genetic causes of osteogenesis imperfecta, juvenile osteoporosis, and some other syndromes characterized by osteoporosis.

1.1 Osteogenesis imperfecta

Osteogenesis imperfecta is a clinically and genetically heterogeneous group of connective tissue disorders that are commonly referred to as brittle bone disease.93 In patients with osteogenesis imperfecta, low bone mass and fractures are associated with extraosseus connective tissue symptoms, such as blue sclerae and dentinogenesis imperfecta. The original classification of this group of disorders, published in 1979, recognized only four osteogenesis imperfecta types (I– IV).94 Subsequently, the number of osteogenesis imperfecta types listed in the Online Mendelian Inheritance in Man® (OMIM®) database95 proliferated with the discovery of each new genetic defect. At the time of writing, 16 causative genes have been identified, corresponding to osteogenesis imperfecta types 1–16 in the OMIM database, and this number will probably continue to rise with the use of next-generation sequencing technologies (Table 1).96,97

Introduction – Chapter 1

Table 1 | Overview of the types of osteogenesis imperfecta. Disease type Inheritance Gene OMIM®entry Protein product Function Type 1: nondeforming AD COL1A1 Type I Collagen α1 (I) chain Collagen type 1 OI with blue sclerae #166,200 biosynthesis AD COL1A2 Type I Collagen α2 (I) chain Collagen type 1 #166,200 biosynthesis Type 2: perinatally AD COL1A1 Type II Collagen α1 (I) chain Collagen type 1 lethal #166,210 biosynthesis AD COL1A2 Type II Collagen α2 (I) chain Collagen type 1 #166,210 biosynthesis AR CRTAP Type VII Cartilage-associated Collagen type 1 #610,682 protein biosynthesis AR LEPRE1 Type VIII Prolyl 3-hydroxylase 1 Collagen type 1 #610,915 biosynthesis AR PPIB Type IX Peptidyl-prolyl cis–trans Collagen type 1 #259,440 isomerase B biosynthesis Type 3: progressively AD COL1A1 Type III Collagen α1 (I) chain Collagen type 1 deforming #259,420 biosynthesis AD COL1A2 Type III Collagen α2 (I) chain Collagen type 1 #259,420 biosynthesis AR BMP1 Type XIII Bone morphogenetic Collagen type 1 #641,856 protein 1 biosynthesis AR CRTAP Type VII Cartilage-associated Collagen type 1 #610,682 protein biosynthesis AR FKBP10 Type XI Peptidyl-prolyl cis–trans Collagen type 1 #610,968 isomerase FKBP10 biosynthesis AR LEPRE1 Type VIII Prolyl 3-hydroxylase 1 Collagen type 1 #610,915 biosynthesis AR PLOD2 Unknown type Procollagen-lysine, Collagen type 1 2-oxoglutarate biosynthesis 5-dioxygenase 2 AR PPIB Type IX Peptidy-lprolyl cis–trans Collagen type 1 #259,440 isomerase B biosynthesis AR SERPINF1 Type VI Pigment epithelium Osteoid #613,982 derived factor mineralization AR SERPINH1 Type X Serpin H1 Collagen type 1 #613,848 biosynthesis AR TMEM38B Type XIV Trimeric intracellular Unknown #615,066 cation channel type B AR WNT1 Type XV Proto-oncogene Wnt1 Bone formation, #615,220 osteoblast activity AR CREB3L1 Unknown type Cyclic AMP-responsive Collagen type 1 element binding protein biosynthesis 3-like protein 1 Type 4: moderate AD COL1A1 Type IV Collagen α1 (I) chain Collagen type 1 severity with normal #166,220 biosynthesis sclerae AD COL1A2 Type IV Collagen α2 (I) chain Collagen type 1 #166,220 biosynthesis AR CRTAP Type VII Cartilage-associated Collagen type 1 #610,682 protein biosynthesis AR PPIB Type IX Peptidyl-prolyl cis–trans Collagen type 1 #259,440 isomerase B biosynthesis AR SP7 Type XII Transcription factor Sp7 Osteoblast #613,849 differentiation Type 5: calcification in AD IFITM5 Type V Interferon-induced Unknown interosseous #610,967 transmembrane protein 5 membranes Abbreviations: AD, autosomal dominant; AR, autosomal recessive; OMIM, Online Mendelian Inheritance in Man database at http://www.omim.org . Permission obtained from Wiley © Van Dijk, F. S. & Sillence, D. O. Osteogenesis imperfecta: Clinical diagnosis, nomenclature and severity assessment. Am. J. Med. Genet. 164 (Part A), 1470–1481 (2014).90

Introduction – Chapter 1

In 2009, therefore, the International Working Group on Constitutional Disorders of Bone divided the known osteogenesis imperfecta types into five main groups on the basis of the specific clinical features and severity of the disease, although the importance of the different genetic causes is still recognized (Table 1).96,98 This revised classification was published in its 2010 nosology.98 Accordingly, five major types of osteogenesis imperfecta (I–V) are currently recognized.96,98

Despite the increasing number of known disease-causing genes, 90% of patients with osteogenesis imperfecta have a mutation in COL1A1 or COL1A2.99 In these patients, osteogenesis imperfecta is inherited in an autosomal dominant manner. Depending on the type and location of the mutation, variants in COL1A1 or COL1A2 can cause osteogenesis imperfecta types I, II, III or IV. Osteogenesis imperfecta type V is also inherited in an autosomal dominant manner, but only one disease-causing gene (IFITM5, which encodes interferon-induced transmembrane protein 5) has been identified to date (Table 1).93,97,100 In around 10% of patients, osteogenesis imperfecta is inherited in a recessive manner; so far, 13 causative genes have been identified. Mutations in some of these genes (CRTAP, PPIB and LEPRE1) can cause several different types of osteogenesis imperfecta, as for COL1A1 and COL1A2.

The majority of genes linked to osteogenesis imperfecta are involved in the collagen type 1 biosynthesis pathway.93,96,101 However, in one study, the expression of transforming growth factor β (TGF-β) receptors was increased in osteoblasts from patients with osteogenesis imperfecta versus those from healthy control individuals.102 Furthermore, altered TGF-β signalling is observed in several mouse models of osteogenesis imperfecta, and has been suggested as the primary pathogenetic mechanism of this disease. Consequently, targeting TGF-β signalling could be a promising therapeutic strategy for osteogenesis imperfecta.103 Apart from the genes involved in collagen type 1 biosynthesis and TGF β signalling, some osteogenesis imperfecta disease- causing genes have other roles in bone homeostasis (Table 1). For example, SP7 encodes transcription factor Sp7 (also known as zinc finger protein osterix), a well-known osteoblast- specific transcription factor that regulates osteoblast differentiation.104 SERPINF1 encodes pigment epithelium-derived factor (PEDF), which is required for osteoid mineralization, and is consequently important for bone formation and bone remodelling.101 Finally, Wnt1 is a known activator of the Wnt–β-catenin signalling pathway, which (as noted above) is well-established regulator of bone formation. Mutations in WNT1 can not only cause osteogenesis imperfecta type III, but are also found in some patients with juvenile osteoporosis.105,106

Introduction – Chapter 1

1.2 Juvenile osteoporosis

Juvenile osteoporosis is a group of heritable disorders that present during childhood and are characterized by low BMD and bone fragility, but without the extraskeletal features found in osteogenesis imperfecta. The low bone mass in patients with juvenile osteoporosis is typically the result of decreased osteoblast activity and mainly affects trabecular bone.107,108 As mentioned previously, heterozygous mutations in WNT1 have been found in several patients with autosomal dominant juvenile osteoporosis.105,106 Functional studies demonstrated that the altered Wnt1 protein from these patients has an impaired capacity to activate Wnt–β-catenin signalling.105,106 Besides mutations in WNT1 itself, mutations in LRP5 (which encodes a co-receptor involved in Wnt–β-catenin signalling) can cause juvenile osteoporosis. Patients with heterozygous mutations in LRP5 also show impaired activation of Wnt–β-catenin signalling.107–109

Finally, a third juvenile osteoporosis gene, PLS3 (which encodes plastin-3), was discovered in 2013.110 Disease-causing variants in PLS3 are found in several families with X linked juvenile osteoporosis. Although PLS3 is ubiquitously expressed in solid tissues, the function of plastin-3 in bone remains unclear. Plastin 3 can bind to actin, and is part of a protein family involved in the dynamic assembly and disassembly of the actin cytoskeleton via interactions with monomeric or globular and filamentous actin. More studies are needed to clarify the role of plastin-3 in the regulation of bone mass and in the pathogenesis of juvenile osteoporosis.110,111

2. Monogenic syndromes with low BMD

The final group of monogenic disorders includes several syndromes characterized by osteoporosis with additional features (Table 2).94,112 Osteoporosis–pseudoglioma syndrome and Bruck syndrome are caused by genes that are also linked to juvenile osteoporosis or osteogenesis imperfecta. Osteoporosis–pseudoglioma syndrome is an autosomal recessive disorder caused by homozygous or compound heterozygous mutations in LRP5, and is marked by severe osteoporosis and early-onset blind-ness.107,108,113,114 The importance of LRP5 in the regulation of bone formation is further demonstrated by the fact that mutations in LRP5 can cause a range of disorders characterized by either decreased bone density (osteoporosis–pseudoglioma syndrome, juvenile osteoporosis) or increased bone density (high bone mass phenotypes, osteosclerosis and endosteal hyperostosis), depending on the type or location of the mutation.84,115

Bruck syndrome types 1 and 2 are autosomal recessive disorders caused by mutations in FKBP10 and PLOD2, respectively. Mutations in both these genes are also reported to cause osteogenesis imperfecta type III, which again demonstrates the importance of these proteins in the regulation

Introduction – Chapter 1 of bone mass (Tables 1 and 2). In addition to osteoporosis, patients with Bruck syndrome also have contractures of the large joints and pterygia.96,112 Bruck syndrome and osteogenesis imperfecta type III are both autosomal recessive disorders, and the same mutation can result in either disorder within the same family, probably due to the influence of genetic modifiers.116–118

Table 2 | Overview of syndromes associated with low bone mass. Additional clinical Disorder Inheritance Gene Protein features Osteoporosis– AR LRP5* LDL receptor-related Blindness pseudoglioma syndrome protein 5 Bruck syndrome type 1 AR FKBP10* Peptidyl-prolyl cis–trans Joint contractures, isomerase FKBP10 pterygia Bruck syndrome type 2 AR PLOD2* Procollagen-lysine, Joint contractures, 2-oxoglutarate pterygia 5-dioxygenase 2 Hajdu–Cheney AD NOTCH2 Neurogenic locus Notch Facial abnormalities, acro- syndrome homologue protein 2 osteolysis, hearing loss, renal cysts Gnathodiaphyseal AD ANO5 Anoctamin-5 Purulent osteomyelitis of dysplasia the jaws in adulthood Geroderma AR GORAB RAB6-interacting golgin Wrinkly skin osteodysplasticum Spondylocular AR Unknown Unknown Platyspondyly, cataracts, syndrome retinal detachment, short trunk, facial dysmorphism, immobile spine, kyphosis Cole–Carpenter Unknown Unknown Unknown Craniosynostosis, ocular dysplasia proptosis, distinctive facial features, hydrocephalus Calvarial doughnut AD Unknown Unknown Lumps on the head, lesions with bone dental caries, increased fragility serum alkaline phosphatase levels Cleidocranial dysplasia AD RUNX2 Runt-related transcription Persistent fontanelles, factor 2 hypoplasia or aplasia of clavicles, scoliosis, wide pubic symphysis, dental and digital anomalies ADCAD2 AD LRP6 LDL receptor-related Early-onset coronary protein 6 disease *Mutations in this gene are also linked to osteogenesis imperfecta or juvenile osteoporosis. Abbreviations: ADCAD2, coronary artery disease, autosomal dominant 2; AD, autosomal dominant; AR, autosomal recessive.

3. Genetics of multifactorial osteoporosis

Although bone fracture is the most important clinical outcome in patients with multifactorial osteoporosis, at this moment BMD measurements serve as the most widely used method to diagnose osteoporosis (as low BMD is an important predictor of fracture). Several twin and family studies have shown that 50–85% of the variance in BMD is genetically determined.119 Heritability of fractures is 25–35%, however—substantially lower than that of BMD, which

Introduction – Chapter 1 indicates that fracture is partially independent of BMD. Furthermore, a large number of individuals with osteoporotic fractures do not have osteoporosis according to BMD criteria.120 These observations confirm that, besides BMD, additional parameters such as bone geometry, bone quality and muscle strength are important in determining the risk of osteoporotic fractures.121,122

Several GWASs, performed in populations with ever-larger sample sizes, have highlighted the extremely polygenic nature of the variance in BMD.123–130 Many SNPs—representing >50 loci— are already reported to be associated with BMD; however, they collectively explain only a small fraction (5.8%) of the total genetic variance in BMD.124 Several BMD-associated SNPs are located in or near to genes with a known function in the regulation of bone homeostasis. A large number of these genes are involved in Wnt signalling (LRP5, SOST, WNT1, LRP4), endochondral ossification (RUNX2, SOX9), osteoclastogenesis (TNFRSF11A, TNFSF11 and TNFRSF11B, also known as RANK, RANKL and OPG, respectively) or are causative of rare monogenic bone disorders (Figure 4). Several BMD-related SNPs are also associated with fracture risk, although these associations were detected in a smaller population.124

Only a small fraction of the genetic contribution to variance in BMD has been identified, and more studies are needed with different designs and outcome measures (such as fracture risk or heel BMD) to identify more variants and explain the ‘missing’ heritability.131–133 Use of large sample sizes will increase the power of these studies. In addition, large-scale sequencing efforts will make it possible to identify rare variants that influence osteoporosis risk. Furthermore, part of the genetic contribution of variance in BMD to osteoporosis and bone fragility can probably be explained by structural variations in DNA or by epigenetic mechanisms. Copy-number variations represent a form of structural variation in which the number of copies of a DNA fragment varies between individuals. Only few GWASs have investigated the effect of copy- number variations on osteoporosis and BMD. Two independent studies identified VPS13B131 and UGT2B17132 as susceptibility genes for osteoporosis; however, the association with UGT2B17 could not be replicated in an independent candidate-gene study.134–136 In 2014, a rare deletion in an intergenic region within the 6p25.1 locus was reported to be associated with osteoporotic fractures, although how deletion of this region influences fracture risk is unknown. It is possible that this region contains regulatory elements that affect the transcription of target genes, or that an as-yet unidentified gene is disrupted by the deletion.133

Introduction – Chapter 1

Figure 4 | Genes associated with either BMD or fracture risk.124 The genes depicted all have direct or indirect roles in the three main bone cell types (osteocytes, osteoblasts and osteoclasts), and are implicated in pathways that regulate osteoblastogenesis and osteoclastogenesis. Genes in red boxes are causative of monogenic syndromes with a skeletal phenotype. Abbreviation: BMD, bone mineral density. IV. Epigenetics and osteoporosis risk

Epigenetics refers to stable and heritable changes in phenotype or gene expression attributable to mechanisms other than changes in the underlying DNA sequence. These mechanisms are dependent on interactions between the environment and the genome. Three epigenetic mechanisms have been described: histone modification, transcriptional repression mediated by microRNAs (miRNAs) and DNA methylation.137–139 The role of epigenetics in bone homeostasis and osteoporosis risk has not been extensively studied, but the available evidence suggests that the differentiation and activity of osteoblasts and osteoclasts can be regulated by all three epigenetic mechanisms.138

Of these three mechanisms, histone modification is probably the most complex. Histone modifications are dynamic and regulated by enzymes that can either induce or reverse specific modifications.138,139 For example, the enzyme NAD-dependent protein deacetylase sirtuin 1 (encoded by SIRT1) is a histone deacetyltransferase that regulates bone remodelling by deacetylating the SOST promoter, thereby repressing expression of sclerostin.140 Decreased expression of sclerostin results in increased activity of Wnt–β-catenin signalling and bone

Introduction – Chapter 1 formation.84 In addition to sirtuin 1, several other enzymes can influence bone remodelling by regulating osteoblast and osteoclast differentiation.138

MiRNAs are small noncoding RNA molecules that are incorporated into the RNA-induced silencing complex (RISC) where they can bind to a target mRNA.138,139 RISC is a multiprotein complex that is activated upon binding of an miRNA to its complementary mRNA, which results in degradation or translational repression of the target mRNA.141 Most miRNAs are able to bind several mRNAs, and most mRNAs can be targeted by several miRNAs. Numerous miRNAs are already reported to regulate the differentiation of osteoblasts and osteoclasts, and are associated with osteoporosis. The role of miRNAs in the regulation of osteogenesis and osteoporosis has been extensively reviewed elsewhere.142,143

The third epigenetic mechanism is DNA methylation, a reversible modification of a cytosine residue located 5' to a guanosine residue (CpG). DNA methylation primarily represses gene expression by either inhibiting or facilitating binding between proteins and DNA.138,139 Several studies demonstrated that expression of some important modulators of bone remodelling is regulated by DNA methylation.138 For example, SOST is mainly expressed in osteocytes embedded in mineralized bone. Subsequent studies demonstrated that differentiation of osteoblasts to osteocytes is accompanied by decreased methylation of the SOST promoter, which facilitates osteocyte-specific expression of SOST.144 The RANKL–RANK–osteoprotegerin pathway is another important regulator of bone remodelling. RANKL and OPG are both expressed in osteoblasts, and the RANKL:osteoprotegerin ratio is at least partially regulated by the methylation status of a CpG island downstream of RANKL and a CpG island upstream of OPG.145 DNA methylation is clearly an important co-regulator of bone homeostasis; however, more studies are needed to further elucidate the contribution of DNA methylation to this process.

V. Therapeutic implications

The scientific breakthroughs realized in the past 20 years have clear implications for the prevention and treatment of osteoporosis. Several of the pathways identified in GWASs have high therapeutic potential. With regard to inhibition of bone resorption, the anti-RANKL antibody denosumab has turned out to be very effective. CTSK encodes cathepsin K, a lysosomal protease essential for bone resorption by osteoclasts. Although no associations with BMD or fracture risk were reported by GWASs or candidate-gene studies, disease-causing mutations in CTSK have been identified in patients with pycnodysostosis, a high bone mass disorder

Introduction – Chapter 1 characterized by bone fragility.146 These findings suggest that cathepsin K might be an interesting target in bone fragility disorders, including osteoporosis; moreover, the results of ongoing clinical studies indicate that odanacatib, an antibody against cathepsin K, might have therapeutic potential.147,148 Recent results of the phase III long-term odanacatib fracture trial (LOFT) indeed indicate a reduced risk of osteoporotic fractures, but also revealed that odanacatib treatment increases the risk of stroke.149-151 Further development of odanacatib as a therapy for osteoporosis has therefore been terminated. Maybe even more important is the fact that several alternatives to teriparatide (a recombinant form of PTH, which has an anabolic effect on bone) are expected to become clinically available soon. Moreover, phase II clinical trials of the monoclonal antibodies romosozumab and blosozumab, both of which target the Wnt signalling inhibitor sclerostin, have shown promising results.152,153 Using any of these agents, a substantial increase in bone formation can be generated in a short period of time.147,148 However, long-term follow-up of patients receiving these therapies, potentially in combination with an antiresorptive agent, is needed to evaluate whether the increase in bone mass can be sustained.

VI. Conclusions

Primary osteoporosis is generally considered a disease of elderly individuals, implying that prevention and treatment of this condition deserves increased attention in our ageing societies. The underlying reductions in bone mass and quality that result in an increased risk of fracture have a complex aetiology, as they are influenced by endogenous, environmental and genetic factors. To some extent, osteoporosis seems to be an unavoidable part of the human ageing process. Many processes involved in ageing of other parts of the body also contribute to the deterioration observed in the bony parts of our skeleton. In particular, intracellular oxidative stress increases with age owing to a decrease in the function of antioxidative mechanisms (including a decline in levels of sex steroids) and an increase in ROS production via mitochondrial respiratory chain leakage. The mechanisms of apoptosis and macroautophagy are also deregulated during ageing, partly as a result of increased oxidative stress. All these interconnected processes affect osteoblasts, osteocytes and osteoclasts, and result in unbalanced bone remodelling and a decrease in bone mass over time.

In addition, some environmental factors contribute to an individual’s risk of osteoporosis, which implies that this risk is, to some extent, modifiable. Adequate levels of vitamin D and calcium, a diet rich in antioxidants, and physical activity protect against osteoporosis, whereas smoking and excessive consumption of alcohol have a detrimental effect on bone. However, most of an individual’s risk of developing primary osteoporosis is explained by genetic factors. In a very

Introduction – Chapter 1 small minority of patients, osteoporosis is caused by a defect in one gene. Many such disease- causing mutations have been identified in the past two decades, and have provided important and novel insights into the mechanisms of bone metabolism and homeostasis. With the implementation of high-throughput, next-generation DNA sequencing technologies, identification of additional genes can be expected in the near future, further broadening our knowledge and understanding of this disorder. The efforts of huge consortia have already resulted in the identification of an extended number of risk variants within disease-associated genes. For the moment, the value of these data lies more in the pathogenetic relevance of these genes and pathways, rather than in providing an explanation for the diagnosis or a tool to assess the genetic risk of osteoporosis in an individual. In most patients with a diagnosis of primary osteoporosis, the underlying genetic architecture of the disorder is complex—involving a lot of variants, each with a small effect size. Moreover, the majority of the variance in risk of osteoporosis cannot yet be explained by genetics. In future, some of the missing variance might be revealed by the large-scale sequencing efforts that are currently ongoing. By sequencing the exomes (or even whole genomes) of large groups of individuals, we can expect to find rare variants that have larger effects on BMD than do any of the common variants currently studied in genetic association studies. However, whether even these novel data will lead to a genetic test that has a predictive value close to (for example) BMD measurements is uncertain.

In conclusion, many of the factors involved in primary osteoporosis have been unravelled in the past two decades, resulting not only in a better understanding of its pathogenesis, but also in broadening of the therapeutic options for patients affected by this disorder. Without a doubt, these advances will help the medical community to treat their patients with osteoporosis in a more individualized and efficient manner.

VII. Acknowledgements

The authors’ research was supported by grants G.0197.12N and G.0065.10N to W.V.H. from the Research Foundation—Flanders (FWO) and by the SYBIL (Systems biology for the functional validation of genetic determinants of skeletal diseases) project. SYBIL is funded by the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 602300 to W.V.H. In addition, E.B. holds a postdoctoral fellowship funded by the FWO.

Introduction – Chapter 1

References

1. Abrahamsen, B., van Staa, T., Ariely, R., Olson, M. & Cooper, C. Excess mortality following hip fracture: a systematic epidemiological review. Osteoporos. Int. 20, 1633–1650 (2009). 2. van den Bergh, J. P., van Geel, T. A. & Geusens, P. P. Osteoporosis, frailty and fracture: implications for case finding and therapy. Nat. Rev. Rheumatol. 8, 163–172 (2012). 3. Melton, L. J. 3rd. How many women have osteoporosis now? J. Bone Miner. Res. 10, 175–177 (1995). 4. Randell, A. et al. Direct clinical and welfare costs of osteoporotic fractures in elderly men and women. Osteoporos. Int. 5, 427–432 (1995). 5. Kanis, J. A. & Johnell, O. Requirements for DXA for the management of osteoporosis in Europe. Osteoporos. Int. 16, 229–238 (2005). 6. Bachrach, L. K. Acquisition of optimal bone mass in childhood and adolescence. Trends Endocrinol. Metab. 12, 22–28 (2001). 7. Sapir-Koren, R. & Livshits, G. Osteocyte control of bone remodeling: is sclerostin a key molecular coordinator of the balanced bone resorption-formation cycles? Osteoporos. Int. (2014). 8. Zuo, C. et al. Osteoblastogenesis regulation signals in bone remodeling. Osteoporos. Int. 23, 1653–1663 (2012). 9. Garnero, P. The contribution of collagen crosslinks to bone strength. Bonekey Rep. 1, 182 (2012). 10. Alliston, T. Biological regulation of bone quality. Curr. Osteoporos. Rep. 12, 366–375 (2014). 11. Sroga, G. E. & Vashishth, D. Effects of bone matrix proteins on fracture and fragility in osteoporosis. Curr. Osteoporos. Rep. 10, 141–150 (2012). 12. Bouxsein, M. L. & Karasik, D. Bone geometry and skeletal fragility. Curr. Osteoporos. Rep. 4, 49–56 (2006). 13. Almeida, M. Aging mechanisms in bone. Bonekey Rep. 1, 102 (2012). 14. Manolagas, S. C. From estrogen-centric to aging and oxidative stress: a revised perspective of the pathogenesis of osteoporosis. Endocr. Rev. 31, 266–300 (2010). 15. de Boer, J. et al. Premature aging in mice deficient in DNA repair and transcription. Science 296, 1276–1279 (2002). 16. Tyner, S. D. et al. p53 mutant mice that display early ageing-associated phenotypes. Nature 415, 45–53 (2002).

Introduction – Chapter 1

17. Nojiri, H. et al. Cytoplasmic superoxide causes bone fragility owing to low-turnover osteoporosis and impaired collagen cross-linking. J. Bone Miner. Res. 26, 2682–2694 (2011). 18. Smietana, M. J., Arruda, E. M., Faulkner, J. A., Brooks, S. V. & Larkin, L. M. Reactive oxygen species on bone mineral density and mechanics in Cu, Zn superoxide dismutase (Sod1) knockout mice. Biochem. Biophys. Res. Commun. 403, 149–153 (2010). 19. Ambrogini, E. et al. FoxO-mediated defense against oxidative stress in osteoblasts is indispensable for skeletal homeostasis in mice. Cell Metab. 11, 136–146 (2010). 20. Almeida, M. Unraveling the role of FoxOs in bone—insights from mouse models. Bone 49, 319–327 (2011). 21. Almeida, M., Han, L., Martin-Millan, M., O’Brien, C. A. & Manolagas, S. C. Oxidative stress antagonizes Wnt signaling in osteoblast precursors by diverting β-catenin from T cell factor- to forkhead box O-mediated transcription. J. Biol. Chem. 282, 27298–27305 (2007). 22. Essers, M. A. et al. Functional interaction between β-catenin and FOXO in oxidative stress signaling. Science 308, 1181–1184 (2005). 23. Migliaccio, E. et al. The p66shc adaptor protein controls oxidative stress response and life span in mammals. Nature 402, 309–313 (1999). 24. Nemoto, S. & Finkel, T. Redox regulation of forkhead proteins through a p66shc-dependent signaling pathway. Science 295, 2450–2452 (2002). 25. Giorgio, M. et al. Electron. transfer between cytochrome c and p66Shc generates reactive oxygen species that trigger mitochondrial apoptosis. Cell 122, 221–233 (2005). 26. Pacini, S. et al. p66SHC promotes apoptosis and antagonizes mitogenic signaling in T cells. Mol. Cell Biol. 24, 1747–1757 (2004). 27. Bartell, S. M. et al. Deletion of the redox amplifier p66shc decreases ROS production in murine bone and increases osteoblast resistance to oxidative stress and bone mass [abstract]. J. Bone Miner. Res. 26 (Suppl. 1), S85 (2011). 28. Maier, B. et al. Modulation of mammalian life span by the short isoform of p53. Genes Dev. 18, 306–319 (2004). 29. Wang, X. et al. p53 functions as a negative regulator of osteoblastogenesis, osteoblast- dependent osteoclastogenesis, and bone remodeling. J. Cell Biol. 172, 115–125 (2006). 30. Jilka, R. L. et al. Dysfunctional osteocytes increase RANKL and promote cortical pore formation in their vicinity: a mechanistic explanation for the development of cortical porosity with age [abstract]. J. Bone Miner. Res. 27 (Suppl. 1), S348 (2012). 31. Jilka, R. L. et al. Dysapoptosis of osteoblasts and osteocytes increases cancellous bone formation but exaggerates bone porosity with age. J. Bone Miner. Res. 29, 103–117 (2014).

Introduction – Chapter 1

32. Tomkinson, A., Gevers, E. F., Wit, J. M., Reeve, J. & Noble, B. S. The role of estrogen in the control of rat osteocyte apoptosis. J. Bone Miner. Res. 13, 1243–1250 (1998). 33. Almeida, M. et al. Skeletal involution by age-associated oxidative stress and its acceleration by loss of sex steroids. J. Biol. Chem. 282, 27285–27297 (2007). 34. Tomkinson, A., Reeve, J., Shaw, R. W. & Noble, B. S. The death of osteocytes via apoptosis accompanies estrogen withdrawal in human bone. J. Clin. Endocrinol. Metab. 82, 3128–3135 (1997). 35. Frost, H. M. Micropetrosis. J. Bone Joint Surg. Am. 42-A, 144–150 (1960). 36. Busse, B. et al. Decrease in the osteocyte lacunar density accompanied by hypermineralized lacunar occlusion reveals failure and delay of remodeling in aged human bone. Aging Cell 9, 1065–1075 (2010). 37. Carpentier, V. T. et al. Increased proportion of hypermineralized osteocyte lacunae in osteoporotic and osteoarthritic human trabecular bone: implications for bone remodeling. Bone 50, 688–694 (2012). 38. Manolagas, S. C., O’Brien, C. A. & Almeida, M. The role of estrogen and androgen receptors in bone health and disease. Nat. Rev. Endocrinol. 9, 699–712 (2013). 39. Seeman, E. Periosteal bone formation—a neglected determinant of bone strength. N. Engl. J. Med. 349, 320–323 (2003). 40. Martin-Millan, M. et al. The estrogen receptor-α in osteoclasts mediates the protective effects of estrogens on cancellous but not cortical bone. Mol. Endocrinol. 24, 323–334 (2010). 41. Nakamura, T. et al. Estrogen prevents bone loss via estrogen receptor α and induction of Fas ligand in osteoclasts. Cell 130, 811–823 (2007). 42. Chiang, C. et al. Mineralization and bone resorption are regulated by the androgen receptor in male mice. J. Bone Miner. Res. 24, 621–631 (2009). 43. Notini, A. J. et al. Osteoblast deletion of exon 3 of the androgen receptor gene results in trabecular bone loss in adult male mice. J. Bone Miner. Res. 22, 347–356 (2007). 44. Sinnesael, M. et al. Androgen receptor (AR) in osteocytes is important for the maintenance of male skeletal integrity: evidence from targeted AR disruption in mouse osteocytes. J. Bone Miner. Res. 27, 2535–2543 (2012). 45. Almeida, M. et al. Estrogens attenuate oxidative stress and the differentiation and apoptosis of osteoblasts by DNA-binding-independent actions of the ERα. J. Bone Miner. Res. 25, 769–781 (2010). 46. Kousteni, S. et al. Nongenotropic, sex-nonspecific signaling through the estrogen or androgen receptors: dissociation from transcriptional activity. Cell 104, 719–730 (2001).

Introduction – Chapter 1

47. Kousteni, S. et al. Reversal of bone loss in mice by nongenotropic signaling of sex steroids. Science 298, 843–846 (2002). 48. Marathe, N., Rangaswami, H., Zhuang, S., Boss, G. R. & Pilz, R. B. Pro-survival effects of 17β-estradiol on osteocytes are mediated by nitric oxide/cGMP via differential actions of cGMP-dependent protein kinases I and II. J. Biol. Chem. 287, 978–988 (2012). 49. Seeman, E. Pathogenesis of bone fragility in women and men. Lancet 359, 1841–1850 (2002). 50. Kaufman, J. M. & Vermeulen, A. The decline of androgen levels in elderly men and its clinical and therapeutic implications. Endocr. Rev. 26, 833–876 (2005). 51. Shahnazari, M. et al. Bone turnover markers in peripheral blood and marrow plasma reflect trabecular bone loss but not endocortical expansion in aging mice. Bone 50, 628–637 (2012). 52. Syed, F. A. et al. Effects of chronic estrogen treatment on modulating age-related bone loss in female mice. J. Bone Miner. Res. 25, 2438–2446 (2010). 53. Zebaze, R. M. et al. Intracortical remodelling and porosity in the distal radius and post- mortem femurs of women: a cross-sectional study. Lancet 375, 1729–1736 (2010). 54. Manolagas, S. C. & Parfitt, A. M. For whom the bell tolls: distress signals from long-lived osteocytes and the pathogenesis of metabolic bone diseases. Bone 54, 272–278 (2013). 55. Hocking, L. J., Whitehouse, C. & Helfrich, M. H. Autophagy: a new player in skeletal maintenance? J. Bone Miner. Res. 27, 1439–1447 (2012). 56. Kim, K. H. & Lee, M. S. Autophagy—a key player in cellular and body metabolism. Nat. Rev. Endocrinol. 10, 322–337 (2014). 57. Zhang, L. et al. Pathway-based genome-wide association analysis identified the importance of regulation-of-autophagy pathway for ultradistal radius BMD. J. Bone Miner. Res. 25, 1572–1580 (2010). 58. Rubinsztein, D. C., Marino, G. & Kroemer, G. Autophagy and aging. Cell 146, 682–695 (2011). 59. Selman, C. et al. Ribosomal protein S6 kinase 1 signaling regulates mammalian life span. Science 326, 140–144 (2009). 60. Chang, Y. Y. et al. Nutrient-dependent regulation of autophagy through the target of rapamycin pathway. Biochem. Soc. Trans. 37, 232–236 (2009). 61. DeSelm, C. J. et al. Autophagy proteins regulate the secretory component of osteoclastic bone resorption. Dev. Cell 21, 966–974 (2011). 62. Liu, F. et al. Suppression of autophagy by FIP200 deletion leads to osteopenia in mice through the inhibition of osteoblast terminal differentiation. J. Bone Miner. Res. 28, 2414–2430 (2013).

Introduction – Chapter 1

63. Onal, M. et al. Suppression of autophagy in osteocytes mimics skeletal aging. J. Biol. Chem. 288, 17432–17440 (2013). 64. Bonjour, J. P., Kraenzlin, M., Levasseur, R., Warren, M. & Whiting, S. Dairy in adulthood: from foods to nutrient interactions on bone and skeletal muscle health. J. Am. Coll. Nutr. 32, 251–263 (2013). 65. Christodoulou, S., Goula, T., Ververidis, A. & Drosos, G. Vitamin D and bone disease. Biomed. Res. Int. 2013, 396541 (2013). 66. Eisman, J. A. & Bouillon, R. Vitamin D: direct effects of vitamin D metabolites on bone: lessons from genetically modified mice. Bonekey Rep. 3, 499 (2014). 67. Rajendran, P. et al. Antioxidants and human diseases. Clin. Chim. Acta 436C, 332–347 (2014). 68. Shen, C. L. et al. Fruits and dietary phytochemicals in bone protection. Nutr. Res. 32, 897–910 (2012). 69. Li, J. J. et al. Fruit and vegetable intake and bone mass in Chinese adolescents, young and postmenopausal women. Public Health Nutr. 16, 78–86 (2013). 70. New, S. A. et al. Dietary influences on bone mass and bone metabolism: further evidence of a positive link between fruit and vegetable consumption and bone health? Am. J. Clin. Nutr. 71, 142–151 (2000). 71. Prynne, C. J. et al. Fruit and vegetable intakes and bone mineral status: a cross sectional study in 5 age and sex cohorts. Am. J. Clin. Nutr. 83, 1420–1428 (2006). 72. Zalloua, P. A. et al. Impact of seafood and fruit consumption on bone mineral density. Maturitas 56, 1–11 (2007). 73. Heaney, R. P. & Layman, D. K. Amount and type of protein influences bone health. Am. J. Clin. Nutr. 87, 1567S–1570S (2008). 74. Sebastian, A. Dietary protein content and the diet’s net acid load: opposing effects on bone health. Am. J. Clin. Nutr. 82, 921–922 (2005). 75. Bonjour, J. P. Nutritional disturbance in acid-base balance and osteoporosis: a hypothesis that disregards the essential homeostatic role of the kidney. Br. J. Nutr. 110, 1168–1177 (2013). 76. Bushinsky, D. A. Metabolic alkalosis decreases bone calcium efflux by suppressing osteoclasts and stimulating osteoblasts. Am. J. Physiol. 271, F216–F222 (1996). 77. Leeuwenburgh, C. & Heinecke, J. W. Oxidative stress and antioxidants in exercise. Curr. Med. Chem. 8, 829–838 (2001). 78. Ozcivici, E. et al. Mechanical signals as anabolic agents in bone. Nat. Rev. Rheumatol. 6, 50–59 (2010).

