Choroid Plexus Tumours: Elucidating the Genomic Complexity of Rare Paediatric Brain Tumours

Home , Choroid plexus papilloma, Choroid plexus tumor

Diana Margarita Merino A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Medical Biophysics University of Toronto © Copyright by Diana M. Merino 2015

Choroid Plexus Tumours: Elucidating the Genomic Complexity of Rare Paediatric Brain Tumours

Diana M. Merino

Doctor of Philosophy

Department of Medical Biophysics University of Toronto

2015

Abstract Choroid plexus tumours (CPTs) are rare intraventricular neoplasms, most commonly found in children, with variable clinical presentation and patient outcome. CPT biology is poorly understood largely due to limited tumour availability and the lack of accurate in-vitro and in-vivo models of disease. My thesis work focused on elucidating the molecular alterations of these rare tumours using high-throughput approaches that enabled the identification of recurrent alterations affecting CPTs and facilitated the classification of clinically and molecularly distinct CPT subgroups. Chapter 1 describes the current clinical and biological understanding of CPTs, providing a background on what’s currently known and identifying research gaps addressed in future chapters. Chapter 2 describes the molecular landscape of CPTs, achieved by defining patterns in copy number, gene expression and DNA methylation, as well as telomere lengthening mechanisms observed in these tumours. Chapter 3 describes the creation of a novel mouse model of CPC and the identification of novel oncogenes involved in CPC development using a cross- species genomics approach. Chapter 4 examines future directions in choroid plexus research and suggests novel avenues to pursue in order to improve our collective understanding on the

ii mechanisms driving CPC development and how to best approach the clinical management of patients with CPC. The aim of this thesis is to highlight recent advances in CPT biology and to provide the molecular framework on which to build an evidence-based treatment strategy to effectively diagnose and treat patients affected by the disease.

iii

Acknowledgments

None of this work would have been possible without the support and commitment of extraordinary people who played a key role in the completion of my degree.

Thank you to my supervisor, Dr. David Malkin, for believing in me, for supporting me, providing me with personal and professional development opportunities, and for exemplifying how one can transform and make a difference in childhood cancer care. Your commitment to helping others is inspiring. I have learnt extensively from you and the dedication and integrity you demonstrate in everything you do. Thank you to the members of my supervisory committee

Drs. Thomas Kislinger, Stephen Meyn, and Michael Taylor, for their expertise, guidance and sincere advice.

I have had the blessing to meet extraordinary collaborators and work alongside them during my thesis project. Thank you to Dr. Richard Gilbertson for his guidance and support, and for mentoring me and opening the doors of his laboratory during my research exchange at St. Jude

Children’s Research Hospital. Thank you to the Gilbertson laboratory for their patience and guidance during my time training there. Thank you to Dr. Uri Tabori, for his sincere and accurate feedback and for his expert advice. Thank you to Dr. Jan Korbel for opening the doors of his laboratory at the European Molecular Biology Laboratories (EMBL) so that I can learn analytical techniques, and thank you to the Korbel laboratory for their guidance and support.

Thank you to my research mentors, Drs. Stephen Mack, Marc Remke, Vijay Ramaswamy, and

Adam Shlien, for their guidance, support and intellectual discussions over coffee.

Thank you to my Malkin lab family, past and present, for your amazing support. You have helped me keep my calm through it all. I will never forget our time together, especially the

iv legendary Christmas gift presentations, birthday celebrations and GGB picnics. Thank you to the

BTRC, for making me feel like one of your own and helping me with research ideas, advice and distractions.