Introduction – Chapter 1

79. Rochefort, G. Y., Pallu, S. & Benhamou, C. L. Osteocyte: the unrecognized side of bone tissue. Osteoporos. Int. 21, 1457–1469 (2010). 80. Balemans, W. et al. Increased bone density in sclerosteosis is due to the deficiency of a novel secreted protein (SOST). Hum. Mol. Genet. 10, 537–543 (2001). 81. Balemans, W. et al. Identification of a 52 kb deletion downstream of the SOST gene in patients with van Buchem disease. J. Med. Genet. 39, 91–97 (2002). 82. Li, X. et al. Sclerostin binds to LRP5/6 and antagonizes canonical Wnt signaling. J. Biol. Chem. 280, 19883–19887 (2005). 83. Semenov, M., Tamai, K. & He, X. SOST is a ligand for LRP5/LRP6 and a Wnt signaling inhibitor. J. Biol. Chem. 280, 26770–26775 (2005). 84. Boudin, E., Fijalkowski, I., Piters, E. & Van Hul, W. The role of extracellular modulators of canonical Wnt signaling in bone metabolism and diseases. Semin. Arthritis Rheum. 43, 220–240 (2013). 85. Wang, Y. et al. Wnt and the Wnt signaling pathway in bone development and disease. Front. Biosci. (Landmark Ed.) 19, 379–407 (2014). 86. Maurel, D. B., Boisseau, N., Benhamou, C. L. & Jaffre, C. Alcohol and bone: review of dose effects and mechanisms. Osteoporos. Int. 23, 1–16 (2012). 87. Ronis, M. J., Mercer, K. & Chen, J. R. Effects of nutrition and alcohol consumption on bone loss. Curr. Osteoporos. Rep. 9, 53–59 (2011). 88. Chen, J. R. et al. A role for ethanol-induced oxidative stress in controlling lineage commitment of mesenchymal stromal cells through inhibition of Wnt/β-catenin signaling. J. Bone Miner. Res. 25, 1117–1127 (2010). 89. Chen, J. R., Shankar, K., Nagarajan, S., Badger, T. M. & Ronis, M. J. Protective effects of estradiol on ethanol-induced bone loss involve inhibition of reactive oxygen species generation in osteoblasts and downstream activation of the extracellular signal-regulated kinase/signal transducer and activator of transcription 3/receptor activator of nuclear factor- κB ligand signaling cascade. J. Pharmacol. Exp. Ther. 324, 50–59 (2008). 90. Kanis, J. A. et al. Smoking and fracture risk: a meta-analysis. Osteoporos. Int. 16, 155–162 (2005). 91. Ward, K. D. & Klesges, R. C. A meta-analysis of the effects of cigarette smoking on bone mineral density. Calcif. Tissue Int. 68, 259–270 (2001). 92. Yoon, V., Maalouf, N. M. & Sakhaee, K. The effects of smoking on bone metabolism. Osteoporos. Int. 23, 2081–2092 (2012).

Introduction – Chapter 1

93. Marini, J. C. & Blissett, A. R. New genes in bone development: what’s new in osteogenesis imperfecta. J. Clin. Endocrinol. Metab. 98, 3095–3103 (2013). 94. Sillence, D. O., Senn, A. & Danks, D. M. Genetic heterogeneity in osteogenesis imperfecta. J. Med. Genet. 16, 101–116 (1979). 95. OMIM® Online Mendelian Inheritance in Man®. An Online Catalog of Human Genes and Genetic Disorders [online], http://www.omim.org (2015). 96. Van Dijk, F. S. & Sillence, D. O. Osteogenesis imperfecta: Clinical diagnosis, nomenclature and severity assessment. Am. J. Med. Genet. A 164, 1470–1481 (2014). 97. Lazarus, S., Zankl, A. & Duncan, E. L. Next-generation sequencing: a frameshift in skeletal dysplasia gene discovery. Osteoporos. Int. 25, 407–422 (2014). 98. Warman, M. L. et al. Nosology and classification of genetic skeletal disorders: 2010 revision. Am. J. Med. Genet. A 155A, 943–968 (2011). 99. van Dijk, F. S. et al. EMQN best practice guidelines for the laboratory diagnosis of osteogenesis imperfecta. Eur. J. Hum. Genet. 20, 11–19 (2012). 100. Lazarus, S., Moffatt, P., Duncan, E. L. & Thomas, G. P. A brilliant breakthrough in OI type V. Osteoporos. Int. 25, 399–405 (2014). 101. Eyre, D. R. & Weis, M. A. Bone collagen: new clues to its mineralization mechanism from recessive osteogenesis imperfecta. Calcif. Tissue Int. 93, 338–347 (2013). 102. Gebken, J. et al. Increased cell surface expression of receptors for transforming growth factor-β on osteoblasts from patients with Osteogenesis imperfecta. Pathobiology 68, 106–112 (2000). 103. Grafe, I. et al. Excessive transforming growth factor-β signaling is a common mechanism in osteogenesis imperfecta. Nat. Med. 20, 670–675 (2014). 104. Lapunzina, P. et al. Identification of a frameshift mutation in Osterix in a patient with recessive osteogenesis imperfecta. Am. J. Hum. Genet. 87, 110–114 (2010). 105. Keupp, K. et al. Mutations in WNT1 cause different forms of bone fragility. Am. J. Hum. Genet. 92, 565–574 (2013). 106. Laine, C. M. et al. WNT1 mutations in early-onset osteoporosis and osteogenesis imperfecta. N. Engl. J. Med. 368, 1809–1816 (2013). 107. Fahiminiya, S. et al. Whole-exome sequencing reveals a heterozygous LRP5 mutation in a 6-year-old boy with vertebral compression fractures and low trabecular bone density. Bone 57, 41–46 (2013). 108. Korvala, J. et al. Mutations in LRP5 cause primary osteoporosis without features of OI by reducing Wnt signaling activity. BMC Med. Genet. 13, 26 (2012).

Introduction – Chapter 1

109. Crabbe, P. et al. Missense mutations in LRP5 are not a common cause of idiopathic osteoporosis in adult men. J. Bone Miner. Res. 20, 1951–1959 (2005). 110. van Dijk, F. S. et al. PLS3 mutations in X-linked osteoporosis with fractures. N. Engl. J. Med. 369, 1529–1536 (2013). 111. Fahiminiya, S. et al. Osteoporosis caused by mutations in PLS3: clinical and bone tissue characteristics. J. Bone Miner. Res. 29, 1805–1814 (2014). 112. Laine, C. M. et al. Primary osteoporosis without features of OI in children and adolescents: clinical and genetic characteristics. Am. J. Med. Genet. A 158A, 1252–1261 (2012). 113. Gong, Y. et al. LDL receptor-related protein 5 (LRP5) affects bone accrual and eye development. Cell 107, 513–523 (2001). 114. Narumi, S. et al. Various types of LRP5 mutations in four patients with osteoporosis- pseudoglioma syndrome: identification of a 7.2-kb microdeletion using oligonucleotide tiling microarray. Am. J. Med. Genet. A 152A, 133–140 (2010). 115. Balemans, W. & Van Hul, W. The genetics of low-density lipoprotein receptor-related protein 5 in bone: a story of extremes. Endocrinology 148, 2622–2629 (2007). 116. Puig-Hervas, M. T. et al. Mutations in PLOD2 cause autosomal-recessive connective tissue disorders within the Bruck syndrome—osteogenesis imperfecta phenotypic spectrum. Hum. Mutat. 33, 1444–1449 (2012). 117. Shaheen, R. et al. Mutations in FKBP10 cause both Bruck syndrome and isolated osteogenesis imperfecta in humans. Am. J. Med. Genet. A 155A, 1448–1452 (2011). 118. Schwarze, U. et al. Mutations in FKBP10, which result in Bruck syndrome and recessive forms of osteogenesis imperfecta, inhibit the hydroxylation of telopeptide lysines in bone collagen. Hum. Mol. Genet. 22, 1–17 (2013). 119. Peacock, M., Turner, C. H., Econs, M. J. & Foroud, T. Genetics of osteoporosis. Endocr. Rev. 23, 303–326 (2002). 120. Siris, E. S. et al. Identification and fracture outcomes of undiagnosed low bone mineral density in postmenopausal women: results from the National Osteoporosis Risk Assessment. JAMA 286, 2815–2822 (2001). 121. Duncan, E. L. & Brown, M. A. Clinical review 2: Genetic determinants of bone density and fracture risk—state of the art and future directions. J. Clin. Endocrinol. Metab. 95, 2576–2587 (2010). 122. Ralston, S. H. & Uitterlinden, A. G. Genetics of osteoporosis. Endocr. Rev. 31, 629–662 (2010).

Introduction – Chapter 1

123. Duncan, E. L. et al. Genome-wide association study using extreme truncate selection identifies novel genes affecting bone mineral density and fracture risk. PLoS Genet. 7, e1001372 (2011). 124. Estrada, K. et al. Genome-wide meta-analysis identifies 56 bone mineral density loci and reveals 14 loci associated with risk of fracture. Nat. Genet. 44, 491–501 (2012). 125. Medina-Gomez, C. et al. Meta-analysis of genome-wide scans for total body BMD in children and adults reveals allelic heterogeneity and age-specific effects at the WNT16 locus. PLoS Genet. 8, e1002718 (2012). 126. Richards, J. B. et al. Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet 371, 1505–1512 (2008). 127. Rivadeneira, F. et al. Twenty bone-mineral-density loci identified by large-scale meta-analysis of genome-wide association studies. Nat. Genet. 41, 1199–1206 (2009). 128. Styrkarsdottir, U. et al. Multiple genetic loci for bone mineral density and fractures. N. Engl. J. Med. 358, 2355–2365 (2008). 129. Styrkarsdottir, U. et al. European bone mineral density loci are also associated with BMD in East-Asian populations. PLoS ONE 5, e13217 (2010). 130. Zhang, L. et al. Multistage genome-wide association meta-analyses identified two new loci for bone mineral density. Hum. Mol. Genet. 23, 1923–1933 (2014). 131. Moayyeri, A. et al. Genetic determinants of heel bone properties: genome-wide association meta-analysis and replication in the GEFOS/GENOMOS consortium. Hum. Mol. Genet. 23, 3054–3068 (2014). 132. Oei, L. et al. Genome-wide association study for radiographic vertebral fractures: a potential role for the 16q24 BMD locus. Bone 59, 20–27 (2014). 133. Oei, L. et al. A genome-wide copy number association study of osteoporotic fractures points to the 6p25.1 locus. J. Med. Genet. 51, 122–131 (2014). 134. Chew, S. et al. Homozygous deletion of the UGT2B17 gene is not associated with osteoporosis risk in elderly Caucasian women. Osteoporos. Int. 22, 1981–1986 (2011). 135. Deng, F. Y. et al. Genome-wide copy number variation association study suggested VPS13B gene for osteoporosis in Caucasians. Osteoporos. Int. 21, 579–587 (2010). 136. Yang, T. L. et al. Genome-wide copy-number-variation study identified a susceptibility gene, UGT2B17, for osteoporosis. Am. J. Hum. Genet. 83, 663–674 (2008). 137. Holroyd, C., Harvey, N., Dennison, E. & Cooper, C. Epigenetic influences in the developmental origins of osteoporosis. Osteoporos. Int. 23, 401–410 (2012).

Introduction – Chapter 1

138. Vrtacnik, P., Marc, J. & Ostanek, B. Epigenetic mechanisms in bone. Clin. Chem. Lab. Med. 52, 589–608 (2014). 139. Delgado-Calle, J., Garmilla, P. & Riancho, J. A. Do epigenetic marks govern bone mass and homeostasis? Curr. Genomics 13, 252–263 (2012). 140. Cohen-Kfir, E. et al. Sirt1 is a regulator of bone mass and a repressor of Sost encoding for sclerostin, a bone formation inhibitor. Endocrinology 152, 4514–4524 (2011). 141. Redfern, A. D. et al. RNA-induced silencing complex (RISC) proteins PACT, TRBP, and Dicer are SRA binding nuclear receptor coregulators. Proc. Natl Acad. Sci. USA 110, 6536– 6541 (2013). 142. van Wijnen, A. J. et al. MicroRNA functions in osteogenesis and dysfunctions in osteoporosis. Curr. Osteoporos. Rep. 11, 72–82 (2013). 143. Gamez, B., Rodriguez-Carballo, E. & Ventura, F. MicroRNAs and post-transcriptional regulation of skeletal development. J. Mol. Endocrinol. 52, R179–197 (2014). 144. Delgado-Calle, J. et al. DNA methylation contributes to the regulation of sclerostin expression in human osteocytes. J. Bone Miner. Res. 27, 926–937 (2012). 145. Delgado-Calle, J. et al. Role of DNA methylation in the regulation of the RANKL-OPG system in human bone. Epigenetics 7, 83–91 (2012). 146. Gelb, B. D., Shi, G. P., Chapman, H. A., & Desnick, R. J. Pycnodysostosis, a lysosomal disease caused by cathepsin K deficiency. Science 273, 1236–1238 (1996). 147. Fijalkowski, I., Boudin, E., Mortier, G. & van Hul, W. Sclerosing bone dysplasias: leads toward novel osteoporosis treatments. Curr. Osteoporos. Rep. 12, 243–251 (2014). 148. Tella, S. H. & Gallagher, J. C. Biological agents in management of osteoporosis. Eur. J. Clin. Pharmacol. (2014). 149. McClung, M. R. et al. Odanacatib efficacy and safety in postmenopausal women with osteoporosis: 5-year data from the extension of the Phase 3 Long-Term Odanacatib Fracture Trial (LOFT). J Bone Miner Res 31 (Suppl 1). Available at www.asbmr.org/ education/AbstractDetail?aid=adae99f0-f992-4da6-9c42-6c6d0173dad6.

150. Papapoulos, S. et al. Safety of Odanacatib in postmenopausal women with osteoporosis: 5- year data from the extension of the Phase 3 Long-Term Odanacatib Fracture Trial (LOFT). J Bone Miner Res 31 (Suppl 1). Available at www.asbmr.org/education/ AbstractDetail?aid=a0d5b4c1-17da-4e35-ba08-af3f7e4d76c6.

151. O’Donoghue, M. et al. The Long-Term Odanacatib Fracture Trial (LOFT): cardiovascular safety results. J Bone Miner Res 31 (Suppl 1). Available at www.asbmr.org/education/AbstractDetail?aid=8c23bf70-634f-4d37-9414-42b6b54a369c.

Introduction – Chapter 1

152. McClung, M. R. et al. Romosozumab in postmenopausal women with low bone mineral density. N. Engl. J. Med. 370, 412–420 (2014). 153. Recker, R. et al. A randomized, double-blind phase 2 clinical trial of blosozumab, a sclerostin antibody, in postmenopausal women with low bone mineral density. J. Bone Miner. Res. 30, 216–224 (2015).

Introduction – Chapter 1

Chapter 2

An Introduction to WNT Signaling

Introduction – Chapter 2

As a result of decades of research, the Wingless-type mouse mammary tumor virus integration site (WNT) signaling pathway has become a focal point in developmental biology and biomedical research. As a group of signal transduction pathways, it plays an important role in general embryonic development, the regeneration of tissues and the regulation of a variety of fundamental cellular processes like cell proliferation, migration and apoptosis1,2. Traditionally, the WNT signaling pathway is subdivided in three distinctive WNT pathways; one canonical or β- catenin dependent (WNT/β-catenin) pathway and two non-canonical or β-catenin independent pathways (WNT/Calcium and WNT/Planar Cell Polarity (PCP) pathway). Activation of all WNT signaling pathways classically starts with the binding of one of the 19 WNT proteins (WNTs) to one of the 10 G-protein coupled Frizzled (FZ) receptors1,3. Throughout the years, alternative (co- )receptors, inhibitors and more and more intracellular players of all three WNT signaling pathways were identified, increasing the complexity of this WNT signaling pathway. Nevertheless, numerous research efforts have yet led to a better understanding of the WNT pathway in light of many research fields and vice versa.

While activation of the non-canonical WNT signaling pathways mainly occurs through the binding of a WNT ligand to a FZ receptor, activation of the β-catenin dependent WNT pathway additionally requires the low-density lipoprotein receptor-related protein 5 (LRP5) or LRP6 co-receptor (Figure 1). As soon as the WNT-FZ-LRP5/6 complex is formed, a β-catenin destruction complex consisting of adenomatous polyposis coli (APC), glycogen synthase kinase (GSK)-3β and Axin dissociates and β-catenin stabilizes in the cytoplasm. This will cause local accumulation of β-catenin and subsequent translocation to the nucleus where it will bind the T- cell factor/lymphoid enhancer factor (TCF/LEF) transcription factors for the transcription of canonical target genes. In absence of LRP5/6 or a WNT ligand, the destruction complex remains intact and GSK-3β will phosphorylate β-catenin, thereby labelling it for ubiquitinylation and subsequent degradation by the proteasome. Activation of the WNT/Calcium pathway after WNT and FZ binding will result in an increase in intracellular calcium (Ca2+) levels through the activation of phospholipase C (PLC). This will then influence Ca2+-sensitive enzymes in the cytoplasm, like protein kinase C (PKC), calcineurin (CaN) and Ca2+/calmodulin-dependent protein kinase type II (CaMKII). Activation of these enzymes can then result in transcription of target genes by the nuclear factor of activated T-cells (NFAT) transcription factor. At last, activation of the WNT/PCP pathway leads to the activation of RhoA and Rac small GTPases in the cytoplasm, thereby activating Rho-associated protein kinase (ROCK) and c-jun terminal kinase (JNK), which activate the activator protein 1 (AP-1) transcription factor and can attain cytoskeletal and cell movement changes1-3.

Introduction

54 Figure 1 | The WNT signaling

pathway consists of three different pathways (WNT/β- catenin; WNT/PCP;

WNT/Calcium), all requiring

–

the binding of a WNT ligand 2 Chapter with a Frizzled receptor, whereas WNT/β-catenin signaling additionally requires a LRP5/6 co-receptor. WNT/β- catenin signaling activation is characterized by an accumulation of β-catenin in the cytoplasm, while WNT/PCP signaling typically activates RhoA and Rac GTPases and WNT/Calcium causes an increase in intracellular calcium through PLC. Finally, activating these pathways will result in the activation of the TCF/LEF, AP-1 and NFAT transcription factors. WNT signaling inhibition occurs through mechanisms of WNT sequestration (WIF-1, SFRP) or by targeting LRP5/6. When formation of the WNT- LRP5/6-Frizzled complex is disturbed, a destruction complex (Axin, APC, GSK-3β, β-catenin) assembles, labeling β- catenin for degradation by the

proteasome.

Introduction – Chapter 2

On the other hand, inhibition of WNT signaling can occur at different levels in the pathway by numerous inhibitors. For example, extracellular inhibition of WNT signaling through the sequestration of WNT ligands is done by secreted frizzled-related proteins (SFRPs) and WNT inhibitory factor 1 (WIF-1), via direct binding to WNTs preventing them from interacting with FZ receptors (Figure 1). This mechanism can inhibit all three WNT signaling pathways, while other extracellular modulators, like sclerostin and Dickkopf (DKK), solely inhibit the WNT/β- catenin pathway. Sclerostin and DKK both bind LRP5/6 and prevent the co-receptor from forming the WNT-FZ-LRP5/6 complex required for activation of this pathway (Figure 1). Here, the LRP4 receptor acts as an anchor for sclerostin on the membrane to mediate inhibition, whereas the Kremen receptor will help DKK with the internalization of LRP5/61,3. Altogether, as a complex and diverse system, the mechanisms of activation and inhibition of WNT signaling are tightly regulated to achieve balanced WNT signaling activity in every type of cell context and its specific needs at any time.

As in many other research fields, WNT signaling is also a major topic in the understanding of skeletal homeostasis and the development of skeletal disorders. Here, performing genetic research was and still is a tool to not only understand the general genetic background of bone mineral density (BMD), but also to identify the disease causing genes for both monogenic and complex bone disorders. For example, some sclerosing bone disorders (sclerosteosis, Van Buchem disease…), characterized by an increased bone formation, are caused by mutations in the SOST, LRP4 or LRP5 genes, which are all part of the WNT/β-catenin signaling pathway. Moreover, loss-of-function mutations in the same LRP5 gene were found in patients with osteoporosis-pseudoglioma syndrome, which is characterized by severe thinning of the bones2. At last, Robinow syndrome, a severe skeletal dysplasia, is caused by mutations in the receptor tyrosine kinase-like orphan receptor (ROR) 2, WNT5A and dishevelled-1 (DVL1) gene, all part of non- canonical WNT signaling1,4. Altogether, fundamental alterations in all parts of WNT signaling can give rise to a broad spectrum of skeletal disorders, implying a striking relevance to skeletal health.

Besides a clear role in rare, monogenic bone disorders, players of the WNT signaling pathway also contribute to the development of common bone disorders, like osteoporosis. To explore the complex genetic nature of osteoporosis, many candidate gene and genome wide association studies (GWAS) have been performed the past decade. To date, these studies identified ±60 loci of which many genes cluster within key pathways in bone metabolism: the RANK-RANKL- OPG pathway, mesenchymal stem cell differentiation and the WNT signaling pathway5. Thereafter, newly associated genes that are part of one of these key pathways have been

Introduction – Chapter 2 interesting targets for in vitro and in vivo functional studies to further elucidate their role in bone homeostasis and the pathogenesis of bone disorders as osteoporosis.

Introduction – Chapter 2

References

1. Rao, T.P. & Kuhl, M. An updated overview on Wnt signaling pathways: a prelude for more. Circ Res 106, 1798-806 (2010). 2. Baron, R. & Kneissel, M. WNT signaling in bone homeostasis and disease: from human mutations to treatments. Nat Med 19, 179-92 (2013). 3. Boudin, E., Fijalkowski, I., Piters, E. & Van Hul, W. The role of extracellular modulators of canonical Wnt signaling in bone metabolism and diseases. Semin Arthritis Rheum 43, 220-40 (2013). 4. Bunn, K.J. et al. Mutations in DVL1 cause an osteosclerotic form of Robinow syndrome. Am J Hum Genet 96, 623-30 (2015). 5. Hendrickx, G., Boudin, E. & Van Hul, W. A look behind the scenes: the risk and pathogenesis of primary osteoporosis. Nat Rev Rheumatol 11, 462-74 (2015).

Introduction – Chapter 2

OBJECTIVES

Objectives

Bone mineral density (BMD) is a continuous skeletal parameter that exhibits a Gaussian distribution in a normal, healthy population (Figure 1). BMD is also widely used as a surrogate phenotype for several common and rare bone disorders, like osteoporosis1. Largely due to the substantial role of BMD in the pathogenesis of osteoporosis, the genetic nature of osteoporosis is also complex. Just as well, there are bone disorders with deviating BMD values caused by only one mutation with a large impact on the skeleton. Unlike osteoporosis, these monogenic bone disorders are mostly rare and often result in an extreme high (sclerosing bone disorders) or low BMD (osteogenesis imperfecta,…)2,3. Altogether, these bone disorders (with a low vs. high BMD, common vs. rare, complex vs. monogenic) encompass a multifaceted spectrum (Figure 1). A uniform way to better understand this plethora of bone disorders is to perform genetic and subsequent functional studies as one gene can be involved in the pathogenesis of different sorts of bone disorders. The identification of a gene (co-)responsible for one disorder can thus support knowledge on bone homeostasis in general and benefit the selection of interesting candidate genes in the study, prevention and/or treatment of other bone disorders2,4.

Figure 1. Frequency and genetic nature of (extreme) low and high bone mass. Due to the complex nature of bone mass (BMD), the majority of people have a normal BMD, but some will present with a pathological lowered bone mass (osteoporosis) with a complex genetic nature (green, full line). Though, an extreme low or high BMD is far less frequent and will probably have a monogenic background (orange, full line), resulting in an osteoporotic or sclerosing bone disorder. Nevertheless, there are exceptions (green or orange, dotted lines), like Hyperostosis Cranialis Interna, a monogenic bone disorder characterized by extreme bone overgrowth in the skull, but the remainder of the skeleton is unaffected and therefore categorized as normal.

Objectives

The aim of this thesis is therefore to further elucidate the regulation of BMD by performing genetic and functional research on bone disorders. To achieve this goal, two different bone disorders were selected, both requiring a different methodological approach.

I. Osteoporosis is in most cases a common, complex and low BMD bone disorder. Here, we will verify the role of three specific candidate genes (WNT16, WNT4 and WNT5B) in the pathogenesis of osteoporosis. These three genes were associated with BMD in past genetic studies and are all part of a well-acknowledged signaling pathway in bone metabolism, the WNT signaling pathway.

The results of the analysis of the WNTs are discussed in Part 1 of this thesis. In Chapter 3 and 5, the results of a thorough genetic evaluation of WNT16 and WNT4/5B are discussed, respectively. In Chapter 3 and 4, the results of the follow-up functional analysis of WNT16 are presented.

II. Hyperostosis Cranialis Interna is, contrary to osteoporosis, a monogenic and very rare bone disorder characterized by bone overgrowth exclusively at the level of the skull. The aim here is to identify the disease causing gene and better understand the pathogenesis and the underlying mechanism regulating bone metabolism by additionally performing functional experiments.

The results of the genetic identification, in vitro and in vivo functional validation of SLC39A14 (ZIP14) as disease-causing gene for Hyperostosis Cranialis Interna are presented in Part 2 (Chapter 6) of this thesis.

In conclusion, by investigating two nearly opposite bone disorders by means of genetic research, we jointly aim to increase the overall knowledge on the regulation of BMD, highlight important signaling pathways (and players) and identify possible novel therapeutic targets for the treatment of bone disorders.

Objectives

References

1. Boudin, E., Fijalkowski, I., Hendrickx, G. & Van Hul, W. Genetic control of bone mass. Mol Cell Endocrinol (2015). 2. Fijalkowski, I., Boudin, E., Mortier, G. & Van Hul, W. Sclerosing bone dysplasias: leads toward novel osteoporosis treatments. Curr Osteoporos Rep 12, 243-51 (2014). 3. Hendrickx, G., Boudin, E. & Van Hul, W. A look behind the scenes: the risk and pathogenesis of primary osteoporosis. Nat Rev Rheumatol 11, 462-74 (2015). 4. Appelman-Dijkstra, N.M. & Papapoulos, S.E. From disease to treatment: from rare skeletal disorders to treatments for osteoporosis. Endocrine 52, 414-26 (2016).

Objectives

RESULTS

Part 1

Osteoporosis

Chapter 3

Variation in the Kozak Sequence of WNT16 Results in an Increased Translation and is Associated with Osteoporosis Related Parameters

Gretl Hendrickx1, Eveline Boudin1, Igor Fijałkowski, Torben Leo Nielsen2, Marianne Andersen2, Kim Brixen2 and Wim Van Hul1

1Department of Medical Genetics, University of Antwerp, Belgium

2Department of Endocrinology, Odense University Hospital, Denmark

This chapter is based on the research article published in: Bone 2014; 59:57-65

Results – Chapter 3

Abstract

The importance of WNT16 in the regulation of bone metabolism was recently confirmed by several genome-wide association studies and by a Wnt16 (Wnt16-/-) knockout mouse model. The aim of this study was thus to replicate and further elucidate the effect of common genetic variation in WNT16 on osteoporosis related parameters. Hereto, we performed a WNT16 candidate gene association study in a population of healthy Caucasian men from the Odense Androgen Study (OAS). Using HapMap, five tagSNPs and one multimarker test were selected for genotyping to cover most of the common genetic variation in and around WNT16 (MAF>5%). This study confirmed previously reported associations for rs3801387 and rs2707466 with bone mineral density (BMD) at several sites. Furthermore, we additionally demonstrated that rs2908007 is strongly associated with BMD at several sites in the young, elderly and complete OAS population. The observed effect of these three associated SNPs on the respective phenotypes is comparable and we can conclude that the presence of the minor allele results in an increase in BMD. Additionally, we performed re-sequencing of WNT16 on two cohorts selected from the young OAS cohort, based on their extreme BMD values. On this basis, rs55710688 was selected for an in vitro translation experiment since it is located in the Kozak sequence of WNT16a. We observed an increased translation efficiency and thus a higher amount of WNT16a for the Kozak sequence that was significantly more prevalent in the high BMD cohort. This observation is in line with the results of the Wnt16-/- mice. Finally, a WNT luciferase reporter assay was performed and showed no activation of the β-catenin dependent pathway by Wnt16. We did detect a dose-dependent inhibitory effect of Wnt16 on WNT1 activation of this canonical WNT pathway. Increased translation of WNT16 can thus lead to an increased inhibitory action of WNT16 on canonical WNT signaling. This statement is in contrast with the known activating effect of canonical WNT signaling on bone formation and suggests a stimulatory effect on bone metabolism via non-canonical WNT signaling. More research is required to not only confirm this hypothesis, but also to further elucidate the role of non-canonical WNT pathways in bone metabolism and the general mechanisms of interplay between the different WNT signaling pathways.

Results – Chapter 3

I. Introduction

Osteoporosis is a common disorder affecting hundreds of millions of people worldwide and causing millions of fractures every year. It is characterized by a decreased bone mineral density (BMD) and deteriorated micro-architectural bone structure. Subsequently, osteoporosis patients have a significant higher fracture risk, particularly at the level of hip, wrist and vertebrae. Treatment, hospitalization and follow-up of these fractures are expensive, making preventive therapies very important. Hereto, several antiresorptive bone therapies are available, but many of them have numerous disadvantages. The medical community has therefore been on the lookout for anabolic treatments, of which daily injections of teriparatide is the only one available so far1.

Osteoporosis is a complex disease and its susceptibility depends on both genetic and environmental factors. Twin and family studies confirmed that 50-85% of the variation in peak bone mass is due to genetic variation2,3. These twin studies also state that both the pathogenesis of osteoporosis and geometric bone parameters have genetic features as well3. Nevertheless, this genetic influence is polygenic in most cases of osteoporosis, unlike a number of monogenic sclerosing bone disorders shown to be monogenic. In the latter, SOST, LRP4 and LRP5, all three genes of the Wingless-type mouse mammary tumor virus integration site (WNT) signaling pathway, were found to be disease causing4-6. These findings implicated that the WNT signaling pathway plays a major role in bone homeostasis. To elucidate the genetic architecture of a common disorder as osteoporosis, several candidate gene and genome wide association studies (GWAS) have been performed. In a recent multicenter study, 56 loci have been associated with bone mineral density, explaining 5.8% of the total genetic variance in femoral neck BMD (FN- BMD)7. Furthermore, 14 BMD loci have also been associated with risk for fracture. Among these identified genes, several members of three key biological pathways were present: the RANK- RANKL-OPG pathway, mesenchymal stem cell differentiation and the WNT signaling pathway. As to WNT signaling, WNT16 was one of the newly associated members, next to associations with previously known genes as SOST, LRP4 and LRP5.

The WNT signaling pathway regulates a variety of cellular processes and plays an important role in the embryonic development and the regeneration of tissues. A family of 19 WNT secreted cysteine-rich glycoproteins can bind the Frizzled (FZ) receptors on the cell surface and subsequently activate at least three different WNT signaling pathways, two non-canonical pathways and one canonical or β-catenin dependent pathway. When activated, the canonical pathway results in the stabilization and accumulation of β-catenin in the cytoplasm. Eventually, this will lead to the translocation of β-catenin to the nucleus and transcription of target genes.

Results – Chapter 3

For the non-canonical pathways, the process after WNT ligand binding is more diverse and β- catenin independent. Since SOST, LRP4 and LRP5 are all three involved in the β-catenin dependent pathway, the role and importance of this pathway in bone has already been studied widely8. On the contrary, less is known about the role of the non-canonical pathways and about the way all three signaling pathways work together or against one another in the context of bone homeostasis. Since WNT proteins have a central place in the complete WNT signaling pathway, research on these proteins can create new insights on these topics.

Recent association studies demonstrated that genetic variation in WNT16 is associated with BMD, hip geometry parameters and fracture risk7,9-11. The importance of WNT16 in the regulation of bone metabolism was also confirmed by a Wnt16 (Wnt16-/-) knockout (KO) mouse model. These mice have a significantly reduced total body BMD, thinner cortical bones at the femur midshaft and reduced bone strength of both femur and tibia10,11.

These findings highlight WNT16 as an interesting member of the WNT signaling pathway in the context of bone research. The aim of this study was thus to replicate and further elucidate the effect of common genetic variation in WNT16 on a broad range of osteoporosis related parameters, such as BMD measurements and bone geometry parameters. We therefore performed a WNT16 candidate gene association study on healthy Caucasian men of the Odense Androgen Study (OAS). Based on the BMD t-scores of the young OAS cohort, high and low BMD subgroups from the cohort were also selected for re-sequencing. To further investigate the functional role of WNT16 in vitro, a luciferase reporter assay was performed to examine the effect of WNT16 on the β-catenin dependent pathway. Also, an in vitro translation experiment was performed to investigate whether the difference in BMD between the high and low BMD cohort could be correlated with variation in the translation efficiency of the WNT16 gene.

II. Material and methods 1. Association study 1.1 The Odense Androgen Study

The Odense Androgen Study (OAS) is a population-based, prospective, observational study on the interrelationship between endocrine status, body composition, muscle function and bone metabolism in men. It consists of two population based cohorts: a young cohort of 783 non- stratified men aged 20-29 years and an older cohort of 600 age-stratified men aged 60-74 years. Details on the design and inclusion of the participants in both cohorts were previously reported

Results – Chapter 3 elsewhere12-14. Characteristics and results of DXA-derived data for both the young and the elderly cohort and also for both cohorts combined were reported previously.

1.2 Body composition and bone mineral density

Body weight was measured in light clothing (SECA, DE) to the nearest 0.1 kg while height was determined to the nearest 0.1 cm (Harpenden, Holtain, UK). BMD was measured at the femoral neck (FN) and the lumbar spine (LS) (L2-L4) by dual energy X-ray absorptiometry (DXA) using a Hologic 4500 device (Hologic, Inc., Waltham, MA, USA). Software version 2.3.1 (Hologic, Inc.) was used to calculate total hip (TH) and whole body (WB) BMD.

1.3 Hip structural analysis (HSA)

Geometry parameters at the proximal femur were obtained using the APEX 2.3.1 software, an application for DXA (Hologic, Inc.). This software provides data on hip axis length (HAL) and neck-shaft angle (NSA); and measures the buckling ratio (BR), the cross-sectional area (CSA) and the cross-sectional moment of inertia (CSMI). The strength parameter that is based on the CSMI and is shown instead is the section modulus (Z). These parameters respectively represent structural stability, axial bone strength and bending strength. There are three different regions of interest (ROI) at which the various geometric parameters are calculated. The ROI called narrow neck (NN) is placed at the narrowest portion of the femoral neck. The intertrochanteric (IT) ROI is a bisector of the intersection of the femoral neck and shaft axes. The last ROI, the shaft region (FS, femoral shaft), is placed at the bottom of the global ROI. The default placement of the global ROI is 1 cm below the lesser trochanter.

1.4 SNP selection and genotyping

Selection of the tagSNPs was performed based on data of the CEU population (Utah residents with Northern and Western European ancestry from the CEPH collection) provided by the HapMap project (Rel28 PhaseII+III, August10, on NCBI B36 assembly, dbSNP b126) (http://www.hapmap.org). All common (minor allele frequency (MAF) ≥ 0.05) polymorphisms in the WNT16 gene were included and using the aggressive tagger mode of Haploview we selected five tagSNPs and one multimarker test which cover all common genetic variation in WNT16 (Table 1).

All tagSNPs were genotyped using real-time polymerase chain reaction (PCR) with Taqman allelic discrimination probes (Applied Biosystems, Foster City, CA, USA) according to instructions by the kit manufacturers on the LightCycler 480 system (Roche Applied Science, Mannheim,

Results – Chapter 3

Germany). For all SNPs, at least 5% of the samples were genotyped in duplicate and additional positive and negative control samples were included. The genotyping success rate was over 98% for all SNPs.

Table 1 | Overview of the selected tagSNPs, tagged SNPs (r²≥0.8), minor allele frequency (MAF) and Hardy-Weinberg Equilibrium (HWE). Capital letter indicates the major allele and small letter indicates the minor allele.

HapMap SNPs OASY OASE OAS TagSNP CEU (r²≥0.80) MAF HWE MAF HWE MAF HWE MAF rs3801387 rs3801387 0.25 0.983 0.27 0.954 0.26 0.972 0.26 (T/c) rs3779381 rs3801385 rs3801385 0.11 0.950 0.12 0.999 0.11 0.978 0.09 (A/g) rs2908007 rs2908007 0.37 0.977 0.40 0.999 0.38 0.987 0.38 (T/c) rs2707466 rs2536189 0.45 0.972 0.45 0.954 0.45 0.996 0.41 (G/a) rs2908004 rs2707466 rs2536184 rs2707471 0.22 0.974 0.19 0.997 0.21 0.985 0.15 (T/c) rs2707473 rs2536184 rs2536192 rs2707460 rs2707516 rs2707474 rs2041490 rs2707467 rs2707461 rs4727922 rs2707469 rs2908007, rs3757552 0.13 0.956 0.14 0.984 0.14 0.981 0.12 rs3801387 (T/c)

1.5 Statistics

It was previously shown that the OAS cohort has at least 80% power to detect small variance effects (0.6-0.8%) for all the studied phenotypes and tagSNPs (MAF>0.05) regardless the mode of inheritance, and with a type I error rate probability of 0.0512. Hardy–Weinberg equilibrium (HWE) was tested by the utility program of J. Ott (http://linkage.rockefeller.edu/ott/linkutil.htm) using a standard Chi-square test comparing the expected and observed allele frequencies. The significance threshold for HWE was set to 0.01. The multi-marker test was analyzed using the Plink software (version 1.07)

Results – Chapter 3

(http://pngu.mgh.harvard.edu/purcell/plink/). Plink was used to impute the haplotypes based on multi-marker predictors using the standard expectation-maximization (EM) algorithm15.

Linear regression analysis, adjusted for age, height and weight, was performed using SPSS version 20.0 for (SPSS, Chicago, IL, USA) as previously described for the complete OAS population as well as for the different age groups separately12. All analyses were performed in both an additive and a recessive model, since it was previously shown that testing the additive model alone performs poorly for recessive causal alleles when the MAF is not close to 50%16. P-values less than 0.05 were considered as nominally significant. Correction for multiple testing was performed using the Bonferroni method for the number of SNPs tested. Finally, to determine whether associated SNPs have independent effects conditional analyses were performed using a likelihood ratio test.

1.6 Re-sequencing

Based on the t-score, the individuals of the young OAS cohort with the most extreme BMD values (t-score <-1.38, n=64 and t-score >1.54, n=64) were selected for re-sequencing of the coding regions using Sanger sequencing. These cutoffs were based on Sims et al. (z-score <-1.5 and z-score>1.5) and afterwards we made sure that both BMD groups were equal in size17. Selection of the amplicons and primer design (Primer3 (http://frodo.wi.mit.edu/primer3/input.htm)) were based on the template sequence available in NCBI (www.ncbi.nlm.nih.gov/gene). Primers are designed for four amplicons which comprise all coding exons and the corresponding intron/exon boundaries of all different protein coding transcripts.

Amplification of the selected amplicons was performed using GoTaq DNA polymerase-mediated PCR (Promega Corporation, Madison, WI) and verified by agarose gel electrophoresis simultaneously running a Generuler 100bp Plus DNA ladder (Fermentas International Inc, Ontario, Canada). Primer sequences and PCR conditions are available on request. After amplification by PCR, primers and unincorporated dNTPs were removed using exonuclease I (New England Biolabs, Inc, Ipswich, Massachusetts, USA) and calf intestine alkaline phosphatase (CIAP, Roche Applied Science, Hoffmann–La Roche AG, Basel, Switzerland). Sequencing was carried out directly on purified fragments with the ABI 310 Genetic Analyser (Applied Biosystems, Foster City, California, USA), using an ABI Prism BigDye terminator cycle sequencing ready reaction kit, version 1.1 (Applied Biosystems, Foster City, California, USA). The primers used here were identical to the primers that were used for amplification. The BigDye

Results – Chapter 3

XTerminator purification kit was used as purification method for DNA sequencing with the purpose of removing unincorporated BigDye terminators.

Finally, a Chi² test was used to compare the genotype frequencies of all detected polymorphisms between the high and the low BMD cohort (SPSS version 20.0 for (SPSS, Chicago, IL, USA)). P- values less than 0.05 were considered as significant.

2. Functional evaluation 2.1 WNT luciferase reporter assay

The human embryonic kidney (HEK) cell line 293T was grown in DMEM (Invitrogen) supplemented with FBS (10% v/v, Invitrogen). Twenty-four hours prior to transfection, cells were plated at 0.3 x 105 cells/well in 96-well plates. Cells were transfected using Fugene 6 (Roche Applied Science) according to the manufacturer’s instructions. Mouse Wnt1-V5 (0.4ng), mesdc2 (0.8ng), human wildtype LRP5 (0.8ng), pRL-TK (1ng) and pGL3-OT (20ng) constructs were cotransfected in HEK293T cells. Details on these expression constructs were previously reported elsewhere 18. Instead of LRP5, HEK293T cells were transfected with human wildtype LRP6 (0.8ng, OriGene Technologies, Rockville, MD) as well. Furthermore, different amounts of the pCMV6-Entry vector containing the murine Wnt16 cDNA (0.2-0.4-0.8ng, Origene Technologies, Rockville, MD) were added to the transfection mixture. When needed, empty pcDNA3.1 vector was added to make total DNA amount equal for all transfection experiments 18. Each transfection was carried out in triplicate and repeated independently in three separate experiments. Forty-eight hours after transfection, cells were lysed and firefly and renilla luciferase activity were measured on a Glomax Multi+ Luminometer (Turner Designs, Sunyvale, CA) using the dual luciferase reporter assay system (Promega Corporation, Madison, WI) according to manufacturer’s instructions. The ratio of the firefly and renilla luciferase measurement was calculated. Data are expressed as mean values ± SD. Comparison between two measurements for a single experiment was performed using a Student’s t-test. Values of p<0.05 were considered significant. Statistical test were provided by the SPSS 20.0 software package (SPSS Inc).

2.2 In vitro translation assay

Site-directed mutagenesis was performed following the manufacturer’s instructions on a human WNT16 cDNA clone of the NM_016087.2 transcript variant (WNT16a, OriGene Technologies, Rockville, MD) with the use of the QuickChange II Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA). To investigate the translational effect of the rs55710688 variance in the Kozak sequence, two different WNT16a cDNA clones were generated with this insertion (-/CCCA)

Results – Chapter 3 polymorphism resulting in two different alleles (GCACCC and GCACCCACCC). The underlined nucleotides represent the two different Kozak sequences. As a positive control, ideal translation conditions were created by generating a third WNT16a cDNA clone with the optimal Kozak sequence (GCCACC) 19. Plasmid from single colonies was isolated using the PureYield Plasmid Miniprep System (Promega Corporation, Madison, WI) following the vacuum protocol.

Cell-free transcription/translation experiments were performed using the TNT Quick Coupled Transcription/Translation System (Promega Corporation, Madison, WI). Hereby, 1µg of plasmid DNA was added to 40µl of TNT Quick Master Mix. Methionine (1µl) and Transcend Biotin- Lysyl-tRNAs (1.5µl) were added separately. The synthesis reactions were performed at 30°C for 90 minutes. Also, a parallel reaction with no DNA was performed as a negative control. Analysis of translation products was done by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Hereto, 2µl of each sample was mixed with 4µl of LDS sample buffer (Life Technologies Corporation) and 1.6µl of NuPAGE reducing agent (Life Technologies Corporation). After denaturation, protein translation was verified by SDS-PAGE on a 12% NuPAGE Bis-Tris precast gel (Life Technologies Corporation) simultaneously running a Novex Sharp Pre-Stained Protein Standard Ladder (Life Technologies Corporation). Afterwards, the translation reaction products were blotted to a Protran BA79 nitrocellulose membrane (GE Healthcare Life Sciences) using a Trans-Blot SD Semi-Dry Transfer Cell (Bio-Rad). Blocking of the membrane was done by incubation of the membrane in Tris-buffered Saline (TBS) with 0.5% Tween20 for 1 hour. The membrane was then incubated with a Streptavidin-HRP conjugate (Promega Corporation, Madison, WI) for 1 hour and subsequently washed three times with TBS + 0.5% Tween®20 and three times with water. In order to detect the protein translation products the membrane was incubated with the Transcend™ Chemiluminescent Substrate mixture (Promega Corporation, Madison, WI) for 1 minute and exposed to a Amersham Hyperfilm ECL (GE Healthcare Life Sciences) for 28 minutes. Band density was quantified using the ChemiDoc XRS+ System (Bio-Rad) and the Image Lab Software (Bio-Rad).

III. Results 1. Association study 1.1 Genotyping

Based on the data of the CEU population in HapMap, the location of the SNPs and based on data of previous studies, 5 tagSNPs and 1 multimarker test were selected which cover the genetic variation in and around WNT16 (Table 1). All tagSNPs were genotyped in the young and elderly cohort of the OAS population with a success rate over 98%. All genotype frequencies were

Results – Chapter 3 compatible with HWE (p>0.01) (Table 1). The MAFs of all studied SNPs were fairly similar in both our cohorts and with the MAFs calculated for the CEU population from the HapMap database.

All SNPs were tested for their effect on several BMD and hip geometry parameters both in the young and elderly OAS cohort separately as well as in the combined analysis of both cohorts using multiple linear regression analysis corrected for age, height and weight.

1.2 Age-dependent analysis

In the young OAS (OASY) cohort, multiple linear regression corrected for age, height and body weight resulted in significant associations for rs3801387 and rs2908007 with BMD of the whole body (WB), lumbar spine (LS) and femoral neck (FN) and with the buckling ratio (BR) at the narrowest portion of the femoral neck (the narrow neck) (NN). In addition, we demonstrated that rs3801387 had a significant effect on total hip (TH) BMD (Table 2). After Bonferroni correction for multiple testing, rs3801387 was still significantly associated with WB and FN BMD and rs2908007 was significantly associated with FN BMD (Table 2). The presence of the minor allele of both associated SNPs resulted in an increase in BMD and a decrease in buckling ratio indicating a reduced fracture risk.

Table 2 | Results from the multiple linear regression analysis in the young OAS cohort (n=783) for BMD and hip geometry parameters. P-values are shown in bold, regression coefficients and 95% confidence intervals in italic. Significant associations were found in an additive model and in a recessive test model (underlined). Associations withstanding Bonferroni correction (p<0.01) are indicated with asterisks. Ns: not significant (p>0.05).

rs3801387, rs3801387 rs3801385 rs2908007 rs2707466 rs2536184 rs2908007

0.002* 0.011 0.049 WB BMD 0.018 Ns 0.013 0.015 Ns Ns (0.006; 0.029) (0.003; 0.024) (0.000; 0.030) 0.026 TH BMD 0.017 Ns Ns Ns Ns Ns (0.002; 0.033) 0.007* 0.006* FN BMD 0.020 Ns 0.019 Ns Ns Ns (0.006; 0.035) (0.005; 0.032) 0.015 0.017 LS BMD 0.017 Ns 0.015 Ns Ns Ns (0.003; 0.031) (0.003; 0.028) 0.048 0.036 NN BR -0.213 Ns -0.202 Ns Ns Ns (-0.423;-0.002) (-0.390; -0.013)

Results – Chapter 3

Table 3 | Results from the multiple linear regression analysis in the elderly OAS cohort (n=600) for BMD and hip geometry parameters. P-values are shown in bold, regression coefficients and 95% confidence intervals in italic. Significant associations were found in an additive model and in a recessive test model (underlined). Associations withstanding Bonferroni correction (p<0.01) are indicated with asterisks. Ns: not significant (p>0.05).

rs3801387, rs3801387 rs3801385 rs2908007 rs2707466 rs2536184 rs2908007 0.035 0.028 WB BMD Ns Ns 0.013 0.014 Ns Ns (0.001; 0.026) (0.002; 0.027) 0.015 0.025 TH BMD Ns 0.112 Ns Ns Ns 0.079 (0.022; 0.203) (0.010; 0.148) 0.010 0.027 FN BMD Ns 0.108 Ns Ns Ns 0.071 (0.026; 0.190) (0.008; 0.133) 0.038 HAL Ns Ns Ns -0.730 Ns Ns (-1.419;-0.042) 0.009* 0.024 NN CSA Ns 0.467 Ns Ns Ns 0.308 (0.120; 0.815) (0.041; 0.574) 0.028 0.033 NN BR Ns -1.989 Ns Ns Ns -1.472 (-3.760; -0.218) (-2.822; -0.123) 0.041 FS Z Ns 0.305 Ns Ns Ns Ns (0.013; 0.597) 0.003* FS BR Ns Ns Ns -0.139 Ns Ns (-0.231; -0.046)

Next, we performed multiple linear regression analysis in the elderly OAS (OASE) cohort. In this cohort we could not replicate the results found in the OASY cohort for rs3801387 and rs2908007, except for the results of these SNPs and WB BMD (Table 3). In addition, in an additive test model we also demonstrated that rs2707466 was significantly associated with WB BMD after correction for age, height and body weight. Furthermore in the same model rs2707466 was associated with hip axis length (HAL) and FS BR. The latter association resisted Bonferroni correction (p<0.01) for the number of tested SNPs. The effect of both associated SNPs on BMD was similar to that seen for the associations in the OASY cohort. In addition, the minor allele of rs2707466 resulted in a decrease of the HAL and of the BR. Besides above mentioned associations found in an additive model, we also found significant associations in a recessive test model. This model demonstrated that rs3801385 and the multi-marker test were significantly associated with TH and FN BMD and with cross-sectional area (CSA) and BR at the NN. At last, rs3801385 was also associated with sectional modulus of the femoral shaft. Only the association of rs3801385 with CSA at the NN remained significant in the recessive model after Bonferroni correction for the number of tested SNPs (Table 3).

Results – Chapter 3

1.3 Combined analysis

Finally, we combined the data of the young and the elderly cohort and we performed multiple linear regression analysis. Prior to testing the effect of the selected SNPs on the phenotypes of interest, we investigated if the SNPs had an age specific effect on one of the phenotypes. When a significant age-specific effect (A.E.) was observed, the combined analysis could not be performed (Table 4).

A significant age-specific effect was found for rs2536184 on WB BMD, CSA at the NN and FS and for NN BR (Table 4). Besides this, multiple linear regression analysis demonstrated that rs3801387, rs2908007 and rs2707466 were significantly associated with WB and FN BMD. Moreover, rs3801387 and rs2908007 were also associated with LS BMD and this all when tested in an additive model. In a recessive model we only found a significant association for rs3801385 with NN BR. Only the associations of rs3801387 and rs2908007 with WB BMD, rs3801387 with LS BMD and the association of rs2908007 with FN BMD remained significant after correction for multiple testing (Table 4).

Table 4 | Results from the linear regression analysis in the complete OAS cohort for BMD and hip geometry parameters. P-values are shown in bold, regression coefficients and 95% confidence intervals in italic. Significant associations were found in an additive model and in a recessive test model (underlined). Associations withstanding Bonferroni correction (p<0.01) are indicated with asterisks. A.E.: Significant age-specific effect (p<0.05) and its p-values between brackets. Ns: not significant (p>0.05).

rs3801387, rs3801387 rs3801385 rs2908007 rs2707466 rs2536184 rs2908007 0.002* 0.001* 0.012 WB BMD 0.014 Ns 0.013 0.010 A.E. (0.014) Ns (0.005;0.023) (0.005; 0.021) (0.002-0.018) TH BMD Ns Ns Ns Ns Ns Ns 0.017 0.003* 0.045 FN BMD 0.013 Ns 0.015 0.009 Ns Ns (0.002; 0.024) (0.005; 0.024) (0.000; 0.019) 0.006* 0.011 LS BMD 0.018 Ns 0.015 Ns Ns Ns (0.005; 0.030) (0.003; 0.026) NN CSA Ns Ns Ns Ns A.E. (0.01) Ns 0.024 NN BR Ns -1.390 Ns Ns A.E. (0.003) Ns (-2.594; -0.185) FS Z A.E. (0.043) Ns Ns Ns Ns A.E. (0.027)

FS BR Ns Ns Ns A.E. (0.015) Ns Ns

Results – Chapter 3

Rs3801387 is located in intron 3 of the WNT16 gene and based on the data of HapMap it tags rs3779381 which is also located in an intron (Table 1). The other significantly associated SNP, rs2908007, is located upstream of WNT16 and does not tag any other SNPs according to the HapMap data (Table 1). The minor alleles of both these significantly associated SNPs had a similar effect on BMD. Since both SNPs are in moderate linkage disequilibrium (LD) with each other (data OASY Figure 1, OAS data not shown), we performed a conditional analysis. Hereafter, there was no longer a significant (p<0.05) association between rs2908007 and the corresponding BMD parameters. Both SNPs and their associations were thus dependent of the same causative variant.

Figure 1 | Physical map and LD structure of WNT16. The upper part of the figure represents both transcripts of WNT16 (WNT16a, WNT16b) and their surrounding 5’ and 3’ regions located on the 7q31 locus. Exons of both transcripts are shown as boxes; coding regions are shaded dark grey and light grey boxes represent untranslated regions. The lower part of the figure shows the LD structure of 15 polymorphisms in and surrounding WNT16, according to genotype data of the association study and re- sequencing from two cohorts with a higher and lower BMD. The genotyped tagSNPs and the multimarker test (rs2536184) are shown in italic. Pair-wise r² values are shown in boxes 20. Dark boxes represent high pair-wise r² values of SNPs which are in strong LD with each other; light grey boxes represent weak LD. When no number is shown a pair-wise r² of zero is alluded.

Results – Chapter 3

1.4 Re-sequencing

In order to further investigate the effect of genetic variation in WNT16 on BMD, we selected 128 individuals with the most extreme t-scores for whole body BMD. Using Sanger sequencing we screened all 128 individuals for genetic variation in or nearby all coding regions of both WNT16 transcripts. In this sample cohort, we found 10 known polymorphisms of which three polymorphisms had a significant (p<0.05) difference in genotype frequencies between the high and low BMD cohorts (Table 5). The first SNP (rs17143281) is located upstream of WNT16 and was found more frequently in individuals with a higher BMD. The other two variants are in moderate LD with rs3801387 (Figure 1), which has previously shown to be associated with BMD in the OASY and the OAS population (Table 2 and 4). The rs142005327 polymorphism involves the insertion of two nucleotides (CT) in intron 2 of both WNT16 transcripts and was also found more in the individuals with higher t-scores. The last polymorphism, rs55710688, is located in exon 1 of one of the transcripts of WNT16 (NM_016087.2, WNT16a). This insertion (-/CCCA) polymorphism influences the Kozak sequence located in the UTR of exon 1 by a G>C substitution at position -6. Previous studies already demonstrated that alterations in the Kozak sequence can influence the translation efficiency of a protein and can be disease causing 21-23. As consequence, this variant is interesting for further functional investigations.

2. Functional evaluation 2.1 In vitro translation assay

To further investigate the functional effect of rs55710688 variance in the Kozak sequence of the WNT16a (NM_016087.2) transcript variant, a cell-free transcription/translation experiment was performed. Post-translational analysis showed bands of different densities, all with a molecular weight of approximately 40kDa, identified as WNT16a (NP_057171.2) (Figure 2). Here, a higher amount of WNT16a was observed for the construct with a GCACCCACCC Kozak sequence (rs55710688; CCCA/CCCA) compared to the construct with a GCACCC Kozak sequence (rs55710688; -/-). This result refers to increased translation efficiency, caused by a G to C substitution at position -6 of the Kozak sequence. Also, an even higher amount of WNT16a was detected for the plasmids with the optimal Kozak sequence (GCCACC), indicating an even higher translation efficiency, as expected (Figure 2).

Results – Chapter 3

Table 5 | Genotype frequencies of known variants in the exons and intron/exon boundaries of WNT16 in individuals of the young OAS cohort with the most extreme BMD values (t-score<- 1.38 and t-score>1.54). Variants with a significant difference in genotype frequencies (p <0.05) between the high and low BMD cohorts are indicated with asterisks.

Position Rs-number Genotypes Genotype frequencies Transcript t-score>1.54 t-score<-1.38 5’ UTR rs17143281* CC 0.83 0.95 WNT16a CT 0.17 0.05 TT 0.00 0.00 5’ UTR rs79813790 CC 1.00 0.98 WNT16a CT 0.00 0.02 TT 0.00 0.00 Exon 1 rs55710688* -/- 0.48 0.70 WNT16a CCCA/- 0.45 0.28 CCCA/CCCA 0.06 0.02 Intron 1 rs4727920 AA 0.92 0.97 WNT16a AG 0.08 0.03 GG 0.00 0.00 Intron 1 rs17143285 TT 0.97 0.91 WNT16a TA 0.03 0.09 AA 0.00 0.00 Exon 1 rs35391640 GG 0.94 0.94 WNT16b GC 0.06 0.06 CC 0.00 0.00 Exon 2 rs2908004 GG 0.22 0.17 WNT16a/b GA 0.55 0.42 AA 0.23 0.39 Intron 2 rs142005327* -/- 0.52 0.77 WNT16a/b CT/- 0.45 0.22 CT/CT 0.03 0.02 Exon 3 rs74389152 CC 0.97 0.98 WNT16a/b CA 0.03 0.02 AA 0.00 0.00 Exon 4 rs2707466 CC 0.25 0.38 WNT16a/b CT 0.55 0.45 TT 0.20 0.17

Results – Chapter 3

Figure 2 | Rs55710688 variance in the Kozak sequence of WNT16a (NM_016087.2) influences translation efficiency. The left part of the figure presents an in vitro translation experiment that was performed on three WNT16a cDNA constructs with different Kozak sequences (underlined), showing different amounts of WNT16a protein (40kDa). This difference in translation efficiency is due to rs55710688 variance (CCCA insertion, italic), which causes a G to C substitution on position -6 of the Kozak sequence. GCCACC was the Kozak sequence used as a positive control, since optimal translation efficiency has been reported for this sequence19. As a negative control, no cDNA was added to this sample. The right part of the figure shows a graph in which the measured band densities are presented.

2.2 Wnt luciferase reporter assay

Previous functional and genetic studies demonstrated that the canonical WNT signaling pathway is an important regulator of bone formation. In order to evaluate the effect of WNT16 on canonical WNT signaling a WNT reporter assay with Wnt16 was performed. This experiment demonstrates that, after cotransfection with LRP5 or LRP6, Wnt16 (0.4-0.8-1.6ng) does not activate the TCF-response element of the TOPflash reporter construct, unlike Wnt1 (Figure 3). Consequently, Wnt16 does not activate canonical WNT signaling. Further, when the same amount of Wnt1 (0.4ng) was cotransfected with raising amounts of Wnt16 (0.2-0.4-0.8ng) a significantly (p<0.05) decreased activation of the TCF-response element was observed (Figure 3). This refers to a dominant inhibitory effect of Wnt16 on the β-catenin dependent WNT pathway, which is also stronger in the presence of LRP6 than in presence of LRP5.

Results – Chapter 3

Figure 3 | Wnt16 inhibits WNT1 induced activity of the β-catenin dependent WNT signaling pathway in a dose-dependent manner. Canonical WNT signaling is stimulated by WNT1 in presence and absence of LRP5 or LRP6. Cotransfection of Wnt1 with Wnt16 inhibits this stimulatory effect in a dose-dependent manner with significant (p<0.05) differences between all three dosages of Wnt16. The inhibitory effect of Wnt16 is also stronger in presence of LRP6 than in presence of LRP5.

IV. Discussion

Recently, several independent GWAS, with the goal to detect candidate genes underlying osteoporosis, revealed that polymorphisms in the WNT16 gene are associated with femoral neck, lumbar spine and whole body BMD, bone strength, cortical bone thickness and fracture risk7,9-11. The importance of WNT16 in the regulation of BMD and bone strength was also confirmed by Wnt16 knockout (Wnt16-/-) mice which have reduced whole body BMD, cortical bone thickness and bone strength10,11. Even though it is clear that WNT16 is involved in the regulation of bone metabolism and the susceptibility for osteoporosis, the causative variant is not yet found and the effect of WNT16 on the different WNT pathways is still unclear.

We investigated the effect of all common genetic variation in and around WNT16 (10kb 5’-5kb 3’) on a range of BMD and hip geometry parameters in a population of Danish men from the Odense Androgen Study. This study confirmed previously reported associations for rs3801387 and rs2707466 with BMD at several sites7,9,10. In our study, the associations found by GWAS of

Results – Chapter 3 rs3801387 with femoral neck, lumbar spine and whole body BMD were all three confirmed in the young OAS cohort and the complete OAS cohort. The absence of significant associations in the elderly OAS cohort could possibly be due to reduced power, since no age-specific effect could be detected. In comparison with the association study performed by García-Ibarbia et al.9 only mildly significant associations for rs2707466 were detected in our population. García-Ibarbia et al. also found strong associations of rs2908004 with BMD parameters9. Nevertheless, we did not detect a significant difference in genotype frequencies for rs2908004 between the young Danish men with a lower and higher BMD during re-sequencing. Also, Fig. 1 shows just a r²-value of 0.11 between rs2707466 and rs2908004 for these cohorts with an extreme BMD. Since García-Ibarbia et al.9 used a population of 1083 individuals over 49 years of age, and we only re-sequenced the young OAS cohort (20-29 years), the effect of rs2908004 could be more prominent on later age and thus not picked up by our study method. Furthermore, next to rs3801387 and rs2707466, we additionally demonstrated that rs2908007 is strongly associated with BMD at several sites in the young, elderly and complete OAS population. The observed effect of all three associated SNPs on the respective phenotypes was comparable and we can conclude that the presence of the minor allele results in an increase in BMD.

To further evaluate the influence of genetic variation in WNT16 on BMD, both a high and low BMD subgroup were selected from the young OAS population for re-sequencing of the exons and intron/exon boundaries. Hereafter, a LD plot was made based on the genotypes of the tagSNPs and sequenced polymorphisms from both BMD cohorts. Based on this LD plot a moderate level of LD was detected between rs3801387 and rs2908007 and between rs3801387 and rs2707466. Since there was, after conditional analysis, no longer a significant association of rs2908007 and rs2707466, we conclude that the effect of both SNPs depends on the same causative SNP. Further, by comparing genotype frequencies of all detected SNPs in both BMD cohorts, other interesting SNPs in LD with rs3801387 were identified. One of these, rs55710688 (r²=0.75), was interesting for functional experiments since it is located in the Kozak sequence of WNT16a (NM_016087.2) and has a relative high minor allele frequency (MAF=0.22). Moreover, rs55710688 is an insertion polymorphism, making it impossible to be detected by previous performed candidate gene and genome-wide association studies. Two transcripts, WNT16a or WNT16b (NM_057168.1), are found for the WNT16 gene. WNT16a is expressed in the pancreas while WNT16b is expressed in several tissues. Currently, no data are available for bone. It was previously shown that variation in the Kozak sequence can influence the translation efficiency of a protein which even can be disease causing21-23. Therefore, for WNT16a, an in vitro translation experiment was performed. Here, we observed increased translation efficiency and thus a higher

Results – Chapter 3 amount of WNT16a for the Kozak sequence that was significantly more prevalent in the high BMD cohort. Since rs55710688, causing the Kozak sequence variation, is in high LD (r²= 0.75) with rs3801387, a higher amount of the WNT16a protein can be correlated with the higher bone mass of the subjects. Moreover, this observation is in line with the results of the Wnt16/- mice, showing a reduced total body BMD, thinner cortical bones at the femur midshaft and reduced bone strength of both femur and tibia10,11. Since we showed that the effect of rs3801387 and rs2707466 originated from the same causative SNP, rs55710688 could possibly be this causative SNP due to this additional functional evidence and the high LD value with rs3801387.

Altogether, these findings indicate that WNT16 has a positive effect on BMD and bone strength. When we tested the effect of Wnt16 on WNT signaling in vitro, the results showed no activation of the canonical pathway in the presence of LRP5 or LRP6, as previously already reported24. Moreover, a recent study demonstrated that WNT16 is able to inhibit WNT3A activation of the canonical WNT signaling pathway25. We now tested the effect of Wnt16 on Wnt1 activation of the β-catenin dependent WNT pathway in the presence of either LRP5 or LRP6 and showed that Wnt16 dominantly inhibits activation of canonical WNT signaling in a dose-dependent manner. An increased translation of WNT16 can thus lead to an increased inhibitory action of WNT16 on canonical WNT signaling. However, this is in contrast with the known activating effect of canonical WNT signaling on bone formation. Furthermore, homozygous deletion of Wnt16 in mice results in a decreased BMD, bone strength and cortical bone thickness.

In conclusion, our results strongly suggest that WNT16 is most likely able to increase BMD through the modulation of one of the non-canonical WNT signaling pathways, compensating for its inhibitory effect on canonical WNT signaling. More research, including in vivo studies, is required to not only confirm this hypothesis, but also to further elucidate the role of non- canonical WNT pathways in bone metabolism and the general mechanisms of interplay between the different WNT signaling pathways.

V. Acknowledgments

This work was supported by a grant from the University of Antwerp (TOP-BOF) to W. Van Hul. E. Boudin holds a post-doctoral fellowship of the FWO (Fund for Scientific Research) Vlaanderen. I. Fijałkowski holds a holds a pre-doctoral specialization scholarship from the “Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT- Vlaanderen)”.

Results – Chapter 3

References

1. Baron, R. & Hesse, E. Update on bone anabolics in osteoporosis treatment: rationale, current status, and perspectives. J Clin Endocrinol Metab 97, 311-25 (2012). 2. Gueguen, R. et al. Segregation analysis and variance components analysis of bone mineral density in healthy families. J Bone Miner Res 10, 2017-22 (1995). 3. Pocock, N.A. et al. Genetic determinants of bone mass in adults. A twin study. J Clin Invest 80, 706-10 (1987). 4. Balemans, W. et al. Identification of a 52 kb deletion downstream of the SOST gene in patients with van Buchem disease. J Med Genet 39, 91-7 (2002). 5. Leupin, O. et al. Bone overgrowth-associated mutations in the LRP4 gene impair sclerostin facilitator function. J Biol Chem 286, 19489-500 (2011). 6. Little, R.D. et al. A mutation in the LDL receptor-related protein 5 gene results in the autosomal dominant high-bone-mass trait. Am J Hum Genet 70, 11-9 (2002). 7. Estrada, K. et al. Genome-wide meta-analysis identifies 56 bone mineral density loci and reveals 14 loci associated with risk of fracture. Nat Genet 44, 491-501 (2012). 8. Baron, R. & Kneissel, M. WNT signaling in bone homeostasis and disease: from human mutations to treatments. Nat Med 19, 179-92 (2013). 9. Garcia-Ibarbia, C. et al. Missense polymorphisms of the WNT16 gene are associated with bone mass, hip geometry and fractures. Osteoporos Int (2013). 10. Medina-Gomez, C. et al. Meta-analysis of genome-wide scans for total body BMD in children and adults reveals allelic heterogeneity and age-specific effects at the WNT16 locus. PLoS Genet 8, e1002718 (2012). 11. Zheng, H.F. et al. WNT16 influences bone mineral density, cortical bone thickness, bone strength, and osteoporotic fracture risk. PLoS Genet 8, e1002745 (2012). 12. Boudin, E. et al. A common LRP4 haplotype is associated with bone mineral density and hip geometry in men-data from the Odense Androgen Study (OAS). Bone 53, 414-20 (2013). 13. Nielsen, T.L. et al. Prevalence of overweight, obesity and physical inactivity in 20- to 29- year-old, Danish men. Relation to sociodemography, physical dysfunction and low socioeconomic status: the Odense Androgen Study. Int J Obes (Lond) 30, 805-15 (2006). 14. Nielsen, T.L., Wraae, K., Brixen, K.T., Andersen, M. & Hagen, C. [The Odense Androgen Study]. Ugeskr Laeger 166, 1449-51 (2004).

Results – Chapter 3

15. Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81, 559-75 (2007). 16. Lettre, G., Lange, C. & Hirschhorn, J.N. Genetic model testing and statistical power in population-based association studies of quantitative traits. Genet Epidemiol 31, 358-62 (2007). 17. Sims, A.M. et al. Genetic analyses in a sample of individuals with high or low BMD shows association with multiple Wnt pathway genes. J Bone Miner Res 23, 499-506 (2008). 18. Balemans, W. et al. The binding between sclerostin and LRP5 is altered by DKK1 and by high-bone mass LRP5 mutations. Calcif Tissue Int 82, 445-53 (2008). 19. Kozak, M. At least six nucleotides preceding the AUG initiator codon enhance translation in mammalian cells. J Mol Biol 196, 947-50 (1987). 20. de Bakker, P.I. et al. Efficiency and power in genetic association studies. Nat Genet 37, 1217-23 (2005). 21. De Angioletti, M., Lacerra, G., Sabato, V. & Carestia, C. Beta+45 G --> C: a novel silent beta-thalassaemia mutation, the first in the Kozak sequence. Br J Haematol 124, 224-31 (2004). 22. Gonzalez-Conejero, R. et al. A common polymorphism in the annexin V Kozak sequence (-1C>T) increases translation efficiency and plasma levels of annexin V, and decreases the risk of myocardial infarction in young patients. Blood 100, 2081-6 (2002). 23. Wolf, A. et al. Single base-pair substitutions at the translation initiation sites of human genes as a cause of inherited disease. Hum Mutat 32, 1137-43 (2011). 24. Lu, D. et al. Activation of the Wnt signaling pathway in chronic lymphocytic leukemia. Proc Natl Acad Sci U S A 101, 3118-23 (2004). 25. Green, J.L. et al. Use of a molecular genetic platform technology to produce human Wnt proteins reveals distinct local and distal signaling abilities. PLoS One 8, e58395 (2013).

Results – Chapter 3

Chapter 4

WNT16 Requires Gα Subunits as Intracellular Partners for Both its Canonical and Non-Canonical WNT Signaling Activity in Osteoblasts

Gretl Hendrickx1, Eveline Boudin1 and Wim Van Hul1

1Department of Medical Genetics, University of Antwerp, Belgium

Work in progress

Results – Chapter 4

Abstract Ever since genetic variation in WNT16 was associated with BMD at numerous skeletal sites and with the risk for osteoporotic fractures, it became an interesting target for both in vitro and in vivo functional studies. Even though there are indications that these effects are coming from the activation of both canonical and non-canonical WNT signaling, a clear model of WNT signaling by WNT16 is not yet depicted. We therefore first investigated the modulation of canonical (WNT/β-catenin) and non-canonical (WNT/Calcium and WNT/Planar Cell Polarity) signaling in HEK293T and Saos-2 cells. This demonstrated that WNT16 activates all WNT signaling pathways in osteoblasts, whereas WNT/Calcium was only activated in HEK293T cells. Strikingly, activation of WNT/β-catenin signaling by Wnt16 in osteoblasts was independent of LRP5/6 co- receptors and resistant to the specific inhibitory effects of sclerostin and Dickkopf1. In osteoblasts, we are currently investigating the role of Gα subunits in the signaling activity by Wnt16 by performing knockdown of Gα12, Gα13 and Gαq. So far, successful knockdown of Gα12 and Gαq was achieved and luciferase reporter assays were performed. Knockdown of Gαq significantly decreased activity of WNT/Calcium signaling by Wnt16 in osteoblasts, whereas knockdown of Gα12 did not affect WNT signaling activity. Future studies will investigate whether knockdown of Gα13 or combined knockdown of Gα12 and Gα13 affects WNT signaling activity by Wnt16. Nevertheless, these studies already point out a possible orchestrating role of Gα subunits in the activity of Wnt16 in osteoblasts and could illustrate a mechanism of interplay between the different WNT signaling pathways.

Results – Chapter 4

I. Introduction The past five years, WNT16 was pointed out by numerous candidate gene or genome-wide association studies as an interesting determinant of bone mineral density (BMD) at different skeletal sites and the risk of fragility fractures1-9. The initial functional validation of the relevance of WNT16 in skeletal homeostasis came with the development of a Wnt16 (Wnt16-/-) knockout (KO) mouse model. These mice have a significantly reduced total body BMD, thinner cortical bones at the femur midshaft and reduced bone strength of both femur and tibia with a more severe phenotype seen in female than in male mice6,9,10. Off all skeletal cell types, WNT16 is predominantly expressed in osteoblasts lining the cortical bone surface. Here, WNT16 is known to affect osteoclastogenesis, not osteoclast function, in a direct and indirect manner. Indirect mechanisms include an osteoblast-mediated change in the RANKL/OPG ratio by WNT16, while WNT16 also inhibits osteoclastogenesis in a direct manner. So far, these effects of WNT16 are carefully attributed to activation of the WNT/β-catenin pathway and a non-canonical JNK cascade in osteoblasts, while it solely activates non-canonical WNT signaling in osteoclasts10. Altogether, these findings highlight WNT16 as a key regulator of cortical bone mass and fragility fractures, but the precise effects of WNT16 on the different WNT signaling pathways are poorly understood.

As mentioned in the introduction of this thesis, the WNT signaling pathway is subdivided in three distinctive WNT pathways; the canonical or β-catenin dependent (WNT/β-catenin) pathway and two non-canonical or β-catenin independent pathways (WNT/Calcium and WNT/Planar Cell Polarity (PCP) pathway)11. Here, WNTs give the initial impulse to modulate this multifaceted pathway after binding to a G-protein coupled Frizzled receptor. These WNTs were initially categorized as either canonical or non-canonical, but later it was demonstrated that the same WNT can activate different parts of the WNT signaling pathway depending on the cell- receptor context12. Therefore, depending on which (co-)receptors are available in which cell type, a WNT can modulate WNT signaling differently. Interestingly, WNT16 appears to be such a WNT, since it does not activate WNT/β-catenin signaling in HEK293T cells4 and was reported to activate this pathway in MC3T3-E1 pre-osteoblasts10,13.

Increasing the knowledge on how WNT16 modulates these distinct WNT signaling pathways could therefore undoubtedly contribute to explaining the cellular and bone tissue specific effects by Wnt16 known so far. The aim of this study was thus to investigate the modulation of canonical and non-canonical WNT signaling by Wnt16 and look for a possible mechanism explaining the observed effects. Therefore, we performed up to four different luciferase reporter

Results – Chapter 4 assays in both HEK293T and Saos-2 cells, an osteoblast-like cell line, to examine the modulation of canonical and non-canonical WNT signaling by WNT16 in both cell types. Based on these results, we decided to further look into the mechanism of action of Wnt16 on osteoblasts. For this purpose we were able to select the Gαq, Gα12 and Gα13 subunits for knockdown experiments in osteoblasts. Here, we knockdown of these specific Gα subunits in Saos-2 cells with overexpression of Wnt16 in NFAT, AP-1 and TCF/LEF sensitive luciferase reporter assays to explore whether Wnt16 requires specific Gα subunits as intracellular partners in osteoblasts.