This thesis is dedicated to my Lord and Saviour Jesus Christ, to whom I owe the gift of life, grace and hope, and for giving me the ability to work on something I enjoy so much; and to my family, who has supported, encouraged, and kept me grounded during this PhD journey. Mom and Dad, you have shown me what true wisdom is all about, you have encouraged me to follow my dreams, no matter how big they may be, and to fight for those dreams with determination while keeping my eyes fixed on things of above. Diego, Rachel and Daniel, thank you for always being there for me, for listening to my science and providing a much needed and refreshing point of view. You have influenced my life in more ways than you realize. Leora Valentina, thank you for filling this last year with your giggles and for reminding me about the miracle of life. To my friends, for helping me find joy in the small things, for your listening ears and for all the memories we have created and now share. Thank you for reminding me that life is not a sprint, but a marathon.

Lastly, I would like to dedicate this work to cancer patients and survivors whose generosity and cooperation allows us to investigate this disease more profoundly, and whose determination in fighting this disease inspire and motivate me to make a difference in this field. Thank you.

Table of Contents Abstract ...... ii

Acknowledgments ...... iv

Table of Contents ...... vi

List of Tables ...... xii

List of Figures ...... xiii

Chapter 1: Introduction ...... 1

1. Acknowledgements ...... 1

2. The choroid plexus ...... 1 2.1. Function ...... 2 2.2. Development ...... 3 2.3. Disease ...... 3

3. Tumours of the choroid plexus ...... 3 3.1. Demographic characteristics of CPT patients ...... 4 3.2. Classification of CPTs ...... 6 3.2.1. Pathology ...... 6 3.2.2. Immunohistochemistry ...... 7 3.3. Clinical presentation ...... 7 3.4. Clinical outcomes of CPT patients ...... 8

4. TP53-more than just a tumour suppressor gene ...... 9 4.1. TP53: The guardian of the genome ...... 9 4.2. TP53 and tumourigenesis ...... 9 4.3. Mutant TP53 and its oncogenic role ...... 11

5. Choroid plexus tumours and familial cancer syndromes ...... 12 5.1. CPC and the Li-Fraumeni syndrome ...... 12 5.2. CPTs and genetic disease ...... 13

6. Genomic instability in cancer: aneuploidy and chromothripsis ...... 14

6.1. Aneuploidy: causes and consequences ...... 14 6.1.1. Causes of aneuploidy ...... 15 6.1.2. Consequences of aneuploidy ...... 15 6.1.3. Aneuploidy and cancer ...... 17 6.1.4. Aneuploidy and TP53 ...... 18 6.2. Chromothripsis ...... 19 6.2.1. Chromothripsis and cancer ...... 19 6.2.2. Chromothripsis and TP53 ...... 20

7. Molecular background of choroid plexus tumours...... 20 7.1. TP53 and CPCs ...... 20 7.2. Chromosomal aberrations in CPTs ...... 22 7.3. Abnormalities in genes associated with CPT ...... 25 7.4. Telomere maintenance ...... 26

8. Animal Models of CPT ...... 28 8.1. Transgenic mouse models of CPP ...... 29 8.2. Conditional allele CPC mouse model ...... 29 8.3. Orthotopic zebrafish model of CPC ...... 30

9. Summary ...... 31

Chapter 2: Defining the genomic landscape of choroid plexus tumours ...... 32

1. Acknowledgements ...... 32

2. Abstract ...... 33

3. Introduction ...... 34 3.1. Choroid plexus tumours ...... 34 3.2. Mutant TP53 and its association with choroid plexus carcinomas ...... 34 3.3. Telomere maintenance in choroid plexus tumours ...... 35 3.4. Chromosomal instability is prevalent in choroid plexus tumours ...... 35