II. Material and methods

1. Cell culture, transfection and luciferase reporter assays

The human embryonic kidney (HEK) cell line 293T and the Saos-2 human osteosarcoma cell line (ATCC) were grown in DMEM supplemented with FBS (10% v/v, Life Technologies, Carlsbad,

CA). Cells were maintained at 37°C in 5% CO2 and 95% air atmosphere. Both HEK293T and Saos-2 cells were passaged when confluent by standard trypsinization with TrypLE™ Express (Life Technologies, Carlsbad, CA) and reculturing procedure. Twenty-four hours prior to transfection, HEK293T or Saos-2 cells were plated at 0.3 x 105 cells/well in 96-well plates for standard luciferase reporter assay experiments. Fugene 6 and ViaFect Transfection Reagent (Promega Corporation, Madison, WI) were used for transfection of HEK293T and Saos-2 cells, respectively, according to the manufacturer’s instructions. In HEK293T cells, different amounts of the pCMV6-Entry vector containing murine Wnt16 cDNA (1-10-40ng) were cotransfected with pRL-TK (1ng) and a pGL4.30 (20ng, Promega Corporation, Madison, WI) or pGL4.44 (20ng, Promega Corporation, Madison, WI) luciferase reporter vector with NFAT or AP-1 response elements, respectively. In Saos-2 cells, Wnt1-V5 (40ng), mesdc-2 (8ng), LRP5 (8ng) or LRP6 (8ng), Sost-HA (40ng), DKK1 (40ng), pRL-TK (2ng) and pGL3-OT (30ng) were cotransfected together with different amounts of pCMV6-Entry-Wnt16 (40-80ng) to monitor WNT/β-catenin signaling. Cotransfection of different amounts of pCMV6-Entry-Wnt16 (1-10- 40ng) with pRL-TK (2ng) and pGL4.30 (30ng), pGL4.44 (30ng) or pCRE-Luc (30ng, Stratagene, La Jolla, CA) in Saos-2 cells was performed to monitor WNT/Calcium, WNT/PCP or cAMP signaling. When needed, empty pcDNA3.1 vector was added to make total DNA amount equal for all transfection experiments. Each transfection was carried out in triplicate and repeated independently in three separate experiments. Forty-eight hours after transfection, cells were lysed and firefly and renilla luciferase activity were measured on a Glomax Multi+ Luminometer (Turner Designs, Sunnyvale, CA) using the dual luciferase reporter assay system (Promega

Results – Chapter 4

Corporation, Madison, WI) according to manufacturer’s instructions. The ratio of the firefly and renilla luciferase measurement was calculated. Data are expressed as mean values ± SD.

2. GNA12, GNA13, GNAQ DsiRNAs and transfection For knockdown of GNA12 (NM_007353), GNA13 (NM_006572) and GNAQ (NM_002072), three TriFECTa RNAi Kits (Integrated DNA Technologies, Coralville, IA) were purchased, each including three Dicer-substrate 27mer duplexes (DsiRNAs) targeting one specific gene. Prior to transfection, DsiRNAs were heated to 94°C for 2 minutes to have fully resuspended DsiRNAs in a stable and double-stranded form. To obtain optimal transfection conditions, cotransfection of empty pcDNA3.1 vector and a TYE 563-labeled control siRNA duplex (Integrated DNA Technologies, Coralville, IA) in Saos-2 cells was performed and transfection efficiency was verified using fluorescence microscopy. A scrambled (NC1, Integrated DNA Technologies, Coralville, IA) and human hypoxanthine phosphoribosyltransferase (HPRT) siRNA duplex were used as negative and positive controls, respectively, to confirm transient transfection and knockdown efficiency. For efficient knockdown of the selected Gα subunits, 20nM of one TriFECTa siRNA duplex targeting GNA12 and 10nM of one TriFECTa siRNA duplex targeting GNAQ were transfected into Saos-2 cells using the ViaFect transfection reagent (Promega Corporation, Madison, WI) under RNase-free conditions following the manufacturer’s protocol. Sense and antisense sequences of the DsiRNAs used in our experiments can be found in Table 1.

Table 1 | Duplex siRNAs (DsiRNAs) used for knockdown experiments in Saos-2 cells. The IDT Catalog ID, the corresponding sense and antisense sequences and concentrations of the DsiRNAs targeting GNA12, GNA13, GNAQ to obtain up to 70% knockdown in Saos-2 cells. No concentration is given for the scrambled negative control DsiRNA (NC1), as this depends on the concentration of the test DsiRNA used in the same experiment (10nM or 20nM).

Gene Concentration DsiRNA sequence 5’ – 3’ IDT Catalog ID (nM) GNA12 (human) Sense GUAUCACUCUACUGAGAAAUCCUGC 20 HSC.RNAI.N007353.12.3 Antisense GCAGGAUUUCUCAGUAGAGUGAUACAC GNA13 (human) Sense CCCAUACAUUUAGCAAAUUGAAUGG 40 HSC.RNAI.N006572.12.1 Antisense CCAUUCAAUUUGCUAAAUGUAUGGGUA GNAQ (human) Sense AGCGCUUAGUGAAUAUGAUCAAGTT 10 HSC.RNAI.N002072.12.3 Antisense AACUUGAUCAUAUUCACUAAGCGCUAC Scrambled Sense CGUUAAUCGCGUAUAAUACGCGUAT / NC1 Negative Control Antisense AUACGCGUAUUAUACGCGAUUAACGAC

Results – Chapter 4

3. Quantitative Real-Time PCR (qRT-PCR) Saos-2 cells were plated at 0.7 x 105 cells/well in 24 well plates. Twelve and twenty-four hours after transfection with DNA and DsiRNAs, cells were harvested and total RNA was prepared by the ReliaPrep RNA Cell Miniprep System (Promega Corporation, Madison, WI). For cDNA synthesis, an average of 160ng total RNA was reverse transcribed to cDNA in 21 µl reactions using the SuperScript III First-Strand Synthesis System (Life Technologies, Carlsbad, CA) according to manufacturer’s instructions. qPCR was performed with the LightCycler 480 system (Roche Applied Science, Mannheim, Germany) using the qPCR Core kits for SYBR Green I, No ROX (Eurogentec, Seraing, Belgium). Briefly, 2 µl of cDNA was amplified by PCR in 15 µl reactions containing 0,45 µl of SYBR green and 0.2 µM of each of the primers (Integrated DNA Technologies, Coralville, IA). The initial HotGoldStar activation for 10 min at 95°C was followed by 40 cycles at 95°C for 15 sec, 60°C for 20 sec and 72°c for 40 sec, a melting curve and cooling phase. Each sample was analyzed in triplicate and B2M, HMBS and SDHA were included as reference genes. Stability of reference genes was verified using geNorm (Biogazelle, Ghent, Belgium) and efficiency of all primer pairs was checked with the qbase+ software (Biogazelle, Ghent, Belgium). Expression of target and reference genes was quantified using qbase+ software. The sequences of all primers used for qRT-PCR are listed in Table 2.

Table 2. Primers used for qRT-PCR. Sequences of forward and reverse qRT-PCR primers used for the quantification of GNA12, GNA13 and GNAQ expression in Saos-2 cells. Gene Primer Sequence 5’ – 3’ Transcript ID GNA12 (human) Forward CTGGTGAAGATCCTGCTGCT NM_007353 Reverse AAGAACCCTTGAGCCCTTGA GNA13 (human) Forward AATCCAAAGCCAGAGGGTCCTG NM_006572 Reverse CCATGGCCTGTACTGCCTGTTATG GNAQ (human) Forward TGAGCACAATAAGGCTCATGCAC NM_002072 Reverse AGCAGACACCTTCTCCACATCAAC HPRT1 (human) Forward TGACACTGGCAAAACAATGCA NM_000194 Reverse GGTCCTTTTCACCAGCAAGCT HMBS (human) Forward GGCAATGCGGCTGCAA NM_000190 Reverse GGGTACCCACGCGAATCAC SDHA (human) Forward TGGGAACAAGAGGGCATCTG NM_004168 Reverse CCACCACTGCATCAAATTCATG B2M (human) Forward TGCTGTCTCCATGTTTGATGTATCT NM_004048 Reverse TCTCTGCTCCCCACCTCTAAGT

Results – Chapter 4

4. Enzyme-Linked Immunosorbent Assay (ELISA) Saos-2 cells were plated at 0.5 x 105 cells/well in 24 well plates. Twenty-four hours after transfection with DNA and DsiRNAs, cells were washed with PBS and trypsinized with TrypLE Express. Cell suspensions underwent three centrifugation steps (5 minutes, 600 x g, RT) after which the cell pellet was washed with cold PBS every time. Hereafter, cell lysis was attained by performing two freeze-thaw cycles. Next, lysed cells underwent centrifugation (15 minutes, 1000 x g, 4°C) and supernatant was kept for further analysis. ELISA kits from MyBioSource (San Diego, CA) were ordered to measure protein levels of human GNA12 (#MBS9317428), GNA13 (#MBS9340640) and GNAQ (#MBS7234399). All ELISAs were performed according to the manufacturer’s instructions. Each sample was analyzed in duplicate and the NC1 DsiRNA with empty vector DNA was used as a negative control. At last, to measure and correct for the total protein concentration of all samples, the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA) was used.

5. Luciferase reporter assays with Gα12, Gα13 and Gαq knockdown Saos-2 cells were plated at 0.1 x 105 cells/well in 96-well plates twenty-four hours prior to transfection. DsiRNAs were cotransfected with pCMV6-Entry-Wnt16 (40ng), pRL-TK (2ng) and pGL3-OT (30ng), pGL4.30 (30ng) or pGL4.44 (30ng). As negative controls for each luciferase reporter assay with DsiRNAs, a NC1 scrambled duplex was cotransfected with empty pcDNA3.1 vector or Wnt16. Twenty-four and forty-eight hours after transfection, cells were lysed and firefly and renilla luciferase activity were measured on a Glomax Multi+ Luminometer (Turner Designs, Sunnyvale, CA) using the dual luciferase reporter assay system (Promega Corporation, Madison, WI) according to manufacturer’s instructions under RNase-free conditions. The ratio of the firefly and renilla luciferase measurement was calculated. Data of twenty-four hours post- transfection are shown and expressed as mean values ± SD.

6. Statistical analysis All experiments were performed independently at least three times of which the result of only one is shown in the figures and represents the result of its triplicates. Comparison between two sets of measurements was performed on the relative mean of the values coming from the three repetitions using a Student’s t-test. Here, p<0.05 (*) or p< 0.01 (**) were considered as the different levels of significance. Statistical tests were provided by the SPSS 22.0 software package (SPSS Inc).

Results – Chapter 4

III. Results

1. Wnt16 modulates WNT signaling differently in HEK293T and Saos-2 cells According to the literature, Wnt16 does not activate WNT/β-catenin signaling in HEK293T cells4, while activation of the pathway in MC3T3-E1 preosteoblasts has been reported10,13. To further investigate this activation in osteoblasts, co-transfection of Wnt16 or Wnt1 with LRP5/6 and sclerostin or DKK1 was performed in Saos-2 cells. In these osteoblast-like cells, we were able to confirm dose-dependent activation of WNT/β-catenin signaling by Wnt16 (Figure 1).

Figure 1 | Wnt16 activates all WNT signaling pathways in Saos-2 cells. (A) As a positive control, transfection of Wnt1 alone demonstrates an increase in WNT/β-catenin signaling (TCF/LEF) that increases after co-transfection with LRP5/6. Adding the canonical inhibitors sclerostin (SOST) or dickkopf1 (DKK1), inhibits this activation by Wnt1, as expected. (B-C) For Wnt16 we see a dose- dependent (40-80ng) activation of WNT/β-catenin signaling. Remarkably, after co-transfection with LRP5/6, signaling activity remains the same or decreases. Adding sclerostin or dickkopf1 also does not inhibit, even increases, signaling activity. (D) Wnt16 activates WNT/Ca2+ signaling (NFAT) and WNT/PCP signaling (AP-1) in a dose-dependent manner in Saos-2 cells. *: p<0.05; **: p<0.01.

Results – Chapter 4

This activation by Wnt16 was stronger compared to co-transfecting the same amount of Wnt1 (Figure 1A-B). Moreover, activation of WNT/β-catenin signaling by Wnt16 does not increase when co-transfecting LRP5, unlike for Wnt1. On the contrary, a significant (p<0.05) decrease in activity was noticed for 80ng of Wnt16 and LRP5 (Figure 1C). Co-transfection of LRP6 with 40ng of Wnt16 did not increase WNT/β-catenin signaling, unlike for Wnt1, but a slight (not significant) increase in activity was observed with 80ng of Wnt16. Moreover, co-transfection of sclerostin or DKK1, two known inhibitors of WNT/β-catenin signaling, with Wnt16 and LRP5 or LRP6 did not decrease the signaling activity by Wnt16 (Figure 1A-C). In contrast, for LRP6, a significant (p<0.05) increase in WNT/β-catenin signaling was observed when sclerostin or Dickkopf1 were added. For sclerostin and a higher amount of Wnt16, the activity of the pathway doubled compared to that of Wnt16 alone. Altogether, our results confirm activation of WNT/β- catenin signaling by Wnt16, which occurs in a LRP5/6 independent manner and with a stimulatory effect of sclerostin and Dickkopf1, whereas this is not observed in presence of Wnt1 as ligand.

In order to explore and compare the effects of Wnt16 on non-canonical WNT signaling, a set of WNT luciferase reporter assays was performed in both HEK293T and Saos-2 cells (Figure 1D- 2). To quantify non-canonical WNT signaling, different amounts of Wnt16 were co-transfected with a luciferase vector containing NFAT- or AP-1-responsive elements to monitor WNT/Ca2+ or WNT/PCP pathway activity. These experiments demonstrate that Wnt16 activates the NFAT- responsive element in a dose-dependent manner, but inhibits the AP-1 responsive element in a

Figure 2 | Wnt16 modulates WNT/Ca2+ and WNT/PCP signaling differently in HEK293T. Wnt16 activates WNT/Ca2+ signaling (NFAT) in a dose-dependent manner in HEK293T cells, whereas WNT/PCP signaling (AP-1) was inhibited by Wnt16 in a dose-dependent manner. 99

Results – Chapter 4 dose-dependent manner in HEK293T cells (Figure 2). In Saos-2 cells, both the NFAT and AP-1 responsive reporter assays show dose-dependent activation by Wnt16 (Figure 1D). Altogether, Wnt16 activates WNT/Ca2+ signaling in HEK293T cells, but inhibits WNT/PCP signaling in a dose-dependent manner, while Wnt16 dose-dependently activates both non-canonical pathways in Saos-2 cells.

2. Selection and knockdown of Gα12, Gα13 and Gαq in Saos-2 cells Since Wnt16 modulates the NFAT, AP-1 and TCF/LEF transcription factors in osteoblasts and activation of TCF/LEF occurs in an atypical manner, the recruitment of specific Gα subunits towards the Frizzled receptor could explain these observations. Based on the literature, Gα subunits from the Gαq and Gα12/13 subfamily have already been linked to modulation of these specific transcription factors, making these subunits interesting for further investigation. Members of the Gαi and Gαs subfamily are known to respectively decrease or increase cAMP levels. To possibly include or exclude these subunits in subsequent functional studies, the cAMP level was monitored after overexpression of Wnt16 in Saos-2 cells with a cAMP-sensitive luciferase reporter assay using the pCRE-Luc vector. Here, after co-transfecting different amounts of Wnt16, no increase or decrease of pCRE-Luc activity was detected, implying stable levels of cAMP (data not shown). Since the NFAT, AP-1 and LEF/TCF transcription factors are activated by Wnt16, Gα subunits of the Gα12/13 and Gαq family were selected for knockdown experiments in Saos-2 cells.

Subsequently, GNA12 (NM_007353), GNA13 (NM_006572) and GNAQ (NM_002072) were targeted for knockdown, to verify the role of Gα12, Gα13 and Gαq, respectively, in the modulation of WNT signaling by Wnt16 in Saos-2 cells. Knockdown was verified on RNA and protein level by performing qPCR and ELISA, respectively. Here, sufficient knockdown of Gα12 and Gαq was achieved when transfecting 20nM DsiGNA12 and 10nM DsiGNAQ. On mRNA level, knockdown was between 50 and 70% for GNA12 (66.5%±16.0) and GNAQ (68.4%±15.6). ELISA confirmed knockdown on protein level for GNA12 (75.5%±24.3) and GNAQ (46.9±23.2).

3. Gα subunits are intracellular partners of Wnt16 in osteoblasts When performing the NFAT, AP-1 and LEF/TCF responsive luciferase reporter assays in Saos- 2 cells overexpressing Wnt16 with knockdown of Gαq alone, we detected a significant (p<0.05) decrease in NFAT reporter activity compared to co-transfection of Wnt16 with the NC1 negative control DsiRNA (Figure 3). A remarkable decrease in AP-1 reporter activity was also detected,

100

Results – Chapter 4 but did not reach significance levels. Overexpression of Wnt16 and knockdown of Gα12 alone did not alter the NFAT, AP-1 and LEF/TCF luciferase reporter activities (data not shown).

Figure 3 | Knockdown of Gαq (GNAQ) subunit affects non-canonical WNT/Ca2+ signaling by Wnt16. Signaling activity of WNT/β-catenin signaling (TCF/LEF), WNT/Ca2+ signaling (NFAT) and WNT/PCP signaling (AP-1) by Wnt16 was measured before and after knockdown of GNAQ (siGNAQ). A significant decrease in WNT/Ca2+ signaling activity by Wnt16 was observed after knockdown of GNAQ. *: p<0.05.

101

Results – Chapter 4

IV. Discussion

Ever since genetic variation in WNT16 was associated with BMD at numerous skeletal sites and with the risk for osteoporotic fractures by several independent candidate gene and genome-wide association studies, it became an interesting target for both in vitro and in vivo functional studies. The importance of WNT16 in the regulation of BMD and bone strength was subsequently confirmed by Wnt16 knockout (Wnt16-/-) mice which have a reduced whole body BMD, cortical bone thickness and bone strength6,9,10. Additionally, on a cellular level, WNT16 is mainly produced by the osteoblasts neighboring cortical bone tissue where it inhibits osteoclastogenesis in a direct and indirect manner, with consequent beneficial effects for the surrounding bone tissue10. Even though there are indications that the effects seen in vitro and in vivo are coming from the activation of both canonical and non-canonical WNT signaling in osteoblasts and only non- canonical signaling in osteoclasts, a clear model of WNT signaling by WNT16 is not yet depicted.

Therefore, we first investigated the effect of Wnt16 on the different WNT signaling pathways in two cell types, HEK293T and Saos-2 cells. Here, HEK293T cells were taken as a standard model, while Saos-2 cells are osteosarcoma cells that are known to be a valid osteoblast-like cell model. Previous findings already indicated that Wnt16 activates WNT/β-catenin signaling in MC3T3-E1 cells10,13, while no activation could be detected in HEK293T cells4. In this study, the previously reported activation of WNT/β-catenin signaling by Wnt16 in MC3T3-E1 cells is now also confirmed in Saos-2 cells, in a dose-dependent manner. Furthermore, we additionally demonstrated that this mechanism of activation is independent of LRP5/6 co-receptors and is resistant to the specific inhibitory effects of sclerostin and Dickkopf1 on canonical WNT signaling. Even more, we observed an increase in WNT/β-catenin signaling for sclerostin and Dickkopf1 in presence of LRP6 and overexpression of LRP5 slightly decreased the activity of WNT/β-catenin signaling. On the contrary, Wnt1, as a classical canonical WNT ligand did show the expected modulations in WNT/β-catenin signaling after co-transfection with LRP5/6, sclerostin and Dickkopf1. Based on these results, we demonstrate that activation of canonical WNT signaling by Wnt16 must occur through a non-classical mechanism. This hypothesis however is different from the one Movérare Skrtic et al. recently formulated10. Based on an observed increase in phosphorylated (thus activated) LRP6, Axin2 expression and non- phosphorylated (thus free and active) β-catenin in osteoblasts after WNT16 treatment, their conclusion was that WNT16 has the capacity to signal through the classical canonical pathway in osteoblasts. They also observed that this activation was albeit less efficient than seen with classical WNT ligands such as WNT3A. Still, in the classical model of WNT/β-catenin signaling,

102

Results – Chapter 4 sclerostin and DKK1 are inhibitors by interfering directly with LRP5/611. In our study, in presence of LRP6, we could not detect any decrease, but even an increase in WNT/β-catenin activity after co-transfection of Wnt16 with sclerostin or Dickkopf in osteoblasts. That is why there must be a deviant mechanism of activation fully upstream of the pathway with possible intracellular consequences for the phosphorylation of LRP6, the release of β-catenin from its destruction complex and the binding with TCF/LEF transcription factors after its translocation to the nucleus.

Next to studying the WNT/β-catenin pathway, we wanted to explore the effects of Wnt16 on the non-canonical WNT/Calcium and WNT/PCP pathway in both cell models based on the downstream activity of the NFAT and AP-1 transcription factors, respectively. In HEK293T cells we detected an activation of the WNT/Calcium pathway, while the WNT/PCP pathway was inhibited by Wnt16. In Saos-2 cells, however, Wnt16 was able to dose-dependently activate both transcription factors and thus non-canonical WNT pathways. Recently, Movérare-Skrtic et al. also reported non-canonical activity of WNT16 in MC3T3-E1 cells, which was solely based on increased phosphorylated (thus activated) JNK and c-JUN10. Since JNK is known to activate the AP-1 transcription factor and AP-1 is composed of proteins belonging to c-JUN, c-FOS, ATF and JDP families, we were able to confirm this finding in Saos-2 cells. Altogether, studying the modulation of WNT signaling by WNT16 in two different cell types clearly demonstrates that WNT16 is a textbook example of a WNT that is able to differentially modulate WNT signaling in different cell types. While only activating the WNT/Calcium pathway in HEK293T cells, WNT16 activates WNT/β-catenin, WNT/Calcium and WNT/PCP signaling in osteoblasts. This suggests that the availability of certain receptors or intra/extracellular modulators varies markedly between cell types, which stresses the importance of investigating WNTs in suitable cell models to draw relevant conclusions.

Due to activity of the AP-1, NFAT and the TCF/LEF transcription factors by Wnt16 in osteoblasts, of which activation of WNT/β-catenin signaling comes from a deviant mechanism of activation, we examined a possible upstream role for Gα subunits in the attainment of these downstream effects. It was previously shown that Frizzled receptors are G-protein coupled receptors and are thus able to recruit G-proteins composed of α, β and γ subunits14. Based on previous studies and the transcription factors of which we so far know are activated by Wnt16 in osteoblasts, we selected the Gα subunits (Gαi, Gαs, Gαq, Gα12/13) as most interesting for our study. However, since we did not detect an increase (Gαs) or decrease (Gαi) of cAMP levels after overexpression of Wnt16 in Saos-2 cells, we decided to select Gαq and Gα12/13 subunits for

103

Results – Chapter 4 knockdown experiments. Here, after knockdown of the Gαq subunit alone, we noticed a slight decrease in activity of all WNT signaling pathways, compared to co-transfection of Wnt16 with a negative control siRNA duplex. This decrease in activity was only significant (p<0.05) for the WNT/Calcium pathway. The correlation between active Gαq and an increase in calcium levels with subsequent NFAT activation has been well-documented and could explain the significant decrease in WNT/Calcium activation by Wnt16 when Gαq alone was knocked down15,16. Knockdown of Gα12 alone did not alter any signaling activity by Wnt16. As we still have to optimize knockdown of Gα13 and believe that Gα12 and Gα13 could work together as partners, combined knockdown of these Gα subunits in the future could result in significant results. Gα12 and Gα13 are indeed well-known to activate RhoA through a direct interaction with Rho guanine nucleotide exchange factors (RhoGEFs), connecting these subunits directly to the activity of Rho-dependent kinase (ROCK), JNK and the AP-1 transcription factor15-17. Moreover, Gα12 and Gα13 can interact with the cytoplasmic domain of cadherins and thereby provide a nonstandard mechanism for β-catenin release from cadherins and subsequent WNT/β-catenin signaling activation18-20. Therefore, the involvement of cadherin could suit our findings on the non-classical way of activating canonical signaling by Wnt16.

In conclusion, we have shown that WNT16 modulates canonical and non-canonical WNT signaling differently in disparate cell types, emphasizing the relevance of investigating WNTs in an appropriate cell model. In osteoblasts, WNT16 activates WNT/β-catenin, WNT/Calcium and WNT/PCP signaling of which activation of WNT/β-catenin signaling occurs in an untypical way. Knockdown studies of Gα12 and Gαq already point out a possible orchestrating role of Gα subunits in the activity of Wnt16 in osteoblasts and could illustrate a mechanism of interplay between the different WNT signaling pathways. Still, more research is required to further verify the consequences for osteoblast function and its communication with the osteoclasts, after activation of the NFAT, AP-1 and TCF/LEF transcription factors by Wnt16 in osteoblasts.

104

Results – Chapter 4

V. Acknowledgements This work was supported by the Systems Biology for the Functional Validation of Genetic Determinants of Skeletal Diseases (SYBIL) project funded by the European Commission. G.H. holds a Rosa Blanckaert Legacy Grant for Young Investigators from the University of Antwerp (Project ID 32435). E.B. holds a post-doctoral fellowship of the FWO Flanders (FWO personal grant 12A3814N).

105

Results – Chapter 4

References

1. Chesi, A. et al. A trans-ethnic genome-wide association study identifies gender-specific loci influencing pediatric aBMD and BMC at the distal radius. Hum Mol Genet 24, 5053-9 (2015). 2. Estrada, K. et al. Genome-wide meta-analysis identifies 56 bone mineral density loci and reveals 14 loci associated with risk of fracture. Nat Genet 44, 491-501 (2012). 3. Garcia-Ibarbia, C. et al. Missense polymorphisms of the WNT16 gene are associated with bone mass, hip geometry and fractures. Osteoporos Int 24, 2449-54 (2013). 4. Hendrickx, G. et al. Variation in the Kozak sequence of WNT16 results in an increased translation and is associated with osteoporosis related parameters. Bone 59, 57-65 (2014). 5. Koller, D.L. et al. Meta-analysis of genome-wide studies identifies WNT16 and ESR1 SNPs associated with bone mineral density in premenopausal women. J Bone Miner Res 28, 547-58 (2013). 6. Medina-Gomez, C. et al. Meta-analysis of genome-wide scans for total body BMD in children and adults reveals allelic heterogeneity and age-specific effects at the WNT16 locus. PLoS Genet 8, e1002718 (2012). 7. Moayyeri, A. et al. Genetic determinants of heel bone properties: genome-wide association meta-analysis and replication in the GEFOS/GENOMOS consortium. Hum Mol Genet 23, 3054-68 (2014). 8. Zhang, L. et al. Multistage genome-wide association meta-analyses identified two new loci for bone mineral density. Hum Mol Genet 23, 1923-33 (2014). 9. Zheng, H.F. et al. WNT16 influences bone mineral density, cortical bone thickness, bone strength, and osteoporotic fracture risk. PLoS Genet 8, e1002745 (2012). 10. Moverare-Skrtic, S. et al. Osteoblast-derived WNT16 represses osteoclastogenesis and prevents cortical bone fragility fractures. Nat Med 20, 1279-88 (2014). 11. Boudin, E., Fijalkowski, I., Piters, E. & Van Hul, W. The role of extracellular modulators of canonical Wnt signaling in bone metabolism and diseases. Semin Arthritis Rheum 43, 220-40 (2013). 12. Mikels, A.J. & Nusse, R. Purified Wnt5a protein activates or inhibits beta-catenin-TCF signaling depending on receptor context. PLoS Biol 4, e115 (2006). 13. Jiang, Z., Von den Hoff, J.W., Torensma, R., Meng, L. & Bian, Z. Wnt16 is involved in intramembranous ossification and suppresses osteoblast differentiation through the Wnt/beta-catenin pathway. J Cell Physiol 229, 384-92 (2014).

106

Results – Chapter 4

14. Schulte, G. & Bryja, V. The Frizzled family of unconventional G-protein-coupled receptors. Trends Pharmacol Sci 28, 518-25 (2007). 15. Dorsam, R.T. & Gutkind, J.S. G-protein-coupled receptors and cancer. Nat Rev Cancer 7, 79-94 (2007). 16. Worzfeld, T., Wettschureck, N. & Offermanns, S. G(12)/G(13)-mediated signalling in mammalian physiology and disease. Trends Pharmacol Sci 29, 582-9 (2008). 17. Riobo, N.A. & Manning, D.R. Receptors coupled to heterotrimeric G proteins of the G12 family. Trends Pharmacol Sci 26, 146-54 (2005). 18. Fujino, H. & Regan, J.W. FP prostanoid receptor activation of a T-cell factor/beta - catenin signaling pathway. J Biol Chem 276, 12489-92 (2001). 19. Meigs, T.E., Fedor-Chaiken, M., Kaplan, D.D., Brackenbury, R. & Casey, P.J. Galpha12 and Galpha13 negatively regulate the adhesive functions of cadherin. J Biol Chem 277, 24594-600 (2002). 20. Meigs, T.E., Fields, T.A., McKee, D.D. & Casey, P.J. Interaction of Galpha 12 and Galpha 13 with the cytoplasmic domain of cadherin provides a mechanism for beta - catenin release. Proc Natl Acad Sci U S A 98, 519-24 (2001).

107

Results – Chapter 4

108

Chapter 5

Genetic Screening of WNT4 and WNT5B in Two Populations with Deviating Bone Mineral Densities

Gretl Hendrickx1, Eveline Boudin1, Ellen Steenackers1, Torben Leo Nielsen2, Marianne Andersen2, Kim Brixen2 and Wim Van Hul1

1Department of Medical Genetics, University of Antwerp, Belgium

2Department of Endocrinology, Odense University Hospital, Denmark

This chapter is based on the research article published in: Calcified Tissue International 2017; 100:244–249

109

Results – Chapter 5

Abstract

A role for WNT4 and WNT5B in bone metabolism was indicated by genome-wide association studies (GWAS) and a Wnt4 knockout mouse model. The aim of this study was therefore to replicate and further investigate the causality between genetic variation in WNT4 and WNT5B and deviating bone mineral density (BMD) values.

A WNT4 and WNT5B mutation screening was performed in patients with craniotubular hyperostosis using Sanger sequencing. Here, no putative causal mutations were detected. Moreover, a high and low BMD cohort were selected from the Odense Androgen Study population for re-sequencing. In WNT4 we detected four variants (three rare, one common), while in WNT5B we detected five variants (two rare, three common). For the common variants, no significant difference in genotype frequencies between the high and low BMD cohort was observed. The SNPs associated in the GWAS were genotyped in these cohorts, but again no significant difference in genotype frequencies was observed.

Despite the findings of the GWAS, we were not able to replicate or further verify the genetic association of polymorphisms in WNT4 and WNT5B with BMD. In order to do so, the intronic regions of both genes could be investigated more thoroughly in more extended populations (or extremes) with greater power. Future genetic and functional studies towards adjacent genes of WNT4 and WNT5B can also be interesting to figure out whether the signal from GWAS could possibly be attributed to genetic variation in these genes.

110

Results – Chapter 5

I. Introduction As a complex trait, bone mineral density (BMD) is subjected to environmental influences, aging- related changes and genetic variation. BMD values are interpreted by the use of t-scores, the number of standard deviations above or below the peak bone mass of young adults. A t-score between -1 and +1 is considered normal or healthy, whereas a t-score below -2.5 or higher than +2.5 indicates severe bone fragility or an abnormally high BMD, respectively1,2. Osteoporosis is a common bone disorder characterized by decreased BMD values (t-score≤-2.5) and micro- architectural changes resulting in an increased facture risk3-5. With an estimated heritability (percentage of variation in a phenotypic trait attributable to genetic variation) ranging from 50 to 85%, genetic background is referred to as an essential determinant of (peak) BMD and therefore indirectly, osteoporosis. It turned out that the general interindividual differences in BMD can be explained by genetic variations in many genes with small effects. This is why, the past decade, genome-wide association studies (GWAS) have been an essential tool in exploring the complex genetic character of BMD6-12.

In one of the largest meta-analyses of GWAS to date, 56 loci were associated with BMD of which 14 were also associated with fracture risk. Several of the genes located in or near these loci clustered within the WNT signaling pathway, which is well-known for its role in skeletal homeostasis and the development of bone disorders7. WNT4 and WNT5B, two WNT genes and members of this pathway, were associated with lumbar spine and femoral neck BMD in this study, whereas WNT4 was also associated with fracture risk7. In previous GWAS, WNT4 had already been associated with total hip and lumbar spine BMD13,14, while WNT5B was one of the newly associated genes in the study by Estrada et al.7. A recent WNT4 and WNT5B candidate gene association study in a population of young Chinese men also identified associations of intronic variants with peak BMD15. Furthermore, transgenic mice overexpressing Wnt4 from osteoblasts demonstrated a protective effect of Wnt4 on bone loss and chronic inflammation by inhibiting both osteoclastogenesis and bone resorption 16. Transgenic mice overexpressing Wnt5b in cartilage had delayed chondrocyte hypertrophy and bone mineralization, whereas Wnt5b knockout (Wnt5b-/-) mice often had slightly elevated bone mass, but this phenotype was not observed consistently17,18.

These findings highlight WNT4 and WNT5B as interesting candidate genes in the context of bone research. We therefore performed a WNT4 and WNT5B mutation screening in patients with craniotubular hyperostosis. Moreover, a high and low BMD cohort selected from the young Odense Androgen Study population were selected for re-sequencing19,20. The SNPs associated 111

Results – Chapter 5 with BMD in the GWAS were genotyped in these BMD cohorts as well. Altogether, with this thorough genetic analysis we aim to gain further insights on how common and rare variation in WNT4 and WNT5B contributes to deviating BMD values.

II. Material and methods 1. Subjects

Our patient population comprises 53 individuals diagnosed with a form of craniotubular hyperostosis (Van Buchem disease, sclerosteosis, endosteal hyperostosis…) and tested negative for mutations in LRP4, LRP5 and SOST, known causative genes for these bone disorders. Details on inclusion of subjects and their characteristics were previously described by Boudin et al. 21.

The Odense Androgen Study (OAS) is a population-based, prospective, observational study on the interrelationship between endocrine status, body composition, muscle function and bone metabolism in men19,20. Here, out of the young OAS population of 783 non-stratified Danish men aged 20-29 years, two cohorts were selected with the most extreme whole body BMD values (t-score<-1.38, n=62 and t-score>1.54, n=62). General characteristics of both BMD cohorts are given in Table 1. Details on the design and inclusion of the participants in the young OAS cohort and the selection of a low and high BMD cohort were previously reported elsewhere19,20,22,23.

Table 1 | General characteristics of the low (t-score<- 1.38) and high (t-score>1.54) BMD cohorts.

t-score <-1.38 t-score >1.54 Age 25.64±2.73 25.36±2.50 (years) Height 179.9±8.1 183.5±6.4 (cm) BMI 22.7±2.9 26.1±3.1 (kg/m²) Whole Body BMD -1.69±0.27 1.92±0.36 (t-score)

2. Sanger sequencing

Selection of the amplicons and primer design (Primer324,25) for WNT4 and WNT5B were based on the template sequence available in NCBI (www.ncbi.nlm.nih.gov/gene). Primers are designed for eight amplicons which comprise all coding exons and the corresponding intron/exon boundaries of all different protein coding transcripts of WNT4 and WNT5B (Supplementary

112

Results – Chapter 5

Material, Table 1). Moreover, two different primer sets were designed for the ±200bp region spanning rs7521902 and rs2887571, the BMD-associated SNPs within the locus of WNT4 and WNT5B respectively (Supplementary Material, Table 2).

Amplification of the selected amplicons was performed using GoTaq DNA polymerase-mediated PCR (Promega Corporation, Madison, WI) and verified by agarose gel electrophoresis simultaneously running a Generuler 100bp Plus DNA ladder (Fermentas International Inc, Ontario, Canada). After amplification by PCR, primers and unincorporated dNTPs were removed using exonuclease I (New England Biolabs, Inc, Ipswich, Massachusetts, USA) and calf intestine alkaline phosphatase (CIAP, Roche Applied Science, Hoffmann–La Roche AG, Basel, Switzerland). Sequencing was carried out directly on purified fragments with the ABI 310 Genetic Analyser (Applied Biosystems, Foster City, California, USA), using an ABI Prism BigDye terminator cycle sequencing ready reaction kit, version 1.1 (Applied Biosystems, Foster City, California, USA). The primers used here were identical to the primers that were used for amplification, except for exon 5 of both WNT4 and WNT5B that required two additional internal primers just for sequencing (Supplementary Material, Table 3). The BigDye XTerminator purification kit was used as purification method for DNA sequencing with the purpose of removing unincorporated BigDye terminators.

Finally, a Chi² test was used to compare the genotype frequencies of all detected polymorphisms between the high and the low BMD cohort (SPSS version 22.0 for (SPSS, Chicago, IL, USA)). P- values less than 0.05 were considered as significant.

III. Results 1. WNT4

Mutation screening of all coding exons and intron/exon boundaries of WNT4 in our patient population with Sanger sequencing led to the identification of several known polymorphisms. These include rs34228276 (MAF=0.018) which is a rare, synonymous SNP (found in four patients) located in exon 5 and rs41420150 is a common (MAF≥0.05) intronic insertion/deletion polymorphism with no effect on the splicing of WNT4 according to the prediction program MutationTaster (Table 2)26. Moreover, in a 7-year old American boy referred to us with a diagnosis of Van Buchem disease, we detected an unknown 20bp-insertion (c.313+9insGGGCGTGCGGGGCTGCAGGG) downstream of exon 2 of WNT4 which was not present in Exome Aggregation Consortium (ExAC) database, Cambridge, MA (http://exac.broadinstitute.org) [September 2016]. Van Buchem disease (OMIM 239100) is an 113

Results – Chapter 5

Table 2 | Genotype frequencies of known and unknown variants in the exons and intron/exon boundaries of WNT4 and WNT5B in individuals of the young OAS cohort with the most extreme BMD values (t-score<−1.38 and t-score>1.54).

Variation Genotype Genotype frequency Position t-score t-score Patients <-1.38 >1.54 c.313+9insGGGCGTGCGGGGCT -/- 1.00 1.00 0.98 GCAGGG GGGCGTGCGGGGCTGCAG Intron 2 GG/- 0.00 0.00 0.02 GGGCGTGCGGGGCTGCAG GG/ 0.00 0.00 0.00 GGGCGTGCGGGGCTGCAG GG rs115547783 CC 0.94 1.00 1.00 Intron 2 CG 0.06 0.00 0.00 WNT4 GG 0.00 0.00 0.00 rs41420150 CA/CA 0.67 0.50 0.51 Intron 4 CA/- 0.29 0.42 0.36 -/- 0.05 0.08 0.13 rs34228276 CC 1.00 0.97 0.94 Exon 5 CT 0.00 0.03 0.06 TT 0.00 0.00 0.00 c.717T>C TT 1.00 0.98 1.00 Exon5 TC 0.00 0.02 0.00 CC 0.00 0.00 0.00 rs35177332 GG 0.89 0.92 0.89 Intron 2 GT 0.11 0.08 0.11 TT 0.00 0.00 0.00 c.81-72_76delTGAA TGAA/TGAA 1.00 0.98 1.00 Intron 2 TGAA/- 0.00 0.02 0.00 -/- 0.00 0.00 0.00 rs58317077 CC 0.50 0.45 0.38 Intron 3 CT 0.42 0.50 0.58 TT 0.08 0.05 0.04 WNT5B rs6489313 AA 0.67 0.67 0.65 Intron 3 AG 0.33 0.33 0.31 GG 0.00 0.00 0.04 rs2270036 CC 0.71 0.75 0.68 Intron 3 CT 0.29 0.25 0.30 TT 0.00 0.00 0.02 rs150051301 CC 1.00 1.00 0.98 Exon 5 CT 0.00 0.00 0.02 TT 0.00 0.00 0.00

autosomal recessive bone disorder characterized by hyperostosis of the skull, mandible, clavicles, ribs and diaphyseal cortices of the long bones. So far, the only causative genetic defect identified for Van Buchem disease is a 52-kb deletion approximately 35 kb downstream of the SOST gene, coding for sclerostin, that removes a SOST-specific regulatory element27. All the patients screened in this study were negative for the deletion in this SOST-specific regulatory element. Sclerostin is an inhibitor of the WNT signalling pathway, the pathway of which WNT4 is a modulator.

114

Results – Chapter 5

Nevertheless, MutationTaster predicts this variant to be a polymorphism (data not shown). Subsequent mutation screening of the unaffected father demonstrated that he carries the same 20bp-insertion, so we can conclude that this variant is not the cause of the phenotype observed in this patient.

In order to further investigate the effect of genetic variation in or nearby all coding regions of WNT4 on BMD, we selected 122 individuals with the most extreme t-scores for whole body BMD. In this sample cohort, we found three known polymorphisms in WNT4 of which none had a significant (p<0.05) difference in genotype frequencies between the high and low BMD cohort. One unknown synonymous (c.717T>C) SNP in exon 5 of WNT4 was detected once in the high BMD cohort (Table 2).

Next to screening the exons and intron-exon boundaries of WNT4, the SNPs that were associated with BMD at multiple skeletal sites in the GWAS were genotyped in our two BMD cohorts as well. Here, genotype frequencies of rs7521902 (WNT4) in the two cohorts were different, but this difference in genotype frequencies did not reach significance (p=0.195) when tested in an additive model. In a dominant model we observed a trend towards significance (p=0.071), suggesting an association with BMD in the same direction as the associations seen in the GWAS (Table 3).

Table 3 | Genotype frequencies of rs7521902 and rs2887571, two GWAS SNPs of WNT4 and WNT5B, in individuals of the young OAS cohort with the most extreme BMD values (t-score<−1.38 and t- score>1.54).

Genotype Genotype frequency p-value t-score t-score Additive model Dominant model <-1.38 >1.54 WNT4 CC 0.46 0.62 rs7521902 CA 0.47 0.33 0.195 0.071 AA 0.07 0.05 WNT5B AA 0.62 0.63 rs2887571 AG 0.33 0.32 0.892 0.745 GG 0.05 0.05

2. WNT5B

For WNT5B, mutation screening of all coding exons and intron/exon boundaries detected five known polymorphisms of which rs150051301 (MAF=0.0005) is a rare, synonymous variant that we detected in only one patient. The other four polymorphisms (rs35177332, rs58317077, rs6489313, rs2270036) are common intronic SNPs (MAF≥0.05) (Table 2).

115

Results – Chapter 5

Genetic screening of WNT5B in the two extreme BMD cohorts identified four known polymorphisms with no significant difference in genotype frequencies between the high and low BMD cohort. Moreover, an unknown 4bp-deletion (c.81-72_76delTGAA) in intron 2 of WNT5B was detected once in the high BMD cohort with no effect on the splicing of the WNT5B gene according to MutationTaster (data not shown)26(Table 2).

Finally, genotype frequencies of the GWAS SNP rs2887571 (WNT5B) were also compared between the high and low BMD cohort. Here, we observed nearly identical genotype frequencies in both BMD cohorts (Table 3).

IV. Discussion One of the latest meta-analyses of GWAS revealed that polymorphisms around the WNT4 (rs7521902) and WNT5B (rs2887571) gene are associated with femoral neck and lumbar spine BMD, while rs7521902 was also associated with the risk for fragility fractures7. Here, WNT5B was one of the newly associated genes, whereas the WNT4/ZBTB40 locus had already been associated with total hip and lumbar spine BMD by Styrkarsdottir et al.14. A transgenic mouse model expressing Wnt4 in osteoblasts further confirmed a role for Wnt4 in skeletal homeostasis, while the bone phenotype of Wnt5b knockout (Wnt5b-/-) mice was not consistent enough to draw conclusions16,17.

In this study we further investigated the causality between both common and rare genetic variation in WNT4 and WNT5B and deviating BMD values in two different populations. For this, we screened the exons and intron-exon boundaries of both genes with Sanger sequencing. No putative causal mutations were detected in our patient population, which implies that mutations in WNT4 and WNT5B are not a common cause of craniotubular hyperostosis. Next to our patient population, we screened two cohorts of healthy Danish men from the Odense Androgen Study population with either a high (t-score>1.54) or low (t-score<-1.38) whole body BMD. For the detected common variants in WNT4 and WNT5B, no significant difference in genotype frequencies between the high and low BMD cohort was observed. Similarly, the unknown variants we detected once in the high BMD cohort were not interesting for further functional studies, since they were either synonymous or intronic with no predicted effect on splicing. The SNPs that were associated with BMD at multiple skeletal sites in the GWAS, did not show a significant difference in genotype frequency in our two extreme BMD cohorts as well. This suggests that no association of genetic variation in WNT4 and WNT5B with whole body BMD was detected in our study with young Danish men. These findings do not fully correlate

116

Results – Chapter 5 with the current literature on WNT4 and WNT5B. Zheng and colleagues investigated the association between polymorphisms in WNT4 and WNT5B (among other) and peak BMD in a population of 1214 young Chinese men from 399 nuclear families by using SNPscan™ 15. Here, one SNP (rs10917157) in WNT4 and three SNPs (rs2240506, rs7308793, rs4765830) in WNT5B were associated with peak BMD. First of all, the difference in methodology could explain the different results coming from our genetic analyses. Zhang et al. studied the genetic character of peak BMD, whereas we selected our BMD cohorts based on extreme whole body BMD values. Due to differences in heritability at different skeletal sites, this could therefore explain the divergent genetic results from different studies on the same genes1. The use of subjects with a different ethnical background and part of nuclear families could also result in different genotype frequency patterns and therefore a difference in genotype-phenotype associations. Moreover, SNPscan™ was used to genotype a selection of SNPs across the complete WNT4 and WNT5B gene, both intronic and exonic. Since all SNPs with significant associations in their study were located in the deeper intronic regions of WNT4 and WNT5B, this could also demonstrate why we did not detect differences in genotype frequencies in the regions (exons and intron-exon boundaries) we screened with Sanger sequencing. Next, in contrast to our study, GWAS were able to identify associations of rs7521902 (WNT4) and rs2887571 (WNT5B) with BMD. Compared to the sample size from the GWAS, the absence of significant associations of these GWAS SNPs with whole body BMD in our study could be due to reduced power. Nevertheless, previous studies confirmed that the selection of an extreme high and extreme low BMD cohort from a larger population significantly increases the power28. Power calculation of our BMD cohorts indicates that we have the ability to detect genetic variations explaining 2% of the variance in BMD with a power of 80% (http://pngu.mgh.harvard.edu/~purcell/gpc/). Even though this is a large effect size, the cohorts have extreme low and high BMD values and the same BMD cohorts were used in our previous study on genetic variation in WNT16 and was here proven to be a successful tool in the identification of common and rare variants with a putative effect on BMD23. We can therefore conclude that the effect of genetic variation in WNT4 and WNT5B on BMD is probably too small to be picked up by these BMD cohorts, unlike for WNT16. Next to a possible difference in power, the allocation of the true causal variant(s) behind the (intronic) GWAS SNPs often remains elusive. It is therefore possible that the causal variant for rs7521902 and rs2887571 localizes in a neighboring gene. Adjacent genes of WNT4 and WNT5B are the ‘Zinc finger and BTB domain-containing protein 40’ (ZBTB40) and ‘ELKS/Rab6- interacting/CAST family member 1’ (ERC1) gene, respectively. So far, little is known on the ERC1 gene regarding its role in skeletal homeostasis, but ZBTB40 is a gene well-known for its

117

Results – Chapter 5 associations with BMD at different skeletal sites10-12,29-32. In other studies, associations with the ‘WNT4/ZBTB40’ locus (1p36.12) were detected14,33. Notwithstanding the numerous replications of the 1p36.12 association with BMD, this could therefore complicate the search towards the functional variant(s) with an effect on BMD, looking at the WNT4 gene alone.

In conclusion, despite the findings of the GWAS, we were not able to replicate or further verify the genetic association of polymorphisms in WNT4 and WNT5B with BMD. In order to do so, the intronic regions of both genes could be investigated more thoroughly in more extended populations (or extremes) with greater power to detect more rare variants and interesting common variants with smaller effect sizes. Future genetic and functional studies towards adjacent genes of WNT4 and WNT5B can also be interesting to figure out whether the signal from the GWAS could possibly be attributed to genetic variation in these genes.

V.--Acknowledgements This work was supported by the Systems Biology for the Functional Validation of Genetic Determinants of Skeletal Diseases (SYBIL) project funded by the European Commission and by a grant from the FWO (Fund for Scientific Research) Flanders (FWO grant G031915N). G.H. holds a Rosa Blanckaert Legacy Grant for Young Investigators from the University of Antwerp (Project ID 32435). E.B. holds a post-doctoral fellowship of the FWO Flanders (FWO personal grant 12A3814N).

118

Results – Chapter 5

References

1. Boudin, E., Fijalkowski, I., Hendrickx, G. & Van Hul, W. Genetic control of bone mass. Mol Cell Endocrinol (2015). 2. Siris, E.S. et al. Identification and fracture outcomes of undiagnosed low bone mineral density in postmenopausal women: results from the National Osteoporosis Risk Assessment. JAMA 286, 2815-22 (2001). 3. Hendrickx, G., Boudin, E. & Van Hul, W. A look behind the scenes: the risk and pathogenesis of primary osteoporosis. Nat Rev Rheumatol 11, 462-74 (2015). 4. Melton, L.J., 3rd. How many women have osteoporosis now? J Bone Miner Res 10, 175-7 (1995). 5. Randell, A. et al. Direct clinical and welfare costs of osteoporotic fractures in elderly men and women. Osteoporos Int 5, 427-32 (1995). 6. Duncan, E.L. et al. Genome-wide association study using extreme truncate selection identifies novel genes affecting bone mineral density and fracture risk. PLoS Genet 7, e1001372 (2011). 7. Estrada, K. et al. Genome-wide meta-analysis identifies 56 bone mineral density loci and reveals 14 loci associated with risk of fracture. Nat Genet 44, 491-501 (2012). 8. Medina-Gomez, C. et al. Meta-analysis of genome-wide scans for total body BMD in children and adults reveals allelic heterogeneity and age-specific effects at the WNT16 locus. PLoS Genet 8, e1002718 (2012). 9. Richards, J.B. et al. Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet 371, 1505-12 (2008). 10. Rivadeneira, F. et al. Twenty bone-mineral-density loci identified by large-scale meta- analysis of genome-wide association studies. Nat Genet 41, 1199-206 (2009). 11. Styrkarsdottir, U. et al. European bone mineral density loci are also associated with BMD in East-Asian populations. PLoS One 5, e13217 (2010). 12. Zhang, L. et al. Multistage genome-wide association meta-analyses identified two new loci for bone mineral density. Hum Mol Genet 23, 1923-33 (2014). 13. Hsu, Y.H. & Kiel, D.P. Clinical review: Genome-wide association studies of skeletal phenotypes: what we have learned and where we are headed. J Clin Endocrinol Metab 97, E1958-77 (2012). 14. Styrkarsdottir, U. et al. Multiple genetic loci for bone mineral density and fractures. N Engl J Med 358, 2355-65 (2008).

119

Results – Chapter 5

15. Zheng, Y. et al. Polymorphisms in Wnt signaling pathway genes are associated with peak bone mineral density, lean mass, and fat mass in Chinese male nuclear families. Osteoporos Int (2016). 16. Yu, B. et al. Wnt4 signaling prevents skeletal aging and inflammation by inhibiting nuclear factor-kappaB. Nat Med 20, 1009-17 (2014). 17. Brommage, R. et al. High-throughput screening of mouse gene knockouts identifies established and novel skeletal phenotypes. Bone Res 2, 14034 (2014). 18. Yang, Y., Topol, L., Lee, H. & Wu, J. Wnt5a and Wnt5b exhibit distinct activities in coordinating chondrocyte proliferation and differentiation. Development 130, 1003-15 (2003). 19. Nielsen, T.L. et al. Prevalence of overweight, obesity and physical inactivity in 20- to 29- year-old, Danish men. Relation to sociodem ography, physical dysfunction and low socioeconomic status: the Odense Androgen Study. Int J Obes (Lond) 30, 805-15 (2006). 20. Nielsen, T.L., Wraae, K., Brixen, K.T., Andersen, M. & Hagen, C. [The Odense Androgen Study]. Ugeskr Laeger 166, 1449-51 (2004). 21. Boudin, E. et al. No mutations in the serotonin related TPH1 and HTR1B genes in patients with monogenic sclerosing bone disorders. Bone 55, 52-6 (2013). 22. Boudin, E. et al. Mutations in sFRP1 or sFRP4 are not a common cause of craniotubular hyperostosis. Bone 52, 292-5 (2013). 23. Hendrickx, G. et al. Variation in the Kozak sequence of WNT16 results in an increased translation and is associated with osteoporosis related parameters. Bone 59, 57-65 (2014). 24. Koressaar, T. & Remm, M. Enhancements and modifications of primer design program Primer3. Bioinformatics 23, 1289-91 (2007). 25. Untergasser, A. et al. Primer3--new capabilities and interfaces. Nucleic Acids Res 40, e115 (2012). 26. Schwarz, J.M., Cooper, D.N., Schuelke, M. & Seelow, D. MutationTaster2: mutation prediction for the deep-sequencing age. Nat Methods 11, 361-2 (2014). 27. Balemans, W. et al. Identification of a 52 kb deletion downstream of the SOST gene in patients with van Buchem disease. J Med Genet 39, 91-7 (2002). 28. Sims, A.M. et al. Genetic analyses in a sample of individuals with high or low BMD shows association with multiple Wnt pathway genes. J Bone Miner Res 23, 499-506 (2008). 29. Chao, T.H., Yu, H.N., Huang, C.C., Liu, W.S. & Lu, K.H. Opposite associations of osteoprotegerin and ZBTB40 polymorphisms with bone mineral density of the hip in postmenopausal Taiwanese women. J Chin Med Assoc 75, 335-40 (2012).

120

Results – Chapter 5

30. Ho-Pham, L.T., Nguyen, S.C., Tran, B. & Nguyen, T.V. Contributions of Caucasian- associated bone mass loci to the variation in bone mineral density in Vietnamese population. Bone 76, 18-22 (2015). 31. Liu, J.M. et al. Analysis of recently identified osteoporosis susceptibility genes in Han Chinese women. J Clin Endocrinol Metab 95, E112-20 (2010). 32. Park, S.E. et al. Association of osteoporosis susceptibility genes with bone mineral density and bone metabolism related markers in Koreans: the Chungju Metabolic Disease Cohort (CMC) study. Endocr J 61, 1069-78 (2014). 33. Zheng, H.F. et al. Whole-genome sequencing identifies EN1 as a determinant of bone density and fracture. Nature 526, 112-7 (2015).

121

Results – Chapter 5

122

Part 2

Hyperostosis Cranialis Interna

123

124

Chapter 6

Conditional Mouse Models Support the Role of SLC39A14 (ZIP14) in Hyperostosis Cranialis Interna and in Bone Homeostasis

Gretl Hendrickx1*, Vere M. Borra1*, Ellen Steenackers1, Timur A. Yorgan2, Christophe Hermans3, Eveline Boudin1, Jérôme J. Waterval4, Ineke D.C. Jansen5, Tolunay Beker Aydemir6, Niels Kamerling7, Geert J. Behets8, Christine Plumeyer2, Patrick C. D’Haese8, Björn Busse2, Vincent Everts5, Martin Lammens9, Geert Mortier1, Robert J. Cousins6, Thorsten Schinke2, Robert J. Stokroos10, Johannes J. Manni10 and Wim Van Hul1

1Center of Medical Genetics, University and University Hospital of Antwerp, Belgium; 2Department of Osteology and Biomechanics (IOBM), University Medical Center Hamburg-Eppendorf, Germany; 3Center for Oncological Research Antwerp (CORE), University of Antwerp, Belgium; 4Department of Otorhinolaryngology, Radboud University Medical Center, The Netherlands; 5Department of Periodontology and Oral Cell Biology, Academic Center of Dentistry Amsterdam and VU University Amsterdam, The Netherlands; 6Food Science and Human Nutrition Department and Center for Nutritional Sciences, College of Agricultural and Life Sciences, University of Florida, Gainesville, FL, USA; 7Department of Neurosurgery, University Hospital Antwerp, Belgium; 8Department of Pathophysiology, University of Antwerp, Belgium; 9Department of Pathological Anatomy, University Hospital Antwerp, Belgium; 10Department of Otorhinolaryngology and Head & Neck Surgery, Maastricht University Medical Center, Maastricht, The Netherlands

* These authors contributed equally

This chapter is based on the research article submitted to PLoS Genetics

125

Results – Chapter 6

Abstract

Hyperostosis Cranialis Interna (HCI) is a rare bone disorder characterized by intracranial hyperostosis and osteosclerosis of the skull base and calvaria. We previously localized the disease- causing gene of HCI on chromosome 8p21. In this study, whole-exome sequencing detected a mutation (L441R) in the SLC39A14 (ZIP14) gene, encoding a zinc transporter. The L441R mutation results in trafficking defects of ZIP14 towards the plasma membrane and accumulates intracellular zinc in vitro. Immunohistochemistry indicates expression of ZIP14 in osteoblasts and osteoclasts. We therefore generated two conditional knock-in mouse models, overexpressing L438R Zip14 in osteoblasts and osteoclasts respectively. Remarkably, the calvariae of these mice were unaffected, whereas the rest of the skeleton was affected by L438R Zip14. Both knock-in models, and mainly the osteoblast-specific knock-in mice, exhibit an increased cortical thickness due to an increased endosteal bone formation, which is the underlying cause of intracranial osteosclerosis in patients with HCI. Trabecular bone volume was differentially modulated by L438R Zip14, i.e. decreased (osteoblast knock-in) or unaffected (osteoclast knock-in). As cAMP- CREB and NFAT signaling were also increased by L441R ZIP14, we postulate that actions of L441R ZIP14 are similar to estrogen and IGF-1 and opposite to PTH. Collectively, our results indicate the identification of the disease-causing gene for HCI and introduce it as a novel gene in the field of bone homeostasis.

126

Results – Chapter 6

I. Introduction

Sclerosing bone disorders are a group of monogenic skeletal dysplasias varying in severity, clinical and radiological presentation and inheritance pattern. These disorders are characterized by an extreme high bone mass due to a decreased bone resorption, an increased bone formation or increased bone turnover. Here, identification of the disease causing genes can give new insights in bone homeostasis and may lead to new therapeutic targets for other bone disorders, like osteoporosis1, 2. Although most of the sclerosing bone disorders affect the whole skeleton, some demonstrate with a significant craniofacial involvement (sclerosteosis, Van Buchem disease, autosomal dominant and recessive osteopetroses…), whereas only few conditions are known with exclusive involvement of the skull, like X-linked calvarial hyperostosis and Hyperostosis Cranialis Interna (HCI)3-6. HCI was first described in 1990 in a Dutch kindred by Manni et al. and to date, this is still the only family known to be described with HCI4. This rare bone disorder is characterized by endosteal hyperostosis and osteosclerosis of the calvaria and the skull base (Figure 1). Remarkably, there is no indication that the remainder of the skeleton is affected in HCI patients. The facial appearance and the shape of the head are normal. The first radiological abnormalities are often seen in the first decade, whereas the first symptoms occur late in the first decade or in adulthood4, 7. CT-scans show bony overgrowth at the level of the temporal bone and narrowing of the internal auditory canals, the optic canals and the orbital fissures. This results in the entrapment and dysfunction of cranial nerves I, II, V, VII and VIII, causing recurrent disturbance of smell, vision, sensation in the face, facial expression, hearing and balance, respectively (Figure 1)7. In addition, increased ocular and intracranial pressure can occur, leading to frequent headaches in some patients. The progressive functional impairment of facial and vestibulocochlear nerve function might be due to direct effect of pressure on the nerve fibers or impairment of the blood supply8, 9. Female HCI patients tend to be more severely affected by the disorder. Remarkable is that they often show severing of symptoms during pregnancy, suggesting a hormonal influence on disease progression7. Besides this, a slow progression of the disease can be seen until the age of 40. Hereafter, radiological progression is minimal and progressive cranial nerve dysfunction is uncommon in most patients. Nevertheless, untimely death may occur in severely affected cases, due to decreased intracranial volume7, 10. To date, no causal treatment is available for HCI, hence patients are treated symptomatically7, 9, 11. This may consist of hearing aids in case of moderate sensorineural hearing loss and in deaf patients, an auditory brainstem implant can be considered. In case of threatened vision due to

127

Results – Chapter 6 optical canal narrowing, optic nerve decompression can be performed. Trigeminal nerve neuralgia can be treated by a percutaneous nerve block or by radiofrequency therapy.

Figure 1 | 3D-reconstsruction of a normal skull base and one of a patient

with Hyperostosis Cranialis Interna. The left image is an unaffected individual in which all neuroforamina can be identified. The right image is an affected individual with thickened calvaria and a bulgy skull base; neuroforamina are hardly visible (from Waterval et al., AJNR 2012).

As previously mentioned, HCI was originally described in one large family from the Netherlands, consisting of three smaller subfamilies, with currently 13 affected family members over four generations. As a monogenic skeletal disorder, HCI has an autosomal dominant inheritance pattern. This genetic background of HCI has been investigated previously by performing a whole-genome scan and linkage analysis, where we assigned the locus for HCI to chromosome 8p21 in this family3. Interesting candidate genes in this region (BMP1, ADAM28, LOXL2) were screened with Sanger sequencing but no mutation was found. The aim of this study was to further look for the disease-causing gene for HCI and get insights in the mechanism underlying HCI. Therefore, whole-exome sequencing was performed, identifying a co-segregating missense mutation (p.L441R) in the SLC39A14 (ZIP14) gene, encoding a zinc transporter. In vitro studies were performed to investigate the subcellular localization, zinc transport function and intracellular zinc accumulation by p.L441R ZIP14. Furthermore, two conditional knock-in mouse models were generated, overexpressing p.L438R Zip14 in osteoblasts and osteoclasts. Thorough skeletal phenotyping of these mice was performed to unravel cell-specific effects of p.L438R Zip14 in vivo. Finally, to learn more about the pathogenesis of this disorder, histology of a HCI skull biopsy specimen was performed and luciferase reporter assays were done to look for aberrations in signaling pathways by p.L441R

128

Results – Chapter 6

ZIP14.

II. Material and methods

1. Patients

The family with HCI used in this study originates from The Netherlands and has been described in detail previously3, 4, 7.

2. Exome sequencing and gene identification

Peripheral blood was collected from 24 family members and five non-related partners. Genomic DNA was isolated from these blood samples using standard procedures. Exome sequencing was performed on a female patient with the NimbleGen SeqCap EZ Human Exome V2 enrichment panel on the HiSeq2000 (Illumina Inc., San Diego, CA, USA). Data analysis was performed with DNA Nexus (DNAnexus Inc., Mountain View, CA, USA; dnanexus.com). Variants were filtered for their absence in dbSNP and non-coding and synonymous variants were excluded. As published previously, we already defined a linkage region on chromosome 8 (chr8: 21,593,210-28,256,787). Variants present in this specific region were selected for further investigation. Possible variants were confirmed with Sanger sequencing on other family members. Non-covered exons were amplified by GoTaq DNA polymerase-mediated PCR (Promega Corporation, Madison, WI, USA) with primers covering the exons and the intron-exon boundaries. Sequencing was carried out with the ABI 310 Genetic Analyser (Thermo Fisher Scientific, Waltham, MA, USA), using an ABI Prism BigDye terminator cycle sequencing kit, version 1.1 (Thermo Fisher Scientific, Waltham, MA, USA). The BigDye XTerminator purification kit (Thermo Scientific, Waltham, MA, USA) was used to remove unincorporated BigDye terminators.

3. Expression constructs and in vitro mutagenesis

Wildtype (WT) human full length ZIP14 cDNA (NM_001128431.2) cloned in a pCMV6-XL6 vector was obtained from OriGene Technologies (Rockville, MD, USA) and the mutation (c.1322T>G, p.L441R ZIP14) was introduced using the QuickChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA). Similarly, a construct generating a truncated form of ZIP14 was created by inserting an early stopcodon (p.W22X ZIP14). This construct is used as a negative control for transfection experiments. Green fluorescent protein (GFP) fusion proteins for WT, mutant and truncated hZIP14 were generated using the above described expression constructs as template. A PCR amplification was

129

Results – Chapter 6 performed to disrupt the termination codon and create the correct restriction sites. Then, the complete region of interest was subcloned in a pEGFP-N1 vector (Clontech Laboratories Inc., Mountain View, CA, USA). As a control, all cloned products were sequenced with Sanger sequencing as previously described.

4. Subcellular localization

Human embryonic kidney (HEK) 293T cells were grown in DMEM medium with 10% FBS supplemented with 100 U/mL penicillin and 100 U/mL streptomycin (Life Technologies, San

Diego, CA). Cells were incubated at 37°C in humidified air containing 5% CO2. Twenty-four hours prior to transfection, cells were plated at a density of 1 x 105 cells/mL in 35mm glass bottom dishes coated with poly-D-lysine (MatTek Corporation, Ashland, MA, USA). HEK293T cells were transfected with WT, L441R or W22X ZIP14-GFP constructs using Fugene 6 (Promega Corporation, Fitchburg, WI, USA) according to manufacturer’s instructions in a 3:1 ratio (Fugene 6 : DNA). Since the mutation in patients with HCI is dominant, a heterozygous model was also created by co-transfecting WT and L441R ZIP14-GFP constructs. Forty-eight hours after transfection, cells were fixed with 4% formalin and washed with PBS (Thermo Fisher Scientific, Waltham, MA, USA). Fluorescent staining of the plasma membrane was performed by incubating the fixed HEK293T cells with 1µg/mL tetramethylrhodamine conjugate of wheat germ agglutinin (Thermo Fisher Scientific, Waltham, MA, USA) for 10 minutes and washed with PBS. Vectashield antifade mounting medium with 4',6-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA, USA) was used to preserve fluorescence and to stain the nucleus. High resolution images were obtained using an Eclipse Ti-E inverted microscope (Nikon, Tokyo, Japan) attached to a dual spinning disk confocal system (UltraVIEW VoX; PerkinElmer, Waltham, MA, USA) equipped with 405, 488 and 561nm diode lasers for excitation of blue, green and red fluorophores, respectively. Images were acquired and processed using Volocity 6.0.1 software (Improvision, PerkinElmer, Waltham, MA, USA).

5. Zinc transport

Uptake of 65Zn and accumulation of Zn2+ with FluoZin3-AM in HEK293T cells were performed as described before12-14. In short, for 65Zn-uptake, HEK293T cells were plated at a density of 5 x 105 cells/mL and transiently transfected with the WT, L441R or W22X ZIP14 expression vector, using the Effectene Transfection Reagent (Qiagen, Venlo, The Netherlands). An empty vector was used as a transfection control. Forty-eight hours after transfection, cells were washed with HBSS (pH 7.0, Thermo Fisher Scientific, Waltham, MA, USA) and incubated at 37°C in serum- 65 free DMEM containing Zn (GE Healthcare, Little Chalfont, UK) and 4µM ZnCl2 for 15

130

Results – Chapter 6 minutes. Cells were washed three times with wash buffer (0.9% NaCl, 10mM EDTA, 10mM HEPES) and then solubilized with 0.2% SDS and 0.2M NaOH for 1 hour. Uptake of 65Zn was measured with a γ-ray spectrometer. Total protein concentrations were measured with the Pierce BCA protein assay kit (Thermo Fisher Scientific, Waltham, MA, USA) and used as a normalizer. For Zn2+ accumulation, transfected HEK293T cells were incubated with 5µM FluoZin3-AM (Thermo Fisher Scientific, Waltham, MA, USA) in serum-free DMEM for 30 minutes at 37°C.

Cells were then stimulated with 40µM ZnCl2 after which fluorescence was measured at 494/516nm excitation/emission12.

6. Immunohistochemistry of bone tumor tissue

In collaboration with the Tumorbank of the Antwerp University Hospital (UZA, Antwerp, Belgium), tumor tissue specimens of patients with a giant cell tumor of bone and an osteoblastoma were obtained. Tissue specimens were fixed in 4% formaldehyde and paraffin embedded on a routine basis. Five µm-thick sections were subjected to heat-induced antigen retrieval by incubation in 10mM citrate buffer (pH 6.0) for 20 minutes at 97°C. Subsequently, endogenous peroxidase activity was quenched by incubating the slides in peroxidase blocking buffer (DAKO, Glostrup, Denmark) for 10 minutes. Incubation with primary anti-human ZIP14 antibody (PA5-21077, Thermo Fisher Scientific, Waltham, MA, USA, 1:200 dilution) was performed at room temperature for 1 hour. After a washing step, bound antibody was detected with the Envision FLEX+ detection kit (DAKO) using 3,3’-diaminobenzidine chromogen solution (DAKO, Glostrup, Denmark), according to the manufacturers’ instructions. Positive controls were included in each staining run and consisted of pancreas tissue, as expression of ZIP14 in pancreatic tissue is well-documented in the literature. Sections were counterstained with haematoxylin, dehydrated and mounted.

7. Expression of Zip14 in KS483 cells and osteoclasts

The KS483 cell line, a murine pre-osteoblast cell line with mesenchymal characteristics, was used to examine the expression of mZip14 during the differentiation to mature and mineralizing osteoblasts. KS483 cells were grown in α-MEM with GlutaMAX (Thermo Fisher Scientific, Waltham, MA, USA) with 10% FBS (Lonza, Basel, Switzerland) supplemented with penicillin- streptomycin (Thermo Fisher Scientific, Waltham, MA, USA). Cells were plated at a density of 2 4 x 10 cells/mL in a 24-well plate and incubated at 37°C in humidified air containing 5% CO2. After plating, RNA was extracted at day 4, 7, 11, 14, 18, 21, 24 and 28 with the ReliaPrep RNA Cell Miniprep System (Promega, Madison, WI, USA) and reverse transcribed with an oligo-dT primer and Superscript II Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA, USA).

131

Results – Chapter 6

Quantitative real-time PCR analysis was performed on all samples with qPCR Core kit for SYBR Green I, No Rox (Eurogentec, Liège, Belgium). For each sample, mZip14 expression was analyzed and normalized to b2m, rpl13a and ubc expression. These reference genes were selected for their stable expression throughout the differentiation process of KS483 cells. Stability of reference genes was verified using geNorm (Biogazelle, Ghent, Belgium) and efficiency of all primer pairs was checked with the qbase+ software (Biogazelle, Ghent, Belgium). Expression of target and reference genes was quantified using qbase+ software. To assess expression of Zip14 in osteoclasts, bone marrow cells from calvaria and long bones were isolated from mice as previously described15. Expression of mZip14 was checked in osteoclasts cultured on plastic or bovine cortical bone slices with supplementation of M-CSF or M-CSF with RANKL. RNA from cultured bone marrow cells was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and reversed transcribed to cDNA for quantitative real-time PCR. Samples were normalized for the expression of b2m15. All primer sequences are available upon request.

8. Histology of human skull biopsy samples

An occipital skull bone biopsy was taken during neurosurgical intervention from a 29-year old female patient with HCI, after receipt of informed consent by the patient. The biopsy specimen was fixed in 4% paraformaldehyde, decalcified and embedded in paraffin. Sections were stained by standard hematoxylin-eosin staining procedures. As a control sample, an occipital skull bone biopsy was taken during neurosurgical intervention from a 37-year old female patient with a posterior fossa meningioma, after receipt of informed consent. Peripheral blood was collected from the patient for the isolation of genomic DNA and genetic screening of ZIP14 with Sanger sequencing. The biopsy specimen was fixed, decalcified, embedded and stained according to the same procedures as described above. Quantification of the number of Haversian channels and osteocytes was performed on three microscopic images of the patient and control externae/internae of the skull and of the patient vertebral cortex. This collaborative study with the neurosurgical team of the Antwerp University Hospital (UZA) was approved by the Committee for Medical Ethics of the Antwerp University Hospital.

9. µCT of Zip14-/- mice

Heterozygous Zip14 knockout (Zip14+/−) mice of the C57BL/6 strain were obtained from the Mutant Mouse Research Resource Consortium at the University of California, Davis via a contract. A breeding colony was established at the University of Florida, generating homozygous (Zip14+/+) WT and homozygous Zip14 knockout (Zip14−/−) mice16, 17. Zip14-/- (n=7) and Zip14+/+

132

Results – Chapter 6 mice (n=6) were fixed in 10% formalin and stored in 70% EtOH. µCT scans of the calvaria were generated with the SkyScan1076 system (Bruker microCT, Kontich, Belgium). Images were reconstructed with NRecon software and data were analyzed with Dataviewer and CTAn (Bruker microCT, Kontich, Belgium). Bone volume, structural thickness and fractal dimension were measured at the calvariae. Nomenclature, symbols and units used are those recommended by the Nomenclature Committee of the American Society of Bone and Mineral Density18.

10. Generation of a mouse model for HCI

The mutated leucine at amino acid position 441 in ZIP14 of HCI patients is highly conserved in mice and corresponds to the amino acid mL438 of both isoforms of mZip14 (NP_001128624.1; NP_659057.2). As no difference in function between both isoforms was reported, wildtype full length mZip14 cDNA corresponding to NP_001128624.1 cloned in a pCMV6-Entry vector was obtained from OriGene Technologies (MC216777, Rockville, MD, USA). The mutation resulting in the p.L438R amino acid change was inserted using the QuickChange Site-Directed Mutagenesis kit (Agilent Technologies, Santa Clara, CA, USA). This construct was sent to genOway (Lyon, France) to create a mouse model with Zip14L438R through targeted insertion within the ROSA26 locus via homologous recombination in embryonic stem cells. The expression of the mutated Zip14 is under control of a CAG promoter. A loxP-flanked transcriptional STOP cassette is incorporated between the mutated Zip14 and the CAG promoter to allow the expression of the resulting transgene to be dependent upon the Cre recombinase (Figure 2A). For breeding, Sox2-Cre mice, Runx2-Cre mice and CtsK-Cre mice were kindly provided by Vincent Timmerman and Delphine Bouhy19 (University of Antwerp, Belgium), Jan Tuckermann20 (Universität Ulm, Germany) and Rachel Davey21 (University of Canberra, Australia), respectively. Skeletal phenotyping was performed at the age of 6 months, corresponding to the age of 30 years in humans at which the HCI phenotype is prominent22. Since no gender-specific differences were found, only the data from male mice are presented in this manuscript. Mice were given two injections of 30 mg/kg calcein at 9 and 2 days before death to assess dynamic histomorphometric indices. At least six mice per group were subjected to histomorphometry and serum analysis to obtain sufficient results to perform statistical analyses. All mice were maintained on a twelve-hour light-dark cycle, with a regular unrestricted diet available ad libitum. All animal experiments were conducted according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Committee of Medical Ethics of the University of Antwerp. Mice homozygous for the floxed mutant Zip14 allele (Zip14flox/flox) were crossed with heterozygous Sox2-Cre, Runx2-Cre and CtsK-Cre mice.