4. Results ...... 36 4.1. Unsupervised clustering reveals significant segregation of CPCs from CPPs and aCPPs ...... 36

vii

4.2. CPCs are characterized by frequent chromosome-wide aberrations and frequent uniparental disomy events ...... 37 4.3. CPPs and aCPPs share similar molecular signatures which correlate with favourable survival outcomes ...... 39 4.4. Ploidy analysis reveals novel CPC subgroups with unique molecular alterations ..... 41 4.5. Transcriptomic and epigenomic signatures differ in CPCs according to TP53 status 44 4.6. Allele-specific copy number analysis determines number of mutant p53 molecules in CPCs ...... 45 4.7. Methylation analysis of TERT upstream of transcription start site (UTSS) in CPTs reveals increased expression of TERT in a subset of CPCs...... 46 4.8. ALT is a recurrent feature in CPCs and not observed in CPPs ...... 48 4.9. ALT is associated with TP53 mutations in paediatric brain tumours ...... 49 4.10. Patient outcomes ...... 49 4.10.1. Patient outcomes for CPT histological subgroups ...... 49 4.10.2. TP53 status has a significant effect on the overall survival of CPC patients ...... 51 4.10.3. Number of mutant copies of TP53 predicts overall and event-free survival outcomes in CPC patients ...... 52 4.10.4. Mutant TP53 CPC patients exhibit no significant differences in survival outcomes according to ALT status ...... 52

5. Discussion ...... 53

6. Conclusion ...... 56

7. Materials and Methods ...... 56 7.1. Patients and sample preparation for microarray study ...... 56 7.2. Patients and sample preparation for TERT methylation study ...... 58 7.3. Patients and sample preparation for ALT Study ...... 58 7.4. TP53 sequencing ...... 58 7.5. Microarray processing & bioinformatics analysis ...... 58 7.6. Statistical analysis ...... 59

Chapter 3: Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes ...... 61

viii

1. Acknowledgements ...... 61

2. Abstract ...... 61

3. Introduction ...... 62 3.1. Choroid plexus carcinomas ...... 62 3.2. Cross-species approach to identify novel oncogenes ...... 63 3.3. CPC development in mouse ...... 64

4. Results ...... 64 4.1. Deletion of Tp53, Rb and Pten generates CPC in mouse ...... 64 4.2. Cross species analysis reveals candidate CPC oncogenes encoded in human chromosome 1p31.3-ter (mouse chromosome 4qC6-qE2) ...... 65 4.3. Gain of chromosome 4qC6-qE2 is an early event in CPC tumourigenesis ...... 69 4.4. Taf12, Nfyc and Rad54l promote aberrant proliferation of the developing choroid plexus epithelia ...... 70 4.5. Expression of Taf12, Nfyc and Rad54l is required to maintain CPC in vivo ...... 71 4.6. Taf12, Nfyc and Rad54l are required to initiate CPC ...... 72 4.7. Upregulation of Taf12, Nfyc and Rad54l drives CPC ...... 73 4.8. Aberrant DNA metabolism is a significant feature of CPC ...... 74

5. Discussion ...... 76

6. Conclusion ...... 79

7. Methods and Materials ...... 79 7.1. Patients and sample preparation for microarray study ...... 79 7.2. TP53 sequencing ...... 80 7.3. RNA and DNA microarray analysis ...... 80 7.4. Whole genome sequencing data analysis ...... 81 7.5. Statistical analysis ...... 81

Chapter 4: Thesis main findings and future directions in choroid plexus tumour research 82

1. Acknowledgements ...... 82

2. Salient findings obtained from thesis project ...... 82

3. High-throughput genomics and CPT research ...... 83

3.1. Clinical implications of the molecular substratification of CPTs ...... 84 3.1.1. CPPs and aCPPs are molecularly similar ...... 84 3.1.2. Molecular prognostic factors for CPC ...... 86 3.2. The use of next-generation sequencing to investigate CPC aetiology ...... 87 3.2.1. Whole genome sequencing reveals very low mutation rates in CPCs ...... 87 3.2.2. Chromothripsis as a novel mechanism in CPC development ...... 88 3.2.3. Identifying fusion lesions in CPC by RNA sequencing ...... 91

4. Mutant p53 & CPC prognosis ...... 94 4.1. Mutant p53 dosage ...... 94 4.2. Mutant p53 function ...... 96

5. Mouse models ...... 96

Chapter 5: Appendix 1- Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers ...... 99