133

Results – Chapter 6

Figure 2 | Generation and genotyping of a floxed Zip14L438R mouse model. (A) A mouse model with floxed Zip14L438R was generated through targeted insertion within the ROSA26 locus. A loxP-flanked transcriptional STOP cassette is incorporated between Zip14L438R and its CAG promoter to allow the expression of the resulting transgene to be dependent upon the Cre recombinase. (B) A first PCR for genotyping (left) is to detect the Zip14flox and Cre-mediated excised (Zip14L438R) locus, with amplicons of 3428bp and 410bp in size, respectively, whereas the wildtype allele gives no amplification. A second PCR (right) is performed to distinguish homozygous Zip14flox/flox (998bp), heterozygous Zip14flox/- or Zip14L438R/- (998bp + 304bp) and homozygous wildtype (304bp) mice.

Offspring was weaned after 3 weeks and marked by ear clipping. DNA, isolated from the tail tip, was used for genotyping of the ROSA26 locus by performing two PCRs (Figure 2B). The first PCR is to detect the Cre-mediated excision of the loxP flanked STOP cassette within the recombined ROSA26 allele and was performed with a forward primer located within the mutated Zip14 (cDNA) sequence and a reverse primer located downstream of the STOP-cassette within the CAG promoter. Here, a Zip14flox and Cre-mediated excised (Zip14L438R) locus give amplicons of 3428bp and 410bp in size, respectively, whereas a wildtype allele gives no amplification. Additionally, a second PCR is performed to distinguish homozygous Zip14flox/flox (998bp

134

Results – Chapter 6 fragment), heterozygous Zip14flox/- or Zip14L438R/- (998bp+304bp fragments) and homozygous wildtype (304bp fragment) mice. The Expand Long Template PCR System (Roche, Basel, Switzerland) and dNTP solution mix (Bio-Rad Laboratories, Hercules, CA, USA) are used for both genotyping PCRs. Fragments were separated on a 2% agarose gel simultaneously running a GeneRuler 100bp Plus DNA Ladder and GeneRuler 1kb DNA Ladder (Thermo Fisher Scientific, Waltham, MA, USA) (Figure 2B). In offspring from breedings with Runx2-Cre and CtsK-Cre mice, a third PCR is performed to check the corresponding Cre-allele. Here, standard GoTaq DNA polymerase-mediated PCR reactions (Promega Corporation, Madison, WI) were performed and verified by agarose gel electrophoresis simultaneously running a Generuler 100 bp Plus DNA ladder (Thermo Fisher Scientific, Waltham, MA, USA). Mice heterozygously carrying the Runx2-Cre allele show a wildtype band (780bp) and Runx2-Cre band (600bp) by using two forward primers and one reverse primer in the PCR. The CtsK-Cre allele amplifies a fragment of 533bp, whereas wildtype alleles show no amplification.

11. Skeletal phenotyping of mice

Dissected skeletons were fixed in 3.7% PBS-buffered formaldehyde for 18 hours at 4°C and stored in 80% ethanol. All mice were analyzed by contact X-ray and µCT scanning. For the latter, a µCT 40 desktop cone-beam µCT (Scanco Medical) was used and reconstructed slices were examined using the Scanco MicroCT software suite. To assess biomechanical properties of the femora, three-point bending assays and a quantitative backscattered electron imaging (qBEI) analysis were performed as described23-26. The lumbar vertebral bodies (L1-L4) and one tibia were dehydrated in ascending alcohol concentrations and embedded in methylmethacrylate as previously described26. Parameters of structural and cellular histomorphometry were quantified on Von Kossa/Van Gieson and toluidine blue stained sections, respectively, of 4µm thickness. Analysis of bone volume, trabecular number, trabecular spacing, trabecular thickness, and the determination of osteoblast and osteoclast numbers and surface were carried out according to standardized protocols using the OsteoMeasure histomorphometry system (OsteoMetrics, Atlanta, GA, USA). Dynamic histomorphometry was performed on unstained 12µm sections of the vertebral bodies and tibia as previously described26.

12. Biochemical assays

ELISA was used to determine serum concentrations of procollagen I C-terminal propeptide (PICP; SEA570Mu, USCN Business Co., Houston, TX, USA), C-terminal telopeptide (RatLaps (CTX-I) EIA, AC-06F1, Immunodiagnostic Systems, Boldon, UK), osteoprotegerin (OPG;

135

Results – Chapter 6

MOP00, R&D Systems, Minneapolis, MN, USA) and receptor activator of nuclear factor kappa- B ligand (RANKL; MTR00, R&D Systems, Minneapolis, MN, USA).

13. Luciferase reporter assays

HEK293T and Saos-2 cells were grown in DMEM (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with FBS (10% v/v). Twenty-four hours prior to transfection, cells were plated at 0.3 x 105 cells/well in 96-well plates. Cells were transiently transfected with pRL-tK (2,5ng) and pCRE-Luc, NF-kB-Luc or pGL4.30 (NFAT-Luc, Promega Corporation, Madison, WI) (25ng) along with 20ng of empty pcDNA3.1 vector, WT, L441R or W22X ZIP14 expression constructs using Fugene 6 (HEK293T cells) or ViaFect (Saos-2 cells) (Promega Corporation, Madison, WI) according to the manufacturer’s instructions. Each transfection was carried out in triplicate and repeated independently in three separate experiments. Forty-eight hours after transfection, cells were lysed and firefly and renilla luciferase activity were measured on a Glomax Multi+ Luminometer (Turner Designs, Sunnyvale, CA) using the dual luciferase reporter assay system (Promega Corporation, Madison, WI) according to manufacturer’s instructions. Finally, the ratio of the firefly and renilla luciferase measurement was calculated.

14. Statistics

All data are presented as mean values ± SD and analyzed by a one-way ANOVA or a two-tailed Student’s t-test. Both statistical tests were provided by the SPSS v22.0 software package (SPSS Inc, Chicago, IL, USA). Statistical analysis of the mouse phenotyping data was performed by comparing the results of osteoblast knock-in mice and osteoclast knock-in mice with those of heterozygous Zip14flox animals. Here, a value of p<0.05 (*) and p<0.025 (**) were considered statistically signiﬁcant and significant after Bonferroni correction, respectively.

III. Results

1. Identification of SLC39A14 (ZIP14) as disease causing gene for HCI

Whole-exome sequencing (WES) was performed on one affected individual from the family with HCI. The average coverage throughout the whole exome was 66x. After filtering variants for their absence in dbSNP and excluding non-coding and synonymous variants, we focused on the variants present in the linkage region on chromosome 8 (chr8: 21,593,210-28,256,787) after which only two variants remained (Figure 2A). Both variants, one in SCARA3 with a 5x coverage and one in ZIP14 with a 66x coverage, were checked with Sanger sequencing. The variant in the SCARA3 gene appeared to be a false positive, since we could not confirm it in the patient. The

136

Results – Chapter 6 other variant is a heterozygous c.1322T>G substitution, leading to a p.L441R amino acid substitution, in the solute carrier family 39 member 14 (SLC39A14 or ZIP14) gene and co-segregates with the disease in the complete family (Figure 2B). This missense variant was not found in 100 control individuals with the same ethnic background that were screened with Sanger sequencing and is not present in sequence databases, including dbSNP, 1000 Genomes Project and ExAc databases (URLs). Eighteen exons from the linkage region remained partially or completely uncovered by WES. These exons were all amplified by PCR and checked with Sanger sequencing without identification of any unknown variants. These results indicate that the c.1322T>G variant found in SLC39A14 is the only coding variant in the 8p21 region previously linked to HCI, confirming its disease causality.

2. Characterization of the p.L441R variant in ZIP14

The human SLC39A14 gene has four protein coding isoforms according to the National Center for Biotechnology Information (NCBI), all consisting of nine exons. The heterozygous c.1322T>G substitution in exon 8 of SLC39A14 affects all isoforms of the gene and results in a p.L441R substitution (Figure 2C), disturbing a highly conserved amino acid. Accordingly, this missense mutation has a Combined Annotation Dependent Depletion (CADD) score of 29.4, indicating that it belongs to the top 0.11% most deleterious substitutions that can occur in the human genome27. As a zinc (Zn) transporter, ZIP14 has six or eight transmembrane domains, depending on the literature or prediction program used (TMHMM, MEMSAT, PRED-TMR, HMMTOP)28, 29. Nevertheless, the p.L441R mutation is always located at the end of the second- to-last transmembrane domain of ZIP14. All transmembrane prediction programs predict the variant to cause one or more shifts in a preceding, the affected or the following transmembrane domain, due to the replacement of a hydrophobic leucine by a hydrophilic arginine.

3. The p.L441R mutation affects localization and function of ZIP14 in vitro

To evaluate changes in the subcellular localization of mutant (L441R) ZIP14, HEK293T cells were transfected with wildtype (WT), L441R or truncated (W22X) ZIP14-GFP constructs and visualized with confocal microscopy (Figure 3A). WT ZIP14 is located on the plasma membrane and in the cytoplasm, as previously reported14, 30-33. In contrast herewith, L441R ZIP14 is not present on the plasma membrane, but appears to be clustered in the cytoplasm. The heterozygous model (WT/L441R ZIP14) clearly shows increased expression in the cytoplasm (compared to WT) and some co-localization on the plasma membrane. W22X ZIP14 shows expression in the cytoplasm as well as in the nucleus. Moreover, there is a difference in

137

Results – Chapter 6

Figure 2 | Identification of the c.1322T>G (p.L441R) mutation in SLCA39A14 (ZIP14). (A) Whole-exome sequencing was performed on one patient with Hyperostosis Cranialis Interna. Variants were filtered for their absence in dbSNP, by excluding non-coding and synonymous variants and its presence in the linkage region on chromosome 8 (chr8: 21,593,210-28,256,787) after which only two variants remained. (B) Identification of the c.1322T>G mutation in exon 8 of the SLC39A14 (ZIP14) gene by Sanger sequencing. (C) Localization of the p.L441R mutation in the fifth transmembrane domain of ZIP14. cytoplasmic distribution of the different ZIP14 forms, i.e. both WT and L441R ZIP14 appear to be clustered in similar vesicular-shaped structures, whereas W22X ZIP14 is uniformly spread across the cytoplasm (Figure 3A). 65Zn uptake and Zn accumulation studies were performed to assess the basic functionality of p.L441R ZIP14 as a transporter of Zn (and other metals) (Figure 3B). 65Zn uptake experiments revealed that overexpressing WT ZIP14 significantly (p<0.001) increases 65Zn uptake by 4-fold

138

Results – Chapter 6 when compared to cells transfected with empty vector. On the contrary, L441R and W22X ZIP14 showed no sign of 65Zn uptake from the extracellular space into the cell (Figure 3B). This was no surprise, since L441R and W22X ZIP14 were no longer detected on the plasma membrane of the cells. FluoZin3-AM measures the accumulation of labile Zn in the cell. Results show that there is a significant (p<0.05) increase in Zn accumulation in cells overexpressing WT ZIP14. Overexpression of L441R ZIP14 also results in a significant (p<0.001) increase in intracellular Zn accumulation, which is greater than for WT ZIP14, indicating that there might be labile Zn trapped in cells with L441R ZIP14 (Figure 3B). Altogether, L441R ZIP14 no longer reaches the plasma membrane, but still resides in cytoplasmic structures from where it causes an entrapment of labile Zn.

4. ZIP14 is expressed in osteoblasts and osteoclasts

ZIP14 was reported to be expressed in many tissues with increased expression in the liver, pancreas, thyroid gland, heart and intestine, and a low expression in the brain28. Information on the expression of ZIP14 in skeletal cell types (osteoblasts, osteoclasts, osteocytes) has not been reported in the literature. We therefore performed immunohistochemistry on sections of giant cell tumor and osteoblastoma tissue, bone tumors known to be enriched with osteoclast-like giant cells and osteoblasts, respectively. Here, expression of ZIP14 was detected in osteoblasts of osteoblastoma tissue and giant cells from giant cell tumor tissue (Figure 4A). ZIP14 was not expressed in osteocytes of osteoblastoma or giant cell tumor tissue. Moreover, quantitative real- time PCR was performed on KS483 cells, murine mesenchymal stem cells, to assess expression level of murine Zip14 (mZip14) during the different phases of osteoblast differentiation to a mature mineralizing osteoblast. Our results, depicted in Figure 4B, indicate that expression of mZip14 is stable during proliferation (first week) and maturation (second week) of osteoblast differentiation and rises during the mineralization phase (day 18-21). Lastly, Zip14 expression was checked with quantitative real-time PCR in murine osteoclasts derived from calvaria and long bones. Here we also detected expression of mZip14 in both osteoclastic cell populations, but in osteoclasts derived from the calvaria we found an average 2-fold greater expression of mZip14 in osteoclasts (Figure 4C).

139

Results – Chapter 6

Figure 3 | Subcellular localization, 65Zn uptake and labile Zn accumulation of wildtype (WT), mutant (L441R) and truncated (W22X) ZIP14 in HEK293T cells. (A) HEK293T cells were transfected with WT, L441R or W22X ZIP14-GFP. WT ZIP14 is located on the plasma membrane and in the cytoplasm, whereas L441R ZIP14 is no longer present on the plasma membrane, but appears to be clustered in the cytoplasm. The heterozygous model (WT/L441R ZIP14) shows increased expression in the cytoplasm (compared to WT) and some co-localization on the plasma membrane. W22X ZIP14 shows expression in the cytoplasm as well as in the nucleus. Scale bars, 10µm. (B) 65Zn uptake experiments demonstrate that WT ZIP14 significantly (p<0.001) increases 65Zn uptake by 4-fold when compared to cells transfected with empty vector (Empty V.). L441R and W22X ZIP14 show no sign of 65Zn uptake from the extracellular space into the cell. (C) FluoZin3-AM experiments demonstrate a significant (p<0.05) increase in Zn accumulation in cells overexpressing WT ZIP14. Overexpression of L441R ZIP14 results in a stronger (p<0.001) increase in intracellular zinc accumulation. 140

Results – Chapter 6

Figure 4 | Expression of ZIP14 in skeletal cell types. (A) Immunohistochemistry of osteoblastoma and giant cell tumor tissue depicts ZIP14 expression in osteoblasts (black line), in giant osteoclast-like cells (arrow-heads) and not in osteocytes. Scale bars upper figures, 500µm; scale bars lower figures, 100µm (B) Expression of murine Zip14 (mZip14) during differentiation of KS483 cells to mature mineralizing osteoblasts, indicating stable expression of mZip14 during proliferation (day 4-7) and maturation (day 11-14) of osteoblast differentiation and rising expression during mineralization (day 18-21). (C) In vitro generated osteoclasts from murine calvariae and long bones were cultured on plastic or cortical bone slices, supplemented with M- CSF (M) or M-CSF + RANKL (M+R). qPCR analysis indicates expression of mZip14 in osteoclasts derived from calvariae and long bones.

141

Results

142

–

Chapter 6 Chapter

Results – Chapter 6

Figure 5 | Micro-structural analysis of bone biopsy specimens of a control individual and of a patient with Hyperostosis Cranialis Interna (HCI). (A) Microscopic images of the external cortex (externa), diploë and internal cortex (interna) of a patient and control skull biopsy specimen. The structure of the interna and externa of the control are both comparable to the patient externa, whereas the patient interna is clearly affected. The latter demonstrates with a dense amount of well-organized bone and less osteocytes that are grouped around the Haversian channels. (B) The cortex and trabeculae of the first cervical (C1) vertebra of the patient show no micro-structural differences with the externa or diploë of the skull of this same patient. (C) Quantitative histological analysis of the number of Haversian channels (N.Haversian channels) and number of osteocytes (N.Ot) in the patient and control internae and externae and the cortex of the first cervical vertebra (Ct.VC1) of the patient. In the patient, the number of Haversian channels and osteocytes were both significantly lower in the interna, compared to the externa and the cortex of the cervical vertebra. When comparing the internae of patient and control, both parameters were significantly lower in the patient interna. The externae of patient and control only differ in the number of Haversian channels. *:p<0.05; **:p<0.01. Scale bars, 500µm.

5. Hyperostosis Cranialis Interna exclusively affects the inner cortex of the skull

A skull and first cervical vertebra biopsy specimen were obtained from a patient with HCI as well as a skull biopsy from an age- and sex-matched control during a neurosurgical intervention. All fragments were embedded in paraffin, sectioned and stained with H&E to further examine the micro-structure of the internal cortex (interna), diploë and external cortex (externa) of the skull (Figure 5A). First, in the control sample we did not find significant microscopic differences between the interna and externa (Figure 5C), but when we look into the patient samples, it is clear that the interna is severely affected. The number of Haversian channels and osteocytes is significantly lower in the patient interna, compared to the externa and the cortex of the cervical vertebra of the patient (Figure 5C). When we compare the externa of the patient with that of the control, the number of osteocytes was significantly lower (p=0.0054) in the patient (Figure 5C), although osteocyte distribution appears comparable (Figure 5A). Comparing the patient and control internae, however, demonstrates that the patient interna is wider and characterized by a great and dense amount of well-organized bone, suggesting an increased bone formation or decreased bone resorption. Moreover, the number of Haversian channels (p= 0.0075) and the number of osteocytes (p=0.0042) are significantly lower in the patient interna, compared to interna of the control. Remarkably, the osteocytes in the patient interna appear grouped around the Haversian channels. Some osteocyte lacunae, especially further away from the Haversian channels, appear empty, suggesting osteocyte apoptosis. This was not seen in the patient externa or vertebral tissue or in the skull of the control.

143

Results – Chapter 6

-/- Figure 6 | µCT analysis of calvariae of Zip14 knockout (Zip14 ) mice and conditional Zip14L438R knock-in mice. (A) Calvarial thickness (Calv.Th) and porosity (Calv.Po) measured at -/- +/+ the calvariae of Zip14 and Zip14 mice shows no significant (p<0.05) differences. (B) 3D reconstruction of calvariae of controls (Zip14fl/-), osteoblast-specific knock-in mice (Zip14fl/-; Runx2-Cre) and osteoclast-specific knock-in mice (Zip14fl/-; CtsK-Cre). (C) Quantitative µCT analysis of calvarial thickness and porosity. Calvarial porosity appears lower in Zip14fl/-; Runx2- Cre mice, but no significant (p<0.05) differences were observed in Zip14fl/-; Runx2-Cre and osteoclast-specific knock-in mice (Zip14fl/-; CtsK-Cre). N=6 animals/genotype.

6. Zip14-/- mice do not have a calvarial phenotype

Zip14-/- mice were previously generated at the University of Florida, USA16. These Zip14-/- mice exhibit dwarfism and a general osteoporotic phenotype in the appendicular skeleton and vertebral column, characterized by a decrease in trabecular bone volume, whereas the cortical bone appears unaffected34. As no information was available on the calvarial phenotype of these mice, we performed µCT analysis of the calvaria of Zip14+/+ and Zip14-/- mice and analyzed the data with the CTAn software (Bruker microCT, Kontich, Belgium). Calvarial thickness (Calv.Th) and porosity (Calv.Po) were measured, but no significant differences were noticed between Zip14+/+ and Zip14-/- mice (Figure 6A).

144

Results – Chapter 6

7. Ubiquitous expression of Zip14L438R in vivo results in perinatal lethality

An in vivo model to study the effect of ZIP14L441R was generated by creating a floxed mutant Zip14 (Zip14flox) mouse model to express Zip14L438R ubiquitously (Sox2-Cre) or in specific cell types, i.e. osteoblasts (Runx2-Cre) and osteoclasts (CtsK-Cre). Breeding Zip14flox/flox mice with Sox2-Cre mice demonstrated that ubiquitous expression of mutant Zip14 results in perinatal lethality. We therefore focused on mice with conditional expression coming from breedings with Runx2-Cre and CtsK-Cre mice. In total, 6-month old Zip14fl/- controls (n=6), Zip14fl/-; Runx2- Cre (osteoblast-specific knock-ins, n=6) and Zip14fl/-; CtsK-Cre (osteoclast-specific knock-ins, n=6) were collected for skeletal phenotyping. No gender-specific differences were observed (data not shown), so the results presented in this article are solely these from the skeletal analysis of male mice.

8. Osteoblast expression of Zip14L438R differentially modulates cortical and trabecular bone in vivo

µCT analysis of the calvaria and femora was performed to unravel structural differences of osteoblast-specific Zip14L438R knock-in mice versus Zip14fl/- controls. Although calvarial porosity appears lower in these mice there were no significant differences in calvarial thickness or porosity (Figure 6B). In contrast herewith, µCT analysis of the femora showed an obvious skeletal phenotype versus controls (Figure 7). Compared to Zip14fl/- controls, the osteoblast knock-in mice had a significantly increased cortical thickness (Ct.Th, p=6.0E-6) with a decreased cortical porosity (Ct.Po, p=0.0014) and a significantly smaller midshaft diameter (Ms.D, p=4.1E-6) (Figure 7A-B). Furthermore, osteoblast knock-in mice have a significantly decreased trabecular bone volume (BV/TV, p=0.0071), trabecular number (Tb.N, p=0.033) and connecting density (Conn.D, p=0.018) with an increased trabecular separation (Tb.S, p=0.035) (Figure 7C). X-ray radiographs of the whole skeletons indicated a fracture with callus in the tibiae of two osteoblast-specific knock-in mice (arrow, Figure 8A). Moreover, as seen in the µCT analysis, X- rays also revealed severe narrowing of the femoral midshaft in these mice (arrowhead, Figure 8A). Assessment of the biomechanical properties of the femora with three-point bending tests indicated that they bear significant higher stress levels (p=5.0E-4) but work-to-fracture was 42% percent lower (p=0.0086) than of femora of controls, probably due to the observed changes in cortical thickness and midshaft diameter. The elastic modulus (p=0.013) and work to reach ultimate stress levels (p=0.021) were also significantly lower in femora of these mice, suggesting more elastic femora (Figure 8B). Consequently, qBEI analysis indicated a significantly reduced cortical mineralization (Ct.CaMean, p=0.026), contributing to this increased flexibility.

145

Results – Chapter 6

Figure 7 | µCT analysis of femora of controls (Zip14fl/-) and conditional Zip14L438R knock- in mice. (A) 3D reconstruction of whole femora (top row) and a vertical section of cortical (middle row) and trabecular bone (bottom row) of 6-month old Zip14fl/- controls, osteoblast- specific knock-in mice (Zip14fl/-; Runx2-Cre) and osteoclast-specific knock-in mice (Zip14fl/-; CtsK- Cre). Femora of Zip14fl/-; Runx2-Cre mice show an increased cortical thickness and decreased midshaft diameter along with a decreased trabecular bone mass. (B) Quantitative µCT analysis of cortical (Ct) bone parameters confirms a significantly increased cortical thickness (Ct.Th) and decreased midshaft diameter (Ms.D) of Zip14fl/-; Runx2-Cre mice. Both Zip14fl/-; Runx2-Cre and Zip14fl/-; CtsK-Cre mice have a decreased cortical porosity (Ct.Po). (C) Quantitative µCT analysis of trabecular (Tb) bone parameters confirms a significantly decreased trabecular bone volume (BV/TV), number (Tb.N), connecting density (Conn.D) and increased separation (Tb.Sp) in Zip14fl/-; Runx2-Cre mice. N=6 animals/genotype; *: p<0.05; **: p<0.025.

146

Results – Chapter 6

Figure 8 | Biomechanical, structural and cellular properties of the skeletal phenotype of controls (Zip14fl/-), osteoblast-specific knock-in mice (Zip14fl/-; Runx2-Cre) and osteoclast-specific knock-in mice (Zip14fl/-; CtsK-Cre). (A) X-ray images of femora-tibiae of a Zip14fl/-and Zip14fl/-; Runx2-Cre mouse. The latter demonstrates with a tibial fracture (arrow) and narrowing of the femoral midshaft can be observed (arrowhead). (B) Three-point bending analysis indicates that femora of Zip14fl/-; Runx2-Cre mice tolerate higher stress levels and are more elastic. Work-to-fracture is also significantly reduced and qBEI analysis indicates significant lower cortical mineralization (Ct.CaMean) of femora of these mice, compared to controls. (C) Representative undecalcified spine (upper row) and tibia sections (bottom row) from 6 months old Zip14fl/-, Zip14fl/-; Runx2-Cre and Zip14fl/-; CtsK-Cre mice stained with von Kossa/van Gieson. Vertebrae of Zip14fl/-; Runx2-Cre mice show less trabecular bone, whereas tibiae of these mice show an increased cortical thickness and decreased midshaft diameter compared to Zip14fl/- controls. (D) Quantification of the bone surface covered by osteoblasts (Ob.S/BS), osteoblast number per bone perimeter (N.Ob/B.Pm), osteoclast surface per bone surface (Oc.S/BS) and osteoclast number per bone perimeter (N.Oc/B.Pm) in the vertebral bodies analyzed using toluidine blue staining. N=6 animals/genotype; *: p<0.05; **: p<0.025 (compared to Zip14fl/- mice).

147

Results – Chapter 6

Figure 9 | Dynamic histomorphometric analysis of 6-month old control (Zip14fl/-), osteoblast-specific knock-in (Zip14fl/-; Runx2-Cre) and osteoclast-specific knock-in (Zip14fl/-; CtsK-Cre) mice. (A) Dynamic histomorphometry of the tibial endocortical (Ct) bone surface indicates a significant increase in endosteal mineralizing surface (MS/BS) and bone formation rate (BFR/BS) in Zip14fl/-; Runx2-Cre. In Zip14fl/-; CtsK-Cre mice a significant increase in MS/BS was detected. (B) Dynamic histomorphometry of trabecular (Tb) bone surface of vertebral bodies (L3-L4) indicates a significant lower MS/BS and BFR/BS in Zip14fl/-; CtsK-Cre mice. N=6 animals/genotype; *: p<0.05; **: p<0.025 (compared to Zip14fl/- mice).

Undecalcified sections of lumbar vertebral bodies and tibiae were stained with Von Kossa/Van Gieson staining, as depicted in Figure 8C. Quantification of parameters of structural histomorphometry confirmed the trabecular phenotype observed with µCT analysis (data not shown). Sections stained with toluidine blue were analyzed to further investigate the skeletal phenotype on a cellular level. In the osteoblast knock-in mice we observed no significant differences in osteoblast-covered surface (OB.S/BS, p=0.50) or number (OB.N/B.Pm, p=0.63). Surprisingly, the osteoclast-covered surface (OC.S/BS, p=0.043) and number (N.OC/B.Pm, p=0.0012) were both significantly increased, compared to Zip14fl/- controls (Figure 8D). Double calcein labelling allowed us to investigate the (endosteal and periosteal) cortical and trabecular mineralizing surface (MS/BS), bone formation rate (BFR/BS) and mineral apposition rate (MAR) by fluorescence microscopy. Compared to Zip14fl/- controls, osteoblast-specific knock-in mice had an increase in endosteal MS/BS (p=0.012) and even more in BFR/BS (p=0.0012) (Figure 9A), whereas there were no significant differences in periosteal (data not shown) or trabecular bone formation parameters (Figure 9B).

148

Results – Chapter 6

Figure 10 | Biochemical analysis of bone turnover markers, RANKL and OPG. (A) Measurement of the bone turnover markers procollagen type-I C-terminal peptide (PICP; bone formation) and collagen type 1 cross-linked C-telopeptide (Crosslaps; bone resorption) in serum of 6-month old control (Zip14fl/-), osteoblast-specific knock-in (Zip14fl/-; Runx2-Cre) and osteoclast-specific knock-in (Zip14fl/-; CtsK-Cre) mice. (B) Quantification of the serum RANKL and OPG levels and the calculated RANKL/OPG ratio. N=6 animals/genotype.

Serum was collected prior to euthanasia of the animals for measurement of procollagen I C- terminal propeptide (PICP) and C-terminal telopeptide (CTX Crosslaps) as serum markers for bone formation and resorption, respectively. Osteoblast knock-in mice had similar levels of PICP and CTX, compared to Zip14fl/- mice (Figure 10A). Serum levels of OPG and RANKL were both slightly higher (not significant) in these mice, resulting in a similar RANKL/OPG ratio as controls (Figure 10B).

9. Osteoclast expression of Zip14L438R has little effect on bone homeostasis in vivo

µCT analysis of osteoclast-specific Zip14L438R knock-in mice demonstrated a significantly decreased cortical porosity (p=0.016) compared to Zip14fl/- controls, whereas trabecular bone was unaffected (Figure 7). Histological analysis of undecalcified Von Kossa/Van Gieson stained spine and tibia sections confirmed trabecular bone mass to be unaffected in these mice (Figure 8C). Three-point bending tests indicated that biomechanical properties of the femora of these mice were similar to that of Zip14fl/- controls (Figure 8B). Furthermore, toluidine blue stained sections of the tibiae showed a significant decrease in osteoclast-covered bone surface (p=0.024), whereas osteoclast number (p=0.22) and osteoblast-covered surface (p=0.22) and number (p=0.50) were unaltered (Figure 8D). Regarding dynamic histomorphometry, osteoclast knock-in mice

149

Results – Chapter 6

Figure 11 | The effect of wildtype and mutant ZIP14 on cAMP-CREB, NF-kB and NFAT signaling in HEK293T cells. Luciferase reporter assays investigating cAMP-CREB activy (upper graph), NF-κB activity (middle graph) and NFAT activity (bottom graph) by wildtype (WT), L441R and truncated (W22X) ZIP14 in HEK293T cells demonstrate a significant increase in cAMP-CREB and NFAT signaling by L441R ZIP14, compared to WT ZIP14. *: p<0.05; **: p<0.01. presented with a significant increase in endosteal mineralizing surface (p=0.039), whereas trabecular MS/BS (p=0.0086) and BFR/BS (p=0.020) were decreased (Figure 9). Finally, serum PICP levels of osteoclast knock-in mice were slightly increased, but did not reach significance, whereas CTX was at the same level as controls (Figure 10A). The RANKL/OPG ratio of osteoclast knock-in mice was somewhat lower, due to a slight decrease in RANKL and increase in OPG. Again, this did not reach significance (Figure 10B).

10. ZIP14L441R increases cAMP-CREB and NFAT signaling

Zip14 was previously linked to cAMP-CREB signaling35. To evaluate the effect of WT and L441R ZIP14 on the cAMP-CREB signaling activity, a luciferase reporter assay with a cAMP- responsive luciferase construct was applied. Here, overexpression of WT ZIP14 in HEK293T caused a decrease in cAMP-CREB signaling, whereas overexpression of L441R ZIP14 resulted in a significant (p=0.004) 5-fold increase in activity (Figure 11). Next to cAMP-CREB signaling,

150

Results – Chapter 6

ZIP14 is often associated with immune response and inflammation in the literature. We therefore checked both NF-κB and NFAT signaling activity, due to their importance in bone cells and their link with inflammation. No significant difference in NF-κB signaling was observed between WT and L441R ZIP14, but the NFAT signaling activity of L441R ZIP14, on the other hand, was significantly increased (p=0.031) compared to WT ZIP14 in HEK293T cells (Figure 11). All luciferase reporter assays were also performed in Saos-2 cells, i.e. osteoblast-like cells, with similar results (data not shown).

IV. Discussion Hyperostosis Cranialis Interna (HCI) was first described in a Dutch family as a bone disorder that solely affects the calvaria and skull base through intracranial hyperostosis and osteosclerosis4, 7. Until now, this family is still the only family known with HCI. We performed a whole genome linkage scan in the past, identifying the region linked to HCI on chromosome 8 (8p21)3. In this study we additionally performed whole-exome sequencing (WES) on one HCI patient that led to the identification of a heterozygous c.1322T>G (p.L441R) substitution in the SLC39A14 gene that co-segregates with the disorder. SLC39A14 encodes for a Zn transporter that belongs to the SLC39A or Zrt-, Irt-related protein (ZIP) family and is therefore often referred to as ZIP14. Next to the ZIP family, the SLC30A or Zn transporter (ZnT) family is also implicated with Zn transport. ZIP transporters invariably function by replenishing cytosolic Zn from the extracellular space and the lumen of intracellular compartments (influx), whereas ZnT transporters lower cytosolic Zn levels by mediating Zn efflux from cells or influx into intracellular organelles36. Zn generally plays a vital role in cells as it is estimated that about 10% of the human genome encodes proteins with Zn-binding sites. More than half of those are thought to be transcription factors and enzymes, distributed across the different cellular compartments. Local alterations in Zn homeostasis can therefore have significant effects on the functionality of corresponding Zn-dependent proteins36, 37. ZIP14 has previously been localized to the plasma membrane and in the cytosol, i.e. in early and late endosomes14, 30-33. From here, ZIP14 mainly mobilizes Zn, but transport of other divalent cations (iron, manganese, cadmium) into the cytosol is also described38, 39. We demonstrate that p.L441R ZIP14 is still localized in the cytosol, but loses its presence on the plasma membrane, implying trafficking defects of p.L441R ZIP14 in vitro. Accordingly, p.L441R ZIP14 was not able to transport 65Zn from the extracellular space into the cell, due to its disturbed subcellular localization. Accumulation of labile Zn in the cell, however, was increased in p.L441R ZIP14 expressing cells, indicating an aberrant cellular Zn homeostasis. It is essential to note that the location of labile Zn excess in the cell is unknown and

151

Results – Chapter 6 depends on the localization and transport capacity of p.L441R ZIP14 in the cell. After protein synthesis in the endoplasmic reticulum (ER), membrane proteins (like ZIP14) mostly traffic through the Golgi network up to early endosomes. From here, ZIP14 would then first traffic to the plasma membrane, undergo endocytosis into early endosomes before being sorted to either late endosomes and/or the trans-Golgi network. Since trafficking towards the plasma membrane is affected by the mutation, it is possible that p.L441R ZIP14 is retained in the early endosomes. Of note, patients with HCI have a heterozygous p.L441R substitution, indicating that fifty percent of their ZIP14 is wildtype and reaches the plasma membrane (and early/late endosomes), whereas the other fifty percent will reasonably be trapped onto early endosomes. As seen in the Zn accumulation studies, this unbalanced distribution of ZIP14 results in aberrations of cellular Zn homeostasis with putative harmful effects on the functionality of cells. Similarly, mutations in SLC39A4 (ZIP4) and SLC39A13 (ZIP13) have already been linked to defects in Zn homeostasis resulting in acrodermatitis enteropathica and spondylocheiro dysplastic Ehlers-Danlos syndrome , respectively40-43. Alterations in the localization and functionality of Zn transporters thus results in Zn deficiency and/or accumulation in specific cellular compartments, which is thought to be the primary cause of the pathogenesis of these disorders37. In general, ZIP14 is most abundantly expressed in liver, pancreas, thyroid gland, intestine and heart, and shows low expression in the brain28. Until now, no information was available on expression or function of ZIP14 in bone cells. In this study we demonstrate positive expression of ZIP14 in osteoclast and osteoblasts, whereas osteocytes were negative for ZIP14 expression. Important to note is that mutations in SLC39A14 can affect manganese (Mn), cadmium and iron homeostasis in these tissues and cells as well. Recently, homozygous missense (F98V, G383R, N469R), nonsense (E105X) and frameshift (S160Cfs*5) mutations in SLC39A14 were identified in patients with childhood-onset parkinsonism-dystonia, due to defects in Mn homeostasis33. These mutations were all part of transmembrane domains that are not predicted to form a pore (according to MemSatSVM), whereas our L441R mutation is part of this pore. Subcellular localization patterns for all ZIP14 mutants in this study were also similar to that of wildtype ZIP14, Mn uptake was reduced and specifically accumulated in the brain of a mutant zebrafish model33. For our study, we focused on Zn as it is more relevant in light of skeletal homeostasis36, 37, 44. Zn is described to have a stimulatory role on osteoblastic bone formation and mineralization and an inhibitory effect on osteoclastic bone resorption44, 45. Skeletal cell-specific effects of the aberrations in Zn homeostasis by p.L441R ZIP14 were therefore investigated in vivo in mice with a floxed Zip14L438R and a Runx2-Cre or CtsK-Cre allele for osteoblast-specific and osteoclast- specific expression of Zip14L438R, respectively. Six-month old mice were analyzed, since this

152

Results – Chapter 6 matches the age of about 30 years in humans at which the HCI phenotype is prominently present in patients4, 7, 22. First, body weight and femoral length (growth) were similar for all genotypes. This is relevant since Zn deficiency is associated with growth retardation (and other symptoms) in humans and mice36, 37. Similarly, Zip14-/- mice exhibit a Zn deficient phenotype marked by growth retardation and dwarfism, due to abnormal chondrocyte differentiation in the growth plate35. The normal growth in our mice can therefore be explained by the fact that the role of Zip14 in growth was attributed to its effects on the hypertrophy of chondrocytes, whereas we conditionally expressed Zip14L438R in osteoblasts or osteoclasts. Moreover, skeletal growth or height is not affected in patients with HCI as well. Since patients with HCI carry a heterozygous p.L441R mutation and Zip14+/- mice are phenotypically normal35, it could be that the wildtype allele fulfills a compensatory role during skeletal growth of HCI patients and that growth defects in Zip14-/- mice are probably due to a general state of Zn deficiency mediated by the chondrocytes. Moreover, it was documented that ZIP14 has roles in adipose tissue and glucose utilization that can influence growth and development as well17, 32. When examining the skulls of our conditional Zip14L438R knock-in mice, we were surprised to see no calvarial phenotype, especially knowing that the long bones were affected in both conditional knock-in models. This is truly the opposite of what we see in HCI patients. One aspect to be discussed here is the difference in expression of human ZIP14L441R and murine Zip14L438R. In patients with HCI, endogenous Zip14 is expressed in its own spatiotemporal manner, whereas in mice with conditional overexpression of Zip14L438R this is driven by the Runx2 and Cathepsin K promoter. Nevertheless, Cre expression was reported in long bones and calvariae of both Cre-models used in this study20, 21. Still, we analyzed calvariae of 5-9 months old Zip14-/- mice and age-matched Zip14+/+ controls, also with no significant differences between both groups. Loss of endogenous Zip14 in mice therefore does not result in a calvarial phenotype, only in an osteoporotic phenotype in the rest of the skeleton. This suggests that aberrations in Zn homeostasis by Zip14 do not seem to affect the skull of mice, despite the fact that they do affect the rest of the skeleton. Whether this is due to a specific protective mechanism present in murine calvaria but not in humans, remains to be determined. In contrast to the calvaria, the appendicular skeleton and vertebral column were affected by knock-in of Zip14L438R in osteoblasts and osteoclasts. Generally, knock-in of Zip14L438R in osteoblasts resulted in a prominent skeletal phenotype, whereas the skeletal phenotype in osteoclast knock-in mice was milder. A remarkable finding is that both conditional knock-in models had an increased (endo)cortical bone formation rate. Additionally, osteoblast knock-in mice had an increased cortical thickness, where excessive endosteal bone formation even led to

153

Results – Chapter 6 narrowing of the bone marrow cavity and midshaft diameter. Similarly, a study investigating the metabolic activity in the calvariae of HCI patients with 18F-Fluoride PET/CT depicted the highest rates of 18F-Fluoride uptake in the hyperostotic regions of the skull and more specifically at the endosteal side of the diploe (towards the bone marrow)46. Hyperostosis of the inner calvarial cortex of patients with HCI is thus also the result of an increase in endosteal bone formation. Therefore, even though the location of the skeletal defect is different, i.e. in the appendicular skeleton and vertebral column versus the calvaria, the (endo)cortical phenotype and the underlying cause of this appear strikingly similar in Zip14L438R knock-in mice and HCI patients. In order to further elucidate the in vivo effects of Zip14L438R through osteoblasts, we focused on the fact that Zip14L438R has disparate effects on the cortical and trabecular bone compartment in osteoblast-specific knock-in mice. In short, osteoblast expression of Zip14L438R results in an increased cortical bone volume along with a decreased trabecular bone volume. According to the literature, only few hormones are linked to similar skeletal actions and these are parathyroid hormone (PTH)/parathyroid-related protein (PTHrP) and insulin-like growth factor (IGF-1) and estrogen. Of note, Zip14 was previously associated with PTH1R-cAMP-CREB signaling and the GH/IGF-1 axis in Zip14-/- mice35. Next to growth retardation and dwarfism, these knockout mice had trabecular osteopenia, but an unaltered cortical bone volume34, 35. First, when we look into the skeletal phenotype of Pth-/- mice and mice with osteoblast/osteocyte- specific Gsα deficiency (BGsKO), bearing in mind that PTH mediates its effects through Gsα signaling, we see an increased cortical bone mass, decreased bone marrow cavity and a decreased trabecular bone mass in both models47, 48. Even though the phenotype in these mice is more severe, it has the same differential effects on bone as seen in our osteoblast-specific Zip14L438R knock-in mice. A contrasting skeletal phenotype is also seen in mice with PTH/PTHrP receptor overexpression in the osteoblastic lineage49. This suggests that the skeletal phenotype of osteoblast-specific Zip14L438R knock-in mice resembles that of deficient or restrained PTH- signaling in osteoblasts. Next, results of in vivo models of osteoblast-specific overexpression/knock-out of Igf-1 (or its receptor) suggest that IGF-1 can have anabolic effects only on the trabecular or cortical bone compartment or on both, depending for example on the promoter used to drive overexpression of Igf-1 and the genetic background of the mice50. Generally, the GH/IGF-1 axis regulates longitudinal bone growth, skeletal maturation and bone mass acquisition. In osteoblasts, IGF-1 has anabolic effects, mainly by increasing their bone formation efficiency instead of number, but also promotes formation and function of osteoclasts through production of TNF-α and IL-650-54.

154

Results – Chapter 6

Of note, IGF-1 is only associated with anabolic, i.e. not catabolic, effects on bone. In our osteoblast-specific knock-in mice, Zip14L438R exerts obvious anabolic effects in the cortical bone compartment without increasing osteoblast numbers, suggesting that these osteoblasts have a higher bone formation efficiency. Moreover, osteoclast number was higher even though the RANKL/OPG ratio was unaffected in these mice, indicating that an increase in cytokine production by the osteoblasts could be an explanation of the increase in osteoclast number. So it seems that Zip14L438R in osteoblasts exerts IGF-1-mimicking effects with an anabolic effect on cortical bone. Despite the fact that estrogen was not previously associated with Zip14, it is involved in the actions of IGF-1 and exerts opposing actions on bone compared to PTH in osteoblasts55. Moreover, studies show that Zn has actions similar to estrogen on osteoblasts and osteoclasts45. Estrogen is generally known to restrain periosteal and stimulate endosteal bone formation during skeletal growth and remodeling through osteoblast progenitors56, 57. Consequently, postmenopausal sex-steroid deficiency has been associated with an enlargement of the marrow cavity, thinning of the cortex and slight increase in midshaft diameter58. Osteoblast-specific Zip14L438R knock-in mice, on the contrary, have a significantly smaller midshaft diameter, due to a restricted periosteal bone formation, along with an increase in cortical thickness and narrowed bone marrow cavity, caused by a stimulated endosteal bone formation. Moreover, estrogen has protective effects on the resorption of both trabecular and cortical bone, but these are exerted by disparate cell types, i.e. by direct effects on osteoclasts and indirect effects on osteoblasts, respectively56. As mentioned previously, we indeed see estrogen-mimicking effects on the cortical compartment of osteoblast specific knock-in mice, whereas we see catabolic effects in the trabecular compartment. A possible explanation for this is that by sole osteoblastic expression of Zip14L438R, there is no protective (estrogen-mimicking) effect on the resorption of trabecular bone. Altogether, we hypothesize that Zip14L438R takes part in the osteoblast-specific actions of IGF-1 and estrogen on cortical bone and opposes actions of PTH in cortical and trabecular bone compartment. The skeletal phenotype of osteoclast-specific Zip14L438R knock-in mice was milder. Trabecular bone volume was unaffected in these mice, due to a decrease in osteoclast-covered surface (not number) and trabecular bone formation rate, indicating a lowered bone turnover in this compartment. Here, estrogen-mimicking effects through osteoclasts indeed decrease osteoclast resorption in the trabecular compartment, with subsequently lower bone formation as well. It is clear that osteoclasts do not harbor PTH signaling activity by lack of associated receptors and there is no clear consensus on the direct effect of IGF-1 on osteoclastic bone

155

Results – Chapter 6 resorption50, 52, 53. Whether the mild skeletal phenotype of osteoclast-specific Zip14L438R knock-in mice is fully due to a primary defect in osteoclastic bone resorption or indirect stimulation of osteoblasts, therefore remains unclear. Another important indication for a role of estrogen-like signaling by Zip14L438R or ZIP14L441R was found in clinical reports on the disease progression of affected individuals. Female patients exhibit sudden aggravation of HCI symptoms during pregnancy, like abrupt loss of smell or hearing, of which they sometimes recovered after pregnancy. Often, female patients are more severely, albeit not significant, affected by the disorder7. As mentioned in the introduction, radiological abnormalities associated with HCI are often seen in the first decade of life and a slow progression of the disease can be seen until the age of 407, 10. Altogether, these stages in life share critical changes in estrogen levels, i.e. estrogen gain associated with puberty and pregnancy and estrogen loss associated with aging-related sex-steroid deficiency. We therefore hypothesize that an increased estrogen production is comparable to the estrogen-mimicking effects of Zip14L438R, resulting in aggravation of symptoms in (female) HCI patients. Finally, as based on our findings, it seems that the skeletal effects by Zip14L438R are primarily orchestrated by the osteoblasts, we aimed at identifying possible downstream mechanisms or second messengers through which ZIP14 mediates these effects in osteoblasts. As previously mentioned, Zip14 was shown to play an important role in G-protein coupled receptor (GPCR)-mediated signaling by importing Zn into the cytosol and maintaining basal cAMP levels35. We detected a 5-fold increase in cAMP levels in HEK293T and Saos-2 cells transfected with ZIP14L441R. Cyclic AMP is a well-known second messenger for several hormones, like PTH/PTHrP35, 48, 49. However, Zip14L438R expression in osteoblasts did not result in a PTH-mimicking skeletal phenotype in vivo, not to say that it led to a PTH-contrasting phenotype. In the literature, the G-protein-coupled estrogen receptor (GPER) is documented to act predominantly intracellularly and stimulate cAMP production, calcium mobilization and c-Src. GPER is described to play a role in the reproductive system, nervous system and neuroendocrinology, immune system, cardiovascular system, pancreatic function and glucose metabolism and bone growth and chondrocyte metabolism59. Remarkably ZIP14 was documented to be expressed and/or play a role in these metabolic systems as well. This is easily illustrated by the Zip14-/- mice that are characterized by impaired gluconeogenesis, hyperinsulinemic/diabetic pancreatic islets, chronic inflammation state, osteopenia and growth retardation34, 35. We therefore hypothesize that Zip14L438R, as it does not traffic towards the plasma membrane, increases intracellular Zn levels and GPER signaling and cAMP production from the intracellular compartment where it resides, with anabolic effects on osteoblasts. Next, since

156

Results – Chapter 6

Zip14 ablation yields a proinflammatory phenotype with increased systemic IL-6 levels that are coincident with a decrease in BMD in vivo34 and osteoclast numbers were increased in mice with osteoblastic expression of Zip14L438R (with normal RANKL/OPG ratio), we investigated differences in NF-κB and NFAT activity by ZIP14L441R. NF-κB activity was unaltered by ZIP14L441R, whereas we demonstrated a nearly 2-fold increase in NFAT signaling activity in HEK293T and Saos-2 cells. NFAT signaling in osteoblasts has been linked to the production of chemoattractants (TNF-α, IL-6) by osteoblasts to attract osteoclast progenitors and hence increase osteoclast numbers. Moreover, GPER activation is also linked to increased intracellular calcium mobilization, which is known to bind and activate calmodulin and calcineurin, activators of NFAT59. In conclusion, based on our in vitro and in vivo findings, we are convinced that we identified the disease causing gene for Hyperostosis Cranialis Interna. Compared to ZIP14, ZIP14L441R increases endosteal bone formation through estrogen-mimicking effects, which is the result of changes in its subcellular localization, intracellular Zn accumulation and the increase in cAMP levels and NFAT activity. Moreover, its effect on bone homeostasis is predominantly driven by the osteoblasts, but carried out together with the osteoclasts. Regarding the fact that only the skull of patients with HCI is affected, we hypothesize that this is due to differences in micro-environment. ZIP14, along with numerous other ZIPs, ZnTs, Zn-dependent transcription factors and enzymes, define a local and spatiotemporal micro-environment and, for some reason, only that one in the calvariae of HCI patients results in severe bone overgrowth by the heterozygous p.L441R mutation in ZIP14. Whether this is due to a specific deficit in skeletal cells of the calvariae or fortunate differences in the expression pattern of compensatory mechanisms in the rest of the skeleton, remains to be determined.

157

Results – Chapter 6

V. Acknowledgements This work was supported by the Systems Biology for the Functional Validation of Genetic Determinants of Skeletal Diseases (SYBIL) project funded by the European Commission and by a grant from the FWO (Fund for Scientific Research) Flanders (FWO grant G0197.12N), all to W. Van Hul. G. Hendrickx holds a Rosa Blanckaert Legacy Grant for Young Investigators from the University of Antwerp (Project ID 32435). V.B. held a pre-doctoral fellowship of the Agency for Innovation by Science and Technology (IWT) Flanders and E.B. holds a post-doctoral fellowship of the FWO Flanders (FWO personal grant 12A3814N). Experiments with Zip14-/- mice and zinc transport studies were supported by the National Institutes of Health grant DK 94244 to R.J.C. The authors want to thank Pancras C.W. Hogendoorn for his constructive comments on the histological findings described in this manuscript and Natalja E. Leeuwis- Fedorovich and Dilara C. Savci-Heijink for their work in obtaining the patient skull biopsy specimen.

158

Results – Chapter 6

References 1. Appelman-Dijkstra, N.M. & Papapoulos, S.E. From disease to treatment: from rare skeletal disorders to treatments for osteoporosis. Endocrine 52, 414-26 (2016). 2. Fijalkowski, I., Boudin, E., Mortier, G. & Van Hul, W. Sclerosing bone dysplasias: leads toward novel osteoporosis treatments. Curr Osteoporos Rep 12, 243-51 (2014). 3. Borra, V.M., Waterval, J.J., Stokroos, R.J., Manni, J.J. & Van Hul, W. Localization of the gene for hyperostosis cranialis interna to chromosome 8p21 with analysis of three candidate genes. Calcif Tissue Int 93, 93-100 (2013). 4. Manni, J.J., Scaf, J.J., Huygen, P.L., Cruysberg, J.R. & Verhagen, W.I. Hyperostosis cranialis interna. A new hereditary syndrome with cranial-nerve entrapment. N Engl J Med 322, 450-4 (1990). 5. Pagon, R.A., Beckwith, J.B. & Ward, B.H. Calvarial hyperostosis: a benign X-linked recessive disorder. Clin Genet 29, 73-8 (1986). 6. Waterval, J.J., Borra, V.M., Van Hul, W., Stokroos, R.J. & Manni, J.J. Sclerosing bone dysplasias with involvement of the craniofacial skeleton. Bone 60, 48-67 (2014). 7. Waterval, J.J., Stokroos, R.J., Bauer, N.J., De Bondt, R.B. & Manni, J.J. Phenotypic manifestations and management of hyperostosis cranialis interna, a hereditary bone dysplasia affecting the calvaria and the skull base. Am J Med Genet A 152A, 547-55 (2010). 8. Manni, J.J., Huygen, P.L., Noten, J.F. & Kuijpers, W. Eighth cranial nerve dysfunction in hyperostosis cranialis interna. Acta Otolaryngol 112, 75-82 (1992). 9. Waterval, J.J., Stokroos, R.J., De Bondt, R.B. & Manni, J.J. Facial nerve decompression via middle fossa approach for hyperostosis cranialis interna: a feasible therapeutic approach. J Laryngol Otol 123, 1177-80 (2009). 10. Waterval, J.J. et al. Imaging features and progression of hyperostosis cranialis interna. AJNR Am J Neuroradiol 33, 453-61 (2012). 11. Waterval, J.J., Stokroos, R.J., Dings, J., Van Overbeeke, J.J. & Manni, J.J. Cerebral vasospasm after auditory brainstem implantation in a patient with hyperostosis cranialis interna. Clin Neurol Neurosurg 113, 904-8 (2011). 12. Emmetsberger, J. et al. Tissue plasminogen activator alters intracellular sequestration of zinc through interaction with the transporter ZIP4. J Neurosci 30, 6538-47 (2010). 13. Liuzzi, J.P., Aydemir, F., Nam, H., Knutson, M.D. & Cousins, R.J. Zip14 (Slc39a14) mediates non-transferrin-bound iron uptake into cells. Proc Natl Acad Sci U S A 103, 13612-7 (2006).

159

Results – Chapter 6

14. Liuzzi, J.P. et al. Interleukin-6 regulates the zinc transporter Zip14 in liver and contributes to the hypozincemia of the acute-phase response. Proc Natl Acad Sci U S A 102, 6843-8 (2005). 15. Jansen, I.D. et al. Ae2(a,b)-deficient mice exhibit osteopetrosis of long bones but not of calvaria. FASEB J 23, 3470-81 (2009). 16. Aydemir, T.B. et al. Zinc transporter ZIP14 functions in hepatic zinc, iron and glucose homeostasis during the innate immune response (endotoxemia). PLoS One 7, e48679 (2012). 17. Aydemir, T.B., Sitren, H.S. & Cousins, R.J. The zinc transporter Zip14 influences c-Met phosphorylation and hepatocyte proliferation during liver regeneration in mice. Gastroenterology 142, 1536-46 e5 (2012). 18. Dempster, D.W. et al. Standardized nomenclature, symbols, and units for bone histomorphometry: a 2012 update of the report of the ASBMR Histomorphometry Nomenclature Committee. J Bone Miner Res 28, 2-17 (2013). 19. Lakso, M. et al. Targeted oncogene activation by site-specific recombination in transgenic mice. Proc Natl Acad Sci U S A 89, 6232-6 (1992). 20. Rauch, A. et al. Glucocorticoids suppress bone formation by attenuating osteoblast differentiation via the monomeric glucocorticoid receptor. Cell Metab 11, 517-31 (2010). 21. Chiu, W.S. et al. Transgenic mice that express Cre recombinase in osteoclasts. Genesis 39, 178-85 (2004). 22. Flurkey, K. et al. PohnB6F1: a cross of wild and domestic mice that is a new model of extended female reproductive life span. J Gerontol A Biol Sci Med Sci 62, 1187-98 (2007). 23. Hoff, A.O. et al. Increased bone mass is an unexpected phenotype associated with deletion of the calcitonin gene. J Clin Invest 110, 1849-57 (2002). 24. Koehne, T. et al. Trends in trabecular architecture and bone mineral density distribution in 152 individuals aged 30-90 years. Bone 66, 31-8 (2014). 25. Roschger, P., Paschalis, E.P., Fratzl, P. & Klaushofer, K. Bone mineralization density distribution in health and disease. Bone 42, 456-66 (2008). 26. Schinke, T. et al. Impaired gastric acidification negatively affects calcium homeostasis and bone mass. Nat Med 15, 674-81 (2009). 27. Kircher, M. et al. A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet 46, 310-5 (2014).

160

Results – Chapter 6

28. Taylor, K.M., Morgan, H.E., Johnson, A. & Nicholson, R.I. Structure-function analysis of a novel member of the LIV-1 subfamily of zinc transporters, ZIP14. FEBS Lett 579, 427- 32 (2005). 29. UniProt, C. UniProt: a hub for protein information. Nucleic Acids Res 43, D204-12 (2015). 30. Aydemir, T.B., Troche, C., Kim, M.H. & Cousins, R.J. Hepatic ZIP14-mediated Zinc Transport Contributes to Endosomal Insulin Receptor Trafficking and Glucose Metabolism. J Biol Chem 291, 23939-23951 (2016). 31. Guthrie, G.J. et al. Influence of ZIP14 (slc39A14) on intestinal zinc processing and barrier function. Am J Physiol Gastrointest Liver Physiol 308, G171-8 (2015). 32. Troche, C., Aydemir, T.B. & Cousins, R.J. Zinc transporter Slc39a14 regulates inflammatory signaling associated with hypertrophic adiposity. Am J Physiol Endocrinol Metab 310, E258-68 (2016). 33. Tuschl, K. et al. Mutations in SLC39A14 disrupt manganese homeostasis and cause childhood-onset parkinsonism-dystonia. Nat Commun 7, 11601 (2016). 34. Aydemir, T.B. et al. Aging amplifies multiple phenotypic defects in mice with zinc transporter Zip14 (Slc39a14) deletion. Exp Gerontol 85, 88-94 (2016). 35. Hojyo, S. et al. The zinc transporter SLC39A14/ZIP14 controls G-protein coupled receptor-mediated signaling required for systemic growth. PLoS One 6, e18059 (2011). 36. Kambe, T., Tsuji, T., Hashimoto, A. & Itsumura, N. The Physiological, Biochemical, and Molecular Roles of Zinc Transporters in Zinc Homeostasis and Metabolism. Physiol Rev 95, 749-84 (2015). 37. Kambe, T., Hashimoto, A. & Fujimoto, S. Current understanding of ZIP and ZnT zinc transporters in human health and diseases. Cell Mol Life Sci 71, 3281-95 (2014). 38. Fujishiro, H., Yano, Y., Takada, Y., Tanihara, M. & Himeno, S. Roles of ZIP8, ZIP14, and DMT1 in transport of cadmium and manganese in mouse kidney proximal tubule cells. Metallomics 4, 700-8 (2012). 39. Pinilla-Tenas, J.J. et al. Zip14 is a complex broad-scope metal-ion transporter whose functional properties support roles in the cellular uptake of zinc and nontransferrin- bound iron. Am J Physiol Cell Physiol 301, C862-71 (2011). 40. Giunta, C. et al. Spondylocheiro dysplastic form of the Ehlers-Danlos syndrome--an autosomal-recessive entity caused by mutations in the zinc transporter gene SLC39A13. Am J Hum Genet 82, 1290-305 (2008).

161

Results – Chapter 6

41. Jeong, J. et al. Promotion of vesicular zinc efflux by ZIP13 and its implications for spondylocheiro dysplastic Ehlers-Danlos syndrome. Proc Natl Acad Sci U S A 109, E3530- 8 (2012). 42. Kury, S. et al. Identification of SLC39A4, a gene involved in acrodermatitis enteropathica. Nat Genet 31, 239-40 (2002). 43. Wang, K., Zhou, B., Kuo, Y.M., Zemansky, J. & Gitschier, J. A novel member of a zinc transporter family is defective in acrodermatitis enteropathica. Am J Hum Genet 71, 66-73 (2002). 44. Fukada, T., Hojyo, S. & Furuichi, T. Zinc signal: a new player in osteobiology. J Bone Miner Metab 31, 129-35 (2013). 45. Karieb, S. & Fox, S.W. Zinc modifies the effect of phyto-oestrogens on osteoblast and osteoclast differentiation in vitro. Br J Nutr 108, 1736-45 (2012). 46. Waterval, J.J. et al. Bone metabolic activity in hyperostosis cranialis interna measured with 18F-fluoride PET. Eur J Nucl Med Mol Imaging 38, 884-93 (2011). 47. Miao, D., He, B., Karaplis, A.C. & Goltzman, D. Parathyroid hormone is essential for normal fetal bone formation. J Clin Invest 109, 1173-82 (2002). 48. Sakamoto, A. et al. Deficiency of the G-protein alpha-subunit G(s)alpha in osteoblasts leads to differential effects on trabecular and cortical bone. J Biol Chem 280, 21369-75 (2005). 49. Calvi, L.M. et al. Activated parathyroid hormone/parathyroid hormone-related protein receptor in osteoblastic cells differentially affects cortical and trabecular bone. J Clin Invest 107, 277-86 (2001). 50. Tahimic, C.G., Wang, Y. & Bikle, D.D. Anabolic effects of IGF-1 signaling on the skeleton. Front Endocrinol (Lausanne) 4, 6 (2013). 51. Jiang, J. et al. Transgenic mice with osteoblast-targeted insulin-like growth factor-I show increased bone remodeling. Bone 39, 494-504 (2006). 52. Locatelli, V. & Bianchi, V.E. Effect of GH/IGF-1 on Bone Metabolism and Osteoporsosis. Int J Endocrinol 2014, 235060 (2014). 53. Ueland, T. GH/IGF-I and bone resorption in vivo and in vitro. Eur J Endocrinol 152, 327- 32 (2005). 54. Zhang, M. et al. Osteoblast-specific knockout of the insulin-like growth factor (IGF) receptor gene reveals an essential role of IGF signaling in bone matrix mineralization. J Biol Chem 277, 44005-12 (2002).

162

Results – Chapter 6

55. Fritton, J.C. et al. Growth hormone protects against ovariectomy-induced bone loss in states of low circulating insulin-like growth factor (IGF-1). J Bone Miner Res 25, 235-46 (2010). 56. Manolagas, S.C., O'Brien, C.A. & Almeida, M. The role of estrogen and androgen receptors in bone health and disease. Nat Rev Endocrinol 9, 699-712 (2013). 57. Seeman, E. Periosteal bone formation--a neglected determinant of bone strength. N Engl J Med 349, 320-3 (2003). 58. Hendrickx, G., Boudin, E. & Van Hul, W. A look behind the scenes: the risk and pathogenesis of primary osteoporosis. Nat Rev Rheumatol 11, 462-74 (2015). 59. Prossnitz, E.R. & Barton, M. The G-protein-coupled estrogen receptor GPER in health and disease. Nat Rev Endocrinol 7, 715-26 (2011).

163

Results – Chapter 6

164

GENERAL DISCUSSION

166

General Discussion

Bone homeostasis is a process characterized by the constant and lifelong remodeling of bone tissue, mediated by skeletal cell types to sustain bone mineral density (BMD) and quality. BMD is a multifactorial skeletal parameter of which genetic factors are estimated to define up to 80% of its variance. Performing genetic research is therefore a reasonable approach to increase knowledge on BMD and bone homeostasis in general. Moreover, studying BMD by means of genetic skeletal disorders has already been a productive strategy, by associating disease-causing genes with the respective pathological mechanism. In this thesis, two different bone disorders were initially investigated by genetic research and, if applicable, followed up by in vitro and in vivo functional research, resulting in a better understanding of bone homeostasis. Nevertheless, bone homeostasis and the regulation of BMD remain complex and there is still a lot to discover for the future. In this part of my thesis I will therefore discuss current and future strategies to go from a novel gene to novel insights in bone homeostasis and how we applied different strategies for the work presented in this thesis.

Genetic efforts to explore the heritability of BMD

The past decades, many efforts have been undertaken to improve the identification of genes contributing to common skeletal disorders, like osteoporosis. One of the first strategies was the candidate gene approach, where a gene of interest is screened and assessed for associations with a trait or disorder, by direct Sanger sequencing or tagSNP genotyping. The selection of the candidate gene here is however biased and grounded on current knowledge, limiting researchers in the discovery of novel genes1-4. Unraveling the genetic background of osteoporosis is challenging due to the small effect size of variants involved. Nowadays, most of our knowledge on this matter is therefore gathered from genome-wide association studies (GWAS) as they allow us to assess the contribution of genetic variants to skeletal traits (f.e. BMD, fracture risk) across the whole genome in a hypothesis-free and high-throughput manner1-4. The largest published study in relation to BMD and fracture risk to date is a meta-analysis of GWAS and associated 56 loci with BMD of which 14 loci were also associated with fracture risk. Here, pathway analysis showed that the WNT/β-catenin signaling pathway, RANK-RANKL-OPG pathway and pathways regulating endochondral ossification are the main associated pathways. These loci still only explain <6% of the variance in BMD, leaving up to 74% of the genetic contribution to BMD variability unexplained2,5. Just recently, at the 2016 Annual Meeting of the American Society for Bone and Mineral Research (ASBMR), novel GWAS data were presented and showed associations of 376 loci with BMD, jointly explaining 13% of the variation in BMD6. Compared

167

General Discussion to the meta-analysis by Estrada et al., increasing the sample size by 4-fold (up to 140 000 samples) thus increased the number of associated loci by 6-fold. Four years ago, we selected three WNT genes associated with BMD (and fracture risk) in the study by Estrada et al. for further analysis5. WNT4, WNT5B and WNT16 were screened with Sanger sequencing using the candidate gene approach. This screening was performed in two extreme BMD cohorts, selected from a population of healthy Danish men. For WNT16, we additionally performed tagSNP genotyping to look for associations of general variation in WNT16 with BMD and bone quality parameters in a population of healthy Danish men. This genetic screening was overall successful for WNT16, as we identified additional associations with BMD and hip geometry parameters and identified a common insertion polymorphism (rs55710688) in the Kozak consensus sequence of WNT16 with an effect on its translation efficiency (Figure 1). We finally associated a higher translation efficiency, and thus higher amounts of WNT16, with a higher BMD and in vivo studies with Wnt16-/- mice by others confirmed this hypothesis7-9. Using this same methodology was however unsuccessful for WNT4 and WNT5B, despite their associations with BMD in the same meta-analysis of GWAS as WNT165. We therefore hypothesize that, on a genetic level, the effect size of variants in WNT16 appears greater than variants in WNT4 and WNT5B or that the actual functional variant for the associations of WNT4 and WNT5B lies within neighboring genes. Increasing the sample size of replication cohorts, looking deeper into intronic regions and neighboring genes can therefore possibly explain the association of WNT4 and WNT5B with BMD in future studies.

For rare monogenic disorders, gene identification tools have the goal to identify the one disease- causing variant. One of the earliest strategies, and nowadays still used for large families with a monogenic disorder, is positional cloning by using linkage analysis. Here, a specific chromosomal region encompassing the disease-causing variant is linked to the disorder and direct sequencing of candidate genes can result in the identification of the variant of interest. A major advantage of this strategy is that it is unbiased or hypothesis-free, enabling the identification of novel genes and mechanisms2,4,10. A next area of gene identification tools came with the implementation of next-generation sequencing (NGS) technologies, enabling sequencing of whole exomes (WES) and even whole genomes (WGS) to look for a disease-causing or different disease-contributing variants in a hypothesis-free manner. Due to a large output, this NGS approach requires good variant filtering strategies to discriminate the disease-causing variants from harmless genomic variation. Currently, for smaller families with monogenic skeletal disorders or isolated cases, WES (or WGS) is mainly used as primary strategy to identify the causative gene in a hypothesis-free manner1,2,4,10. Prior to this thesis, the monogenic background of Hyperostosis Cranialis Interna

168

General Discussion

(HCI) had already been investigated by linkage analysis on the complete family, identifying the linkage region at chromosome 8p2111. Screening of putative candidate genes in this region was negative and we therefore performed WES on one HCI patient. Here, previous identification of the linkage region contributed to a very efficient filtering strategy, as only variants in this region were considered for filtering. This genetic approach was successful as it led to the identification of SLC39A14 (ZIP14) as the disease-causing gene for HCI, along with in vitro and in vivo functional studies (Figure 1). In this part of my thesis we therefore demonstrate that when a disease-causing gene remains unfound after screening of putative candidate genes in a linkage region, performing WES on just one patient is a great strategy to find the disease-causing gene in a hypothesis-free manner.

From a genetic point of view, this thesis contributed to novel knowledge on BMD by using the candidate gene approach on GWAS hits for osteoporosis and linkage analysis combined with WES for Hyperostosis Cranialis Interna. Nevertheless, there is still plenty to discover on BMD, as only 13% of its variance is explained according to the recent study by Kemp et al.6. To contest this missing heritability in the future, several actions can be undertaken. As mentioned previously, increasing sample sizes of GWAS unveils more and more BMD loci. NGS techniques are currently also used to screen large populations for the discovery of rare variants with large effect sizes. Moreover, as mentioned in the introduction of this thesis, part of the missing heritability can probably be explained by structural variations (f.e. copy-number variations) or epigenetic mechanisms (f.e. histone modification, transcriptional repression by MicroRNAs and DNA methylation regulating expression) as well2-4.

Functional validation as a link to definite genotype-phenotype correlations

Functional validation remains a crucial step to ascertain genotype-phenotype correlations, not only of putative variants involved in monogenic traits, but also of novel GWAS hits associated with complex traits. As the effect size of SNPs associated with BMD in GWAS is relatively small, functional characterization of novel interesting GWAS hits mainly focuses on exploring the function of the gene to which the associated SNP was allocated. This is for example done by checking the expression in skeletal cell types and performing overexpression or knockdown studies to asses changes in differentiation, function or expression of bone formation/resorption markers. Moreover, next to in vitro studies, a novel GWAS hit is often investigated in vivo by creating a (conditional) knock-out or overexpression model. This is illustrated by the identification of WNT4, WNT5B and WNT16 as novel GWAS hits by Estrada et al.. Shortly

169

General Discussion

Figure 1 | Overview of the objectives, strategies and general conclusions of this PhD thesis.

170

General Discussion hereafter, results of a Wnt16-/- mouse model were published with an obvious skeletal phenotype, validating its role in bone homeostasis8,9. Similarly, transgenic mice overexpressing Wnt4 from osteoblasts demonstrated a protective effect of Wnt4 on bone loss and chronic inflammation12 and transgenic mice overexpressing Wnt5b in cartilage had delayed chondrocyte hypertrophy and bone mineralization, while Wnt5b knockout (Wnt5b-/-) mice had no consistent skeletal phenotype13,14. Altogether, functional validation of these WNTs has been performed, but illustrate completely different mechanisms of action by these WNTs. When comparing the different in vivo models, knockout of Wnt16 has a clear, straightforward effect on BMD, whereas the models of Wnt4 and Wnt5b exhibit indirect mechanisms to modulate their effects on BMD. These functional data offer an additional explanation for the negative genetic screening of WNT4 and WNT5B in our respective cohorts. We screened patients with craniotubular hyperostosis, characterized by increased osteoblastic bone formation, and young, healthy men with an extreme BMD value. Both Wnt4 and Wnt5b do not increase osteoblast function in vivo and our populations are not characterized by a state of chronic inflammation, f.e. by aging or sex-steroid deficiency, which was linked to Wnt4 in mice. Therefore, in addition to the suggestions made on a genetic level, it could be worthwhile to screen populations with specific characteristics. Based on the information coming from in vivo models of GWAS hits, we could for example screen WNT4 in an aged population, as aging is associated with a chronic inflammatory state.

Compared to GWAS hits, functional validation of putative disease-causing variants for monogenic skeletal disorders occurs in a different manner. Instead of looking into the general function of a gene in bone homeostasis, the wildtype versus mutant gene or protein are investigated in parallel. Based on in silico predictions, a first set of in vitro assays can be selected. An in silico predicted disruption of cellular localization can be checked by subcellular localization studies. Protein function can be assessed in vitro by overexpression (transient/stable transfection) or knockdown (transient siRNA-based or stable shRNA or CRISPR/Cas-based) in a cellular model. Here, an appropriate cellular model can be selected through expression studies in cells and/or through immunohistochemistry. Looking into the phenotype of existing in vivo models of your gene of interest or of a gene with a known similar function can also contribute to novel insights on the effect of a mutation in a certain gene in vivo. Finally, studying a particular mutation in an in vivo model, f.e. in mice or zebrafish, is often seen as a final phase of functional validation. Thorough phenotyping of in vivo models can then not only prove a genotype-phenotype correlation, but also contributes to broader knowledge on the function of a gene in bone homeostasis and other non-skeletal mechanisms. This is illustrated by the work performed for Hyperostosis Cranialis Interna in this thesis, where in vitro assays were performed and an

171

General Discussion in vivo model underwent thorough phenotyping. Although the results were not identical to the human HCI phenotype, the ZIP14L441R-HCI correlation was proven successfully (Figure 1). What remains to be determined is the exact mechanism(s) through which solely the skull is affected in these patients. As we believe that this is due to a difference in micro- environment on a cellular level, RNA sequencing and a proteomic analysis could be performed to discover differential expression patterns or differences in the abundance of (zinc-dependent) proteins in skeletal cells of the skull and long bones.

Functional validation is essential to link one genetic variant to a specific disorder or complex trait. Due to a great availability of in vitro assays and efficient ways to create different in vivo models nowadays, there are numerous possibilities to do so. Consequently, as underlying mechanisms are identified by performing functional studies, this will contribute to the general knowledge on BMD, bone disorders and bone homeostasis in general.

Studying different bone disorders unveils different aspects of bone homeostasis

After decades of research towards skeletal disorders encompassing the complete BMD spectrum, diverse aspects of bone homeostasis have been revealed and thoroughly investigated. Osteoporosis is the most prevalent bone disorder and has therefore been studied extensively in order to improve its prevention and treatment3. A drawback in this research is that the low BMD and consequential fractures of osteoporosis patients are the result of a plethora of mechanisms that are not solely based on genetic variations with small effects, but also influences of lifestyle and aging. Nevertheless, research efforts towards the pathogenesis of osteoporosis contributed to the knowledge on the effect of sex steroids, physical activity, dietary factors, etc. on the bone homeostasis process. As these results are of direct relevance for osteoporosis patients, they already contributed to a better prevention and/or treatment, for example by encouraging physical activity at a later age3. In this thesis, osteoporosis was investigated in a more indirect manner, as we chose to further investigate three WNT genes and GWAS hits. Motivated by the genetic and functional results of us and others, WNT16 was selected for a follow-up functional study, to unravel its method of action in osteoblasts. For WNT16, we detected an activating effect on canonical WNT signaling, but additionally found non-canonical activity in osteoblasts, which is probably responsible for its communication with the osteoclasts (Figure 1). The role of non-canonical WNT signaling in bone homeostasis is still more elusive than that of canonical WNT signaling, due to less genetic evidence linking it to bone disorders and a big

172

General Discussion network of intracellular partners. By performing these studies, we point out a cooperating role of non-canonical WNT signaling by WNT16 in its beneficial effects on BMD. Moreover, the mechanisms through which WNTs modulate different parts of WNT signaling remain largely elusive. Our ongoing research towards the orchestrating role of Gα subunits in this mechanism could help increasing knowledge on this matter (Figure 1).