1. Abstract ...... 99

2. Introduction ...... 100

3. Results ...... 100

4. Discussion ...... 109

5. Conclusion ...... 109

6. Methods ...... 109 6.1. Patient and sample collection...... 109 6.2. Copy number analysis...... 110

7. References: ...... 110

Chapter 6: Appendix 2- Supplemental Data ...... 113

1. Supplemental Experimental Procedures ...... 113 1.1. Chapter 2: Defining the genomic landscape of choroid plexus tumors ...... 113

1.2. Chapter 3: Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes ...... 115 1.3. Chapter 5: Appendix 1- Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers ...... 118

2. Supplemental Figures ...... 122 2.1. Chapter 3: Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes ...... 122 2.2. Chapter 5: Appendix 1- Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers ...... 128

References ...... 142

List of Tables

Table 1.1: Choroid plexus tumour chromosomal instability reported in literature ...... 23 Table 2.1: Frequency of TP53 mutations in CPTs and characterization of mutation types ...... 45 Table 2.2: Characterization of tumour and patient cohort ...... 50

xii

List of Figures Figure 1.1: Known clinical and molecular characteristics of choroid plexus tumours ...... 5 Figure 1.2: Distribution of germline TP53 mutations mapped to the coding sequence of TP53 . 10 Figure 1.3: Frequency of tumours associated with germline TP53 mutations ...... 11 Figure 1.4: Frequency of TP53 mutations in our cohort of CPCs ...... 21 Table 1.1: Choroid plexus tumour chromosomal instability reported in literature ...... 23 Figure 2.1: Unsupervised clustering of choroid plexus tumours ...... 36 Figure 2.2: Unsupervised clustering analysis of CPTs showed significant segregation of CPCs from CPPs with aCPPs ...... 37 Figure 2.3: Characterization of recurrent chromosome-wide gains and losses for each tumour subtype ...... 38 Figure 2.4: Unsupervised clustering of CPPs and aCPPs reveal no significant molecular differences between the two tumour subgroups ...... 40 Figure 2.5: Copy number analysis reveals aneuploidy on three CPT subgroups with CPCs demonstrating a binary distribution of ploidy values ...... 41 Figure 2.6: Genome-wide characterization of chromosome-wide aberrations and allelic abnormalities in 71 unique CPC samples ...... 42 Figure 2.7: Gene set enrichment analysis (GSEA) comparing the expression of different biological pathways and processes between hypodiploid (red) and hyperdiploid (blue) CPCs ...... 43 Figure 2.8: Unsupervised hierarchical clustering of gene expression and DNA methylation values demonstrate molecular heterogeneity among CPCs associated with TP53 status ...... 44 Table 2.1: Frequency of TP53 mutations in CPTs and characterization of mutation types ...... 45 Figure 2.9: UTSS methylation status and survival in CPTs ...... 47 Figure 2.10: Frequency of alternative lengthening of telomeres (ALT) in CPTs and its association to CPC survival ...... 48 Table 2.2: Characterization of tumour and patient cohort ...... 50 Figure 2.11: Kaplan-Meier curves depicting overall and event-free survival estimates of CPT patients by diagnosis ...... 50 Figure 2.12: Kaplan-Meier curves depicting overall survival and event-free survival estimates of CPCs by ploidy ...... 51