Probably the greatest contribution to general knowledge on bone homeostasis came from the study of rare, monogenic bone disorders. Investigating sclerosing bone disorders, i.e. with an extremely high BMD, led to the identification of novel mechanisms and signaling pathways regulating the function of the different skeletal cell types1,3,10. Logically, an increase in osteoblast function can underlie an increased BMD, as was illustrated by the discovery of the disease- causing genes for Van Buchem disease and sclerosteosis, i.e. SOST, LRP4 and LRP515-19 . Mutations in these genes jointly result in an increase in osteoblast-mediated bone formation and are all known to activate the WNT/β-catenin signaling pathway. Research of these craniotubular hyperostoses therefore unveiled the anabolic potential of the canonical WNT pathway. Another important pathway in osteoblast function is the TGF-β signaling pathway, in which activating genetic defects were found to be causative for Camurati-Engelmann disease, osteopoikilosis and others20,21. Next to improving osteoblastic bone formation, a defect in osteoclastic bone resorption can cause a sclerosing bone disorder. Genetic research of pycnodysostosis and osteopetrosis gained novel insights on this matter22-24. Similarly, other types of osteopetrosis, i.e. osteoclast-poor forms of osteopetrosis, linked defects in osteoclast differentiation to an increased BMD. These types of osteopetrosis highlighted the RANK-RANKL-OPG pathway as a crucial mediator of osteoclastogenesis25,26. Monogenic bone disorders with a low BMD mainly increased knowledge on the biosynthesis of type 1 collagen. This part of the BMD spectrum is dominated by osteogenesis imperfecta, a collective term of heterogeneous connective tissue disorders of which 90% of patients have a mutation in COL1A1 or COL1A227. Moreover, loss-of-function mutations in players of the WNT pathway have been identified as disease-causing for osteogenesis imperfecta (WNT1)28, osteoporosis pseudoglioma syndrome (LRP5)29 and juvenile idiopathic osteoporosis (WNT1, LRP5)28,30,31, indicating a BMD spectrum-wide role for WNT signaling. Hyperostosis Cranialis Interna was investigated in this thesis as a monogenic skeletal disorder characterized by skeletal overgrowth at the skull, whereas the remainder of the skeleton is unaffected. HCI was investigated in a direct manner, as SLC39A14 (ZIP14) was identified as the disease-causing gene for HCI. ZIP14 was not reported to be part of the WNT, TGF-β or RANK-RANKL-OPG pathway, was not known in the field of bone research and was therefore seen as an “orphan” gene. By performing in vitro and in vivo functional studies, we were

173

General Discussion able to unravel the underlying pathological mechanism, as we illustrate estrogen- mimicking and parathyroid hormone (PTH)-opposing effects of p.L441R ZIP14 on bone growth and homeostasis. We therefore believe that, in osteoblasts, ZIP14 is a key component in the axis of growth factors during both bone modeling and remodeling (Figure1).

Altogether, it is clear that performing research towards different skeletal disorders improves the knowledge of different molecular and cellular aspects of bone homeostasis, confirming its relevance. Moreover, even though identifying disease-causing genes for rare skeletal disorders only has direct relevance for this small amount of patients, it is of major indirect relevance for skeletal disorders in general. Increasing the general knowledge on how bone is regulated, broadens the possibilities on how bone disorders generally can be prevented or treated. Therefore, for the future, further research to all kinds of bone disorders is strongly encouraged.

Novel genetic insights in bone homeostasis can result in novel therapeutic approaches

The scientific breakthroughs realized in the past 20 years had clear implications for the prevention and treatment of osteoporosis. The development of several of these therapeutic strategies are based on the study of rare, monogenic skeletal disorders10,32. One way to decrease fracture risk in osteoporosis patients is through inhibition of bone resorption. Consequently, the study of causative genes for osteopetrosis led to the development of an effective monoclonal anti-RANKL antibody (denosumab)33 and small molecules targeting CLCN7 and TCIRG1 are tested for their efficacy, specificity and safety34,35. Similarly, disease-causing mutations in CTSK, encoding cathepsin K, have been identified in patients with pycnodysostosis and was suggested to have therapeutic potential. Subsequently, odanacatib was considered a promising cathepsin K inhibitor, until recent Phase III trials were discontinued due to an increased risk of atrial fibrillation and stroke36. Another approach to treat osteoporosis patients is by increasing bone formation. As mentioned previously, research of Van Buchem disease and sclerosteosis defined the anabolic potential of canonical WNT signaling. Sclerostin, the gene product of SOST, is currently one of the most promising drug targets for osteoporosis. Several pharmaceutical companies are developing sclerostin antibodies, i.e. Romosozumab, Blosozumab and BPS804, and recent studies indicate a lower risk of vertebral and clinical fractures after a 1-year treatment with Romosozumab37. Similar to sclerostin, Dickkopf-related protein 1 (DKK1) is also an

174

General Discussion inhibitor of the canonical WNT pathway and was associated with BMD in the meta-analysis of GWAS by Estrada et al.5. Monoclonal DKK1-neutralizing antibodies are currently studied in clinical trials and are mainly promising for the treatment of multiple myeloma, a cancerous disorder characterized by severe bone pain38.

In conclusion, performing genetic research towards common and rare skeletal disorders greatly improves knowledge on the genetic background of BMD. In this thesis, we present WNT16 and SLC39A14 as two genes with a clear effect on BMD. Still, it remains crucial to functionally validate genes coming from GWAS and from the study of monogenic skeletal disorders in order to confirm the causality of genetic variation and a certain phenotype. For WNT16, we further confirmed genetic association of a variant in its Kozak consensus sequence with BMD by additionally performing in vitro experiments. Definite genotype-phenotype correlation of p.L441R ZIP14 and Hyperostosis Cranialis Interna was achieved by our results coming from in vitro experiments and skeletal phenotyping of conditional mouse models. This led to a broader understanding of the complex and multifaceted process of bone homeostasis and highlighted putative novel targets or mechanisms for prevention or therapeutic intervention of bone disorders. For the future, by performing and improving the genetic and functional possibilities to study skeletal disorders encompassing the complete BMD spectrum, this story will definitely be continued.

175

General Discussion

References

1. Boudin, E., Fijalkowski, I., Hendrickx, G. & Van Hul, W. Genetic control of bone mass. Mol Cell Endocrinol 432, 3-13 (2016). 2. Clark, G.R. & Duncan, E.L. The genetics of osteoporosis. Br Med Bull 113, 73-81 (2015). 3. Hendrickx, G., Boudin, E. & Van Hul, W. A look behind the scenes: the risk and pathogenesis of primary osteoporosis. Nat Rev Rheumatol 11, 462-74 (2015). 4. Rocha-Braz, M.G. & Ferraz-de-Souza, B. Genetics of osteoporosis: searching for candidate genes for bone fragility. Arch Endocrinol Metab 60, 391-401 (2016). 5. Estrada, K. et al. Genome-wide meta-analysis identifies 56 bone mineral density loci and reveals 14 loci associated with risk of fracture. Nat Genet 44, 491-501 (2012). 6. Kemp, J.P. et al. Genome-wide association study of bone mineral density in the UK Biobank Study identifies over 376 loci associated with osteoporosis. American Society of Bone and Mineral Research, Annual Meeting 2016 Abstracts (2016). 7. Hendrickx, G. et al. Variation in the Kozak sequence of WNT16 results in an increased translation and is associated with osteoporosis related parameters. Bone 59, 57-65 (2014). 8. Medina-Gomez, C. et al. Meta-analysis of genome-wide scans for total body BMD in children and adults reveals allelic heterogeneity and age-specific effects at the WNT16 locus. PLoS Genet 8, e1002718 (2012). 9. Zheng, H.F. et al. WNT16 influences bone mineral density, cortical bone thickness, bone strength, and osteoporotic fracture risk. PLoS Genet 8, e1002745 (2012). 10. Fijalkowski, I., Boudin, E., Mortier, G. & Van Hul, W. Sclerosing bone dysplasias: leads toward novel osteoporosis treatments. Curr Osteoporos Rep 12, 243-51 (2014). 11. Borra, V.M., Waterval, J.J., Stokroos, R.J., Manni, J.J. & Van Hul, W. Localization of the gene for hyperostosis cranialis interna to chromosome 8p21 with analysis of three candidate genes. Calcif Tissue Int 93, 93-100 (2013). 12. Yu, B. et al. Wnt4 signaling prevents skeletal aging and inflammation by inhibiting nuclear factor-kappaB. Nat Med 20, 1009-17 (2014). 13. Brommage, R. et al. High-throughput screening of mouse gene knockouts identifies established and novel skeletal phenotypes. Bone Res 2, 14034 (2014). 14. Yang, Y., Topol, L., Lee, H. & Wu, J. Wnt5a and Wnt5b exhibit distinct activities in coordinating chondrocyte proliferation and differentiation. Development 130, 1003-15 (2003). 15. Balemans, W. et al. Identification of a 52 kb deletion downstream of the SOST gene in patients with van Buchem disease. J Med Genet 39, 91-7 (2002).

176

General Discussion

16. Boyden, L.M. et al. High bone density due to a mutation in LDL-receptor-related protein 5. N Engl J Med 346, 1513-21 (2002). 17. Brunkow, M.E. et al. Bone dysplasia sclerosteosis results from loss of the SOST gene product, a novel cystine knot-containing protein. Am J Hum Genet 68, 577-89 (2001). 18. Leupin, O. et al. Bone overgrowth-associated mutations in the LRP4 gene impair sclerostin facilitator function. J Biol Chem 286, 19489-500 (2011). 19. Little, R.D., Recker, R.R. & Johnson, M.L. High bone density due to a mutation in LDL- receptor-related protein 5. N Engl J Med 347, 943-4; author reply 943-4 (2002). 20. Hellemans, J. et al. Loss-of-function mutations in LEMD3 result in osteopoikilosis, Buschke-Ollendorff syndrome and melorheostosis. Nat Genet 36, 1213-8 (2004). 21. Janssens, K. et al. Mutations in the gene encoding the latency-associated peptide of TGF- beta 1 cause Camurati-Engelmann disease. Nat Genet 26, 273-5 (2000). 22. Gelb, B.D., Shi, G.P., Chapman, H.A. & Desnick, R.J. Pycnodysostosis, a lysosomal disease caused by cathepsin K deficiency. Science 273, 1236-8 (1996). 23. Kornak, U. et al. Loss of the ClC-7 chloride channel leads to osteopetrosis in mice and man. Cell 104, 205-15 (2001). 24. Kornak, U. et al. Mutations in the a3 subunit of the vacuolar H(+)-ATPase cause infantile malignant osteopetrosis. Hum Mol Genet 9, 2059-63 (2000). 25. Guerrini, M.M. et al. Human osteoclast-poor osteopetrosis with hypogammaglobulinemia due to TNFRSF11A (RANK) mutations. Am J Hum Genet 83, 64-76 (2008). 26. Sobacchi, C. et al. Osteoclast-poor human osteopetrosis due to mutations in the gene encoding RANKL. Nat Genet 39, 960-2 (2007). 27. Van Dijk, F.S. & Sillence, D.O. Osteogenesis imperfecta: Clinical diagnosis, nomenclature and severity assessment. Am J Med Genet A 164, 1470-81 (2014). 28. Laine, C.M. et al. WNT1 mutations in early-onset osteoporosis and osteogenesis imperfecta. N Engl J Med 368, 1809-16 (2013). 29. Narumi, S. et al. Various types of LRP5 mutations in four patients with osteoporosis- pseudoglioma syndrome: identification of a 7.2-kb microdeletion using oligonucleotide tiling microarray. Am J Med Genet A 152A, 133-40 (2010). 30. Keupp, K. et al. Mutations in WNT1 cause different forms of bone fragility. Am J Hum Genet 92, 565-74 (2013). 31. Korvala, J. et al. Mutations in LRP5 cause primary osteoporosis without features of OI by reducing Wnt signaling activity. BMC Med Genet 13, 26 (2012).

177

General Discussion

32. Appelman-Dijkstra, N.M. & Papapoulos, S.E. From disease to treatment: from rare skeletal disorders to treatments for osteoporosis. Endocrine 52, 414-26 (2016). 33. Leder, B.Z. et al. Two years of Denosumab and teriparatide administration in postmenopausal women with osteoporosis (The DATA Extension Study): a randomized controlled trial. J Clin Endocrinol Metab 99, 1694-700 (2014). 34. Qin, A. et al. V-ATPases in osteoclasts: structure, function and potential inhibitors of bone resorption. Int J Biochem Cell Biol 44, 1422-35 (2012). 35. Schaller, S. et al. The chloride channel inhibitor NS3736 [corrected] prevents bone resorption in ovariectomized rats without changing bone formation. J Bone Miner Res 19, 1144-53 (2004). 36. Mullard, A. Merck &Co. drops osteoporosis drug odanacatib. Nat Rev Drug Discov 15, 669 (2016). 37. Cosman, F. et al. Romosozumab Treatment in Postmenopausal Women with Osteoporosis. N Engl J Med 375, 1532-1543 (2016). 38. Iyer, S.P. et al. A Phase IB multicentre dose-determination study of BHQ880 in combination with anti-myeloma therapy and zoledronic acid in patients with relapsed or refractory multiple myeloma and prior skeletal-related events. Br J Haematol 167, 366-75 (2014).

178

LIST OF ABBREVIATIONS

180

List of Abbreviations

Abbreviations

AP-1 Activator Protein 1 APC Adenomatous Polyposis Coli AR Androgen Receptor B.Pm Bone Perimeter BFR Bone Formation Rate BMD Bone Mineral Density BR Buckling Ratio BS Bone Surface BV Bone Volume CADD Combined Annotation Dependent Depletion CaMKII Calcium/Calmodulin-Dependent Protein Kinase Type II cAMP Cyclic AMP CaN Calcineurin CAT Catalase Conn.D Connecting Density CREB Camp Response Element Binding CSA Cross-Sectional Area CSMI Cross-Sectional Moment Of Inertia Ct.Po Cortical Porosity Ct.Th Cortical Thickness CTX-I C-Terminal Telopeptide DAPI 4',6-Diamidino-2-Phenylindole DKK Dickkopf DNA Deoxyribonucleic Acid DsiRNA Duplex SiRNA DVL1 Dishevelled-1 DXA Dual Enery X-Ray Absorptiometry ELISA Enzyme-Linked Immunosorbent Assay ERα Estrogen Receptor α ExAC Exome Aggregation Consortium FN Femoral Neck FOXO Forkhead Box O FS Femoral Shaft FZ Frizzled GFP Green Fluorescent Protein GPER G-Protein Coupled Estrogen Receptor GPx Glutathione Peroxidase GSH Glutathione Peroxidase GSK-3β Glycogen Synthase Kinase 3β GWAS Genome-Wide Association Studies HAL Hip Axis Length HCI Hyperostosis Cranialis Interna HEK Human Embryonic Kidney

181

List of Abbreviations

HWE Hardy-Weinberg Equilibrium IGF-1 Insulin-Like Growth Factor 1 IL-6 Interleukin 6 IT Intertrochanteric JNK C-Jun Termine Kinase KO Knockout LD Linkage Desequilibrium LRP5/6 Low Density Lipoprotein Receptor-Related Proteins 5/6 LS Lumbar Spine MAF Minor Allele Frequency miRNA MicroRNA MS Mineralizing Surface Ms.D Midshaft Diameter NCBI National Center For Biotechnology NFAT Nuclear Factor Of Activated T-Cells NGS Next-Generation Sequencing NN Narrow Neck NOX NADPH Oxidase NSA Neck-Shaft Angle OAS Odense Androgen Study OASE Elderly OAS Cohort OASY Young OAS Cohort Ob.N Osteoblast Number Ob.S Osteoblast-Covered Surface Oc.N Osteoclast Number Oc.S Osteoclast-Covered Surface OMIM Online Mendelian Inheritance In Man OPG Osteoprotegerin PCP Planar Cell Polarity PCR Polymerase Chain Reaction PICP Procollagen I C-Terminal Propeptide PKC Protein Kinase C PLC Phospholipase C PTH Parathyroid Hormone PTHrP Parathyroid-Related Protein RANK Receptor Activator Of Nuclear Factor Kappab RANKL Receptor Activator Of Nuclear Factor Kappab Ligand RhoGEFs Rho Guanine Nucleotide Exchange Factors RISC RNA-Induced Silencing Complex RNA Ribonucleic Acid ROCK Rho-Associated Protein Kinase ROI Region Of Interest ROR2 Receptor Tyrosine Kinase-Like Orphan Receptor 2 ROS Reactive Oxygen Species SFRP Secreted Frizzled-Related Protein SLC39A14 Solute Carrier Family 39 Member 14

182

List of Abbreviations

SNP Single Nucleotide Polymorphisms SOD Superoxide Dismutase Tb.N Trabecular Number Tb.S Trabecular Separation TBS Tris-Buffered Saline TCF/LEF T-Cell Factor/Lymphoid Enhancer Factor TGF-β Transforming Growth Factor β TH Total Hip TNF-α Tumor Necrosis Factor α TV Tissue Volume WB Whole Body WES Whole-Exome Sequencing WGS Whole-Genome Sequencing WIF-1 WNT Inhibitory Factor 1 WNT Wingless-Type Mouse Mammary Tumor Virus Integration Site WT Wildtype Z Section Modulus ZIP Zrt-, Irt-Related Protein ZnT Zinc Transporter

183

List of Abbreviations

184

CURRICULUM VITAE

186

Curriculum Vitae

Curriculum Vitae – Gretl Hendrickx

Personal Details

Name/First name Hendrickx Gretl Address Ruytenburgstraat 8, 2600 Berchem (Antwerp), Belgium Date of birth 18.12.1989 Email [email protected]

Education and Academic Career

Education 2010 – 2012 Master of Molecular and Cellular Biomedical Science University of Antwerp, Belgium

Master dissertation entitled: “Functional and genetic analysis of WNT proteins in bone formation” › Graduated with Great Distinction

Promotor: Prof. Dr. Wim Van Hul

2007 – 2010 Bachelor in Biomedical Science

2001 – 2007 Mathematics-Science, Instituut Spijker, Hoogstraten

Academic career 2012 – Present PhD in Biomedical Sciences Center of Medical Genetics, University of Antwerp, Belgium

PhD thesis entitled: Bone Homeostasis as a Multifaceted Puzzle: From WNT Ligands to Metal-ion Transporters

Promotor: Prof. Dr. Wim Van Hul Co-promotor: Dr. Eveline Boudin

Additional Scientific Training

April 2016 Skeletal Phenotyping of Mice University Medical Center Hamburg-Eppendorf, Hamburg, Germany

September 2013 European Calcified Tissue Society (ECTS) PhD Training Course University Medical Center Hamburg-Eppendorf, Hamburg, Germany

October 2012 Laboratory Animal Science, FELASA, type C University of Antwerp, Belgium

187

Curriculum Vitae

Grants and Awards

Young Investigator Award by the American Society for Bone and Mineral Research (ASBMR) September 2016, Atlanta, Georgia, United States of America Abstract entitled: A Missense Mutation in the ZIP14 Gene Results in Abnormal Skull Growth in Patients with Hyperostosis Cranialis Interna

Research Foundation Flanders (FWO) Travel Grant for the Annual meeting of the American Society for Bone and Mineral Research (ASBMR) September 2016, Atlanta, Georgia, United States of America

Abstract entitled: A Missense Mutation in the ZIP14 Gene Results in Abnormal Skull Growth in Patients with Hyperostosis Cranialis Interna

Belgian Bone Club (BBC) Travel Award for the Herbert Fleisch Workshop of the International Bone and Mineral Society (IBMS) February 2016, Bruges, Belgium

Abstract entitled: WNT16 requires Gα12, Gα13 and Gαq subunits as intracellular partners for both its canonical and non-canonical WNT signaling activity in osteoblasts

Rosa Blanckaert Legacy Grant for Young Investigators by the University of Antwerp October 2015, Antwerp, Belgium

Project entitled: “Gα subunits as intracellular partners of WNT16 in osteoblasts – a functional analysis”

Publication List

Gretl Hendrickx, Eveline Boudin and Wim Van Hul. WNT16 Requires Gα Subunits as Intracellular Partners for Both its Canonical and Non-Canonical WNT Signaling Activity in Osteoblasts. Manuscript in preparation.

Gretl Hendrickx, Eveline Boudin, Ellen Steenackers, Torben Leo Nielsen, Marianne Andersen, Kim Brixen and Wim Van Hul. Genetic screening of WNT4 and WNT5B in two populations with deviating bone mineral densities. Calcified Tissue International 2017; 100:244– 249.

188

Curriculum Vitae

Gretl Hendrickx, Eveline Boudin and Wim Van Hul. Het ontstaan van primaire osteoporose: een complex samenspel tussen onze genen en onze omgeving. OrthoRheumato 2016; 14: 7-11.

Gretl Hendrickx, Eveline Boudin and Wim Van Hul. A look behind the scenes: the risk and pathogenesis of primary osteoporosis. Nature Reviews Rheumatology 2015; 11;462–474.

Gretl Hendrickx, Eveline Boudin, Igor Fijałkowski, Torben Leo Nielsen, Marianne Andersen, Kim Brixen and Wim Van Hul. Variation in the Kozak sequence of WNT16 results in an increased translation and is associated with osteoporosis related parameters. Bone 2014; 59:57-65.

Eveline Boudin, Igor Fijalkowski, Timur Yorgan, Stephan Sonntag, Ellen Steenackers, Gretl Hendrickx, Silke Peeters, Annelies Demaré, Geert Mortier, Patrick D’Haese, Thorsten Schinke and Wim Van Hul. The Lrp4 R1170Q homozygote knock-in mouse recapitulates the bone phenotype of sclerosteosis in human. Journal of Bone and Mineral Research. Revision submitted.

Eveline Boudin, Igor Fijałkowski, Gretl Hendrickx and Wim Van Hul. Genetic control of bone mass. Molecular and Cellular Endocrinology 2016; 432:3-13.

Oral Presentations

Belgian Society of Human Genetics 17th annual meeting 2017: “Human Genetics goes Somatic”, Louvain-la-Neuve, Belgium. Presentation entitled: “Conditional Mouse Models Support a Role for SLC39A14 (ZIP14) in Hyperostosis Cranialis Interna and in Bone Homeostasis”

American Society for Bone and Mineral Research (ASBMR) – Annual Meeting 2016, Atlanta, Georgia, United States of America. Presentation entitled: “A missense mutation in the SLC39A14 (ZIP14) gene results in abnormal skull growth in patients with Hyperostosis Cranialis Interna”

Refereeravond Departement Neurochirurgie, 8 april 2016, Universitair Ziekenhuis Antwerpen (UZA), Edegem, Belgium. Presentation entitled: “Hyperostosis Cranialis Interna: Validatie van ZIP14 als causatief gen”

Sybil (Systems Biology for the Functional Validation of Genetic Determinants of Skeletal Diseases) – Annual Meeting 2015, Vienna, Austria Presentation entitled: “Hyperostosis Cranialis Interna – Functional validation of the L441R variant in ZIP14.”

European Calcified Tissue Society (ECTS) – Annual Meeting 2014, Prague, Czech Republic. Presentation entitled: “rs55710688 in the Kozak sequence of WNT16 can modulate canonical WNT signaling differently in different cell types.”

Belgian Society of Human Genetics 14th Annual Meeting 2014: “Translational Genetics: From Cage over Bench to Bed”, Antwerp, Belgium. Presentation entitled: “Variation in and around WNT16 affects osteoporosis related parameters.”

189

Curriculum Vitae

European Calcified Tissue Society (ECTS) – PhD training Course 2013, Hamburg, Germany. Presentation entitled: “Variation in the Kozak sequence of WNT16 results in an increased translation and is associated with osteoporosis related parameters.”

Poster Presentations

International Bone and Mineral Society (IBMS) – Herbert Fleisch Workshop 2016, Bruges, Belgium. Poster presentation: “WNT16 requires Gα12, Gα13 and Gαq subunits as intracellular partners for both its canonical and non-canonical WNT signaling activity in osteoblasts”

European Calcified Tissue Society (ECTS) – International Bone and Mineral Society (IBMS), 4th Joint Meeting 2015, Rotterdam, The Netherlands. Poster presentation: “Genetic screening of WNT5B in three populations with extreme bone mineral density values.”

Belgian Society of Human Genetics 15th annual meeting 2015: " Behaviour: Is it genetically determined?", Charleroi, Belgium. Poster presentation: “Genetic screening of WNT5B in three populations with extreme bone mineral density values.”

WNT Symposium 2013: “Stem Cells, Development and Disease”, Heidelberg, Germany. Poster presentation: “Genetic variation in WNT16 is associated with osteoporosis related parameters in healthy Caucasian men.”

Belgian Society of Human Genetics 13th annual meeting 2013: "Genetics of Human Development EXPOsed", Brussels, Belgium. Poster presentation: “Genetic variation in WNT16 is associated with osteoporosis related parameters in healthy Caucasian men.”

Educational Activities

Assistance Master Thesis: “Genetic and functional evaluation of the role of RIN3 in the pathogenesis of Paget’s disease of bone” by Raphaël De Ridder, 2nd Master in Biomedical Sciences, University of Antwerp, Belgium. Academic Year 2014-2015.

Assistance Bachelor Thesis: “De rol van WNT moleculen in botmetabolisme en –pathologieën” by Kenny Van Theemsche, 3rd Bachelor in Biomedical Sciences, University of Antwerp, Belgium. Academic Year 2014-2015.

Assistance Research Internship by Anouk Smits, 1st Master in Biomedical Sciences, University of Antwerp, Belgium. Academic Year 2014-2015.

190

Curriculum Vitae

Assistance Bachelor Thesis: “Nieuwe mogelijkheden voor de preventie of behandeling van osteoporose” by Ben Bouazza Youssef, 3rd Bachelor in Biomedical Sciences, University of Antwerp, Belgium. Academic Year 2013-2014.

Assistance Research Internship by Farnaz Takhsha, 1st Master in Biomedical Sciences, University of Antwerp, Belgium. Academic Year 2013-2014.

Assistance Research Internship by Yuliya Vostrikova, 1st Master in Biomedical Sciences, University of Antwerp, Belgium. Academic Year 2012-2013.

191

Curriculum Vitae

192

DANKWOORD

194

Dankwoord

Wat een tijd, zoveel ervaringen en een doctoraatsthesis rijker. Naast hard werk, is dit boek toch evenzeer het resultaat van geweldige professionele contacten en ondersteuning, in combinatie met de steun van een fantastische achterban van vrienden en familie.

Vooreerst wil ik de leden van mijn interne doctoraatscommissie, Prof. Dr. Vincent Timmerman en Prof. Dr. Marc Cruts, bedanken om m’n werk kritisch te evalueren de afgelopen viereneenhalf jaar. Also thanks to all other members of the jury, Prof. Dr. Bart Loeys, Prof. Dr. Geert Carmeliet and Prof. Dr. Martine Cohen-Solal, for their willingness and time to participate in the evaluation of my doctoral thesis.

Thanks to all the co-authors for their contribution to the work presented in this thesis, especially the work performed for Hyperostosis Cranialis Interna (HCI) and ZIP14. Vere, jij hebt me dit werk vier jaar geleden met dubbele gevoelens overgedragen, na er jaren zelf aan gewerkt te hebben. Het was complexe materie, maar we kunnen er nu samen terecht trots op zijn. Hans en Jérôme, jullie hebben HCI klinisch beschreven en besloten dit verhaal van Maastricht tot bij ons in Antwerpen te brengen. Bedankt voor de fantastische samenwerking, de fijne meetings en jullie oneindige interesse in het genetische luik van HCI. Christophe Hermans wil ik ook bedanken voor zijn bijdrage aan dit werk, jij stond altijd klaar om die enkele coupes van mij telkens samen te kleuren. Lastly, a special thanks goes out to Timur and Thorsten, for their guidance and work in the phenotyping of the mice in Hamburg. I felt truly welcome at your lab in Hamburg and it was a privilege to see and perform the phenotyping there.

Een uitgebreid woord van dank gaat uit naar mijn promotor, Wim. Bedankt voor alle kansen die je me gegeven hebt om me te ontplooien tot de onderzoeker die ik nu ben, maar me toch met de voeten op de grond te houden. Bedankt voor de vrijheid en het vertrouwen dat je me gaf in wat ik zelf graag wou doen met m’n onderzoek, maar me toch op tijd temperde zodat ik niet té veel hooi op m’n vork nam. Algemeen bedankt voor de zeer fijne en gemoedelijke samenwerking gedurende mijn masterthesis en de jaren van mijn doctoraat, tot op de dag van vandaag.

Eveline, jij hebt me intussen ook al vele jaren begeleid, eerst in mijn masterthesis en daarna als co-promotor van mijn doctoraat. Een groot deel van mijn tijd op het Centrum Medische Genetica hebben we naast elkaar gezeten op de bureau. Dit was ideaal voor mij om à la minute mijn wildste ideeën/theorieën, inclusief tekeningen in veel kleuren, met jou te kunnen delen. Hier moest je vaak mee lachen, maar toch was je ook altijd bereid om mee na te denken, kritisch te zijn en mijn werk na te lezen. Bedankt om me telkens met m’n voeten op de grond te houden, maar ook de nodige steun te bieden op momenten dat ik het moeilijk had.

195

Dankwoord

Een vaste waarde in onze onderzoeksgroep waar ik zeer veel aan gehad heb, is Ellen S. Ellen, wat hebben wij toch allemaal met tweetjes meegemaakt in het animalarium. Onze Zip14 muizen waren springers en dat hebben we vaak geweten. Maar goede of slechte ervaringen, we hebben het altijd met een lach gedaan en dat maakte het vele werk een stuk lichter. Ook al zei je vaak dat “dit je job was” als laborante, toch wil ik je uitdrukkelijk bedanken om me te helpen met m’n muizenwerk. We waren hierin een geoliede machine, een dream team. Enorm bedankt!

Dan kom ik tot de bot-obesitas onderzoeksgroep, de BoObies, en de andere leden van onze bureau. Ellen S en Eveline, jullie zaten een lange tijd aan mijn zijde. Naast jullie professionele bijdrage aan dit werk, zijn jullie ook mentaal een steun geweest. Van een luisterend oor tot het wegdromen naar lekker eten en restaurantbezoekjes, het kon allemaal. Bedankt daarvoor! Evi, wij zijn destijds samen gestart aan onze masterthesis en aan een doctoraat in de BoOb-groep. Tijdens het IWT hebben we veel aan elkaar gehad, maar ook daarna was jij er altijd voor steun en een babbel, dankjewel daarvoor. Hetzelfde geldt voor Ellen G, Silke, Raphaël en Sara –heel fijn om jullie nadien mee in de BoOb groep te hebben! Silke, bedankt voor de leuke tijd in Wenen en Berlijn en de (virtuele) babbels na de werkuren. Naast de BoOb groep, vervolledig(d)en Esther en Matthias de bureau. Een volle bureau, maar wat een fantastische groepsdynamiek. Coffee-time zal nooit meer hetzelfde zijn! Bovendien: gaan lunchen, take-out (‘pizzavisjes’) bestellen, escape rooms, persoonlijkheidstesten… We hebben het allemaal gedaan met z’n negenen. Bedankt om zo’n topbureau te zijn!

Daarnaast wil ik alle (ex-)collega’s van het Centrum Medische Genetica, research en diagnostiek, bedanken voor alle kleine of grote hulp en leuke babbels tijdens mijn master/doctoraatsthesis. Mede door jullie allemaal was het ontzettend fijn werken hier.

Aan mijn vriendinnen: wat zou ik zijn zonder jullie? Anouchka, Katrijn, Karlien, Kathleen, Liesbeth, Nele, Hanna, Lore, Anne en Heleen, of kortweg de Toyboy Girls, de meisjes van ‘Woagstroate’: jullie ken ik bijna allemaal al van de lagere/middelbare school en onze band is enkel sterker geworden sinds we in Antwerpen hebben verder gestudeerd. We zijn allemaal heel verschillend in wat we doen, maar delen allen dezelfde gedrevenheid op professioneel vlak en waarden op vriendschappelijk vlak. Stuk voor stuk topvrouwen die sterk in hun schoenen staan. We zijn er altijd voor elkaar en jullie dus ook voor mij, de afgelopen jaren. Ook al durf ik in eerste instantie wel eens bloemen van jullie te weigeren aan mijn voordeur, ik ben jullie ongelooflijk dankbaar voor de talrijke gezellige momenten, tripjes naar IJsland, Praag, Rotterdam, Maastricht… en simpelweg jullie vriendschap. Jessie, Jasmine en Bieke, jullie ken ik sinds de opleiding Biomedische Wetenschappen en zijn ook doorgestroomd in een doctoraat, waardoor

196

Dankwoord we veel hebben gehad aan elkaar op dat vlak. Jessie en Jasmine, onze trip naar Hamburg zal ik niet gauw vergeten, net als de vele momenten bij elkaar thuis of in de nieuwste koffie- of cocktailbar in Antwerpen. Dankjewel! Jessie, jij bent doorheen de jaren een beetje mijn ‘Antwerpse zus’ geworden; we zijn er non-stop voor elkaar geweest en gingen samen sporten, kookten samen, werkten samen en konden samen niets doen. Reizen naar Kroatië, Griekenland, Indonesië, Madrid, Hamburg, Berlijn, Amsterdam, Zeeland… We hebben zoveel gezien en beleefd samen. Bedankt voor al die mooie herinneringen, je gezellige karakter en oprechte vriendschap.

Hiernaast wil ik mijn ‘schoonvrienden’ ook bedanken. Hanne, Lien, Woody, Chloé, Wouter, Lisser, Jos, Dorien, Marjon, David en Robin – jullie zijn terecht ‘de Supers’. Lien en Woody, bedankt voor de vele leuke avonden in Antwerpen en tripjes daarbuiten, inclusief de bijhorende raadsels. Lien, ook in het bijzonder nog dankjewel om me te helpen bij het ontwerp van de cover van dit boek. Hanne, Wouter en Chloé, gezelschapspellen spelen en ‘Sunday chills’ met jullie waren altijd enorm fijn, dankjewel!

Marjan, Aimée, Jelle, Mariekje, Toon en Jade; Bedankt om zo’n fijne en warme schoonfamilie te zijn. Niet enkel voor praktische zaken, maar ook voor een babbel of een luisterend oor, stonden jullie altijd klaar. Jelle en Mariekje, ook met jullie hebben Jasper en ik vele avonden gezelschapsspelen gespeeld. Ook al zijn we vooral verloren van jullie, het waren steeds zeer gezellige avonden, dus bedankt daarvoor!

Moemoe en Vava, op nieuwsgierigheid, gedrevenheid en humor staat duidelijk geen leeftijd voor jullie. Ondanks jullie 19 kleinkinderen hebben, slagen jullie erin ieder van hen te appreciëren zoals ze zijn, op de hoogte te blijven van hoe het met iedereen gaat, wat we allemaal studeren of doen van werk en heel de familie samen te brengen. Ik wil jullie, samen met mijn nonkels en tantes, neven en nichten, dan ook bedanken voor de vele feestdagen, zondagen en weekends dat we met z’n allen hebben doorgebracht.

Martijn, Ties en Simon, als mijn broers en zus betekenen jullie ontzettend veel voor mij. Martijn, grote broer, bedankt voor je rationele kijk op de dingen, je passie voor architectuur en relatief hoge plaagstokgehalte. Leen, bedankt om hem zo goed in toom te houden, maar ook voor je sympathieke nieuwsgierigheid. Ties(je), kleine zus, tot ik verder ging studeren hebben we altijd samen op eenzelfde kamer geslapen, wat een unieke zussenband heeft gecreëerd. Bedankt voor je instant enthousiasme, je passie voor taal en toneel en de wederkerende vraag of je mij ‘een knuffel mocht geven’. Simon(tje), ook al schelen we zeven jaar, wij zijn altijd twee handen op één

197

Dankwoord buik geweest. Bedankt om jezelf te zijn, je passie voor film en bakken en de talrijke telefoontjes en bezoekjes.

Mama en Papa, van jongs af aan wisten jullie blijkbaar al dat ik graag zelfstandig was en goed wist wat ik (niet) wou. Jullie hebben altijd in me geloofd en me weten stimuleren én temperen op de juiste momenten. Van jullie heb ik geleerd wat hard werken is, dat je kritisch moet zijn en dat je moet doen wat je zelf graag wilt doen. Je eigen weg afleggen en niet omkijken. Het gedicht op de laatste pagina van dit boek staat er omdat jullie een zeer gelijkaardige tekst op mijn verjaardagskaart hebben gezet vorig jaar. Jullie kennen me zo ontzettend goed. Naast hard werken, was het voor jullie minstens even belangrijk om tenvolle van het leven te genieten, in alle simpliciteit. Bedankt om tot op heden ook zo’n warm nest te creëren voor Martijn, Ties, Simon en ik.

Jasper, jij krijgt mijn laatste paragraaf van dank. Saving the very best for last. In de zomer van 2014 kwam jij uit het niets voorbij fladderen en was ik eigenlijk meteen verkocht. Sindsdien vormen we een team met twee duidelijke doelen: genieten en gelukkig zijn. Jij brengt mijn work- life balance effectief in ‘balance’ en zorgt niet enkel voor ontzettend lekker eten ’s avonds, maar ook voor een grenzeloze steun in alles dat ik doe. Met twee hebben we Finland, Zweden, Cambodja en Vietnam doorgereisd; avonturen om nooit te vergeten. En nu staat een volgend groot avontuur ons te wachten: Hamburg. Ik kan niet wachten om met jou dit avontuur aan te gaan. Jasper, bedankt voor je goede hart, je humor, je steun, je liefde. Ik zie je graag.

198

199

200

Ballerina

Gebaar voor gebaar reikt

ze haar lichaam over aan haar dromen

en telkens wordt ze meer

dan ze al was en telkens reikt ze weer.

Herman de Coninck.

Uit “De lenige liefde – Een dag als geen ander”