xiii

Figure 2.13: Kaplan-Meier curves depicting overall survival and event-free survival estimates of CPCs by TP53 mutation status ...... 51 Figure 2.14: Kaplan-Meier curves depicting overall and event-free survival estimates of CPC patients by number of mutated copies of TP53, as estimated by Sanger sequencing and allele-specific copy number analysis ...... 52 Figure 2.15: Sample flow-chart for CPT study ...... 57 Figure 2.1: Graphical abstract of cross-species choroid plexus carcinoma study ...... 62 Figure 3.2: Cross-species analysis of syntenic chromosomal gains in human and mouse CPC .. 66 Figure 3.3: A common set of 21 syntenic genes gained and overexpressed on 1p31.3-ter in human and 4qC6-qE2 in mouse, CPC ...... 68 Figure 3.4: Prognostic significance of RAD54L, TAF12 and NFYC expression in human CPC .. 71 Figure 3.5: DNA repair is upregulated in CPC ...... 75 Figure 3.6: Model proposed for the development of CPC from normal choroid plexus epithelium ...... 78 Figure 4.1: CPCs and CPPs are molecularly distinct ...... 85 Figure 4.2: Whole genome sequencing reveals that CPCs exhibit very low mutation rates and no recurrent point mutations ...... 87 Figure 4.3: Whole genome sequencing identifies evidence of chromothripsis event in CPC ...... 89 Figure 4.4: Copy number profile confirms chromothripsis event in CPC affecting chromosomes 1 and 19 ...... 90 Figure 4.5: Diagram of RAF1 and RAF1-TMEM40 fused protein found CPC ...... 91 Figure 4.6: Copy number and gene expression microarray analysis validates the molecular changes associated with the RAF1-TMEM40 fusion ...... 92 Figure 4.7: Physical location of RAF1 and TMEM40 on chromosome 3 ...... 93 Figure 4.8: Fusion profile of CPCs and affected pathways found by whole genome sequencing (WGS) and RNA sequencing analysis (RNAseq) ...... 94 Figure 4.9: Copy number profile of human and Myc-driven mouse CPC ...... 98 Figure 5.1: Somatic mutation frequency in bMMRD ultra-hypermutated cancers ...... 102 Figure 5.2: Consequences of polymerase mutations in bMMRD cancers ...... 104 Figure 5.3: Mutation spectrum of inherited and sporadic cancers ...... 106 Figure 5.4: Mutation threshold and rate in cancers with mismatch and polymerase mutations . 107

xiv

Supplemental Figures: Figure S3.1: A new mouse model of CPC ...... 123 Figure S3.2: Serial analysis of choroid plexus transformation in mice ...... 124 Figure S3.3: Functional in utero assessment of 1p31.3-ter/4qC6-qE2 candidate CPC oncogenes ...... 125 Figure S3.4: Taf12, Nfyc and Rad54l are required to initiate and maintain CPC in mice ...... 126 Figure S3.5: Taf12, Nfyc and Rad54l expression promote CPC in mice ...... 127 Figure S.5.1: Somatic rearrangements in bMMRD cancers ...... 128 Figure S.5.2: Number of single-nucleotide variations (SNVs) in non-neoplastic bMMRD DNA ...... 129 Figure S.5.3: MMR protein expression in non-neoplastic biallelic MMR mutant cells by protein blotting ...... 130 Figure S.5.4: In vitro GT mismatch repair assay ...... 131 Figure S.5.5: Frequency of somatic mutations in DNA replication and repair genes ...... 132 Figure S.5.6: Putative polymerase driver mutations found in ultra-hypermutated bMMRD tumors ...... 134 Figure S.5.7: Putative polymerase driver mutations found in ultra-hypermutated bMMRD tumors...... 134 Figure S.5.8: Timing of the polumerase E mutational signature in bMMRD/POLE cancers .... 135 Figure S.5.9: Correlation of mutation signatures ...... 136 Figure S.5.10: Mutation burden in sporadic colorectal and endometrial cancers ...... 137 Figure S.5.11: Suggested model for the mutation signature of bMMRD malignant cancers ..... 138 Figure S.5.12: Next generation sequencing coverage of bMMRD malignant cancers ...... 139 Figure S.5.13: Validation of POLE and POLD1 mutations ...... 141

Chapter 1: Introduction

1. Acknowledgements This chapter includes modified and reformatted excerpts from the following manuscripts: (1) “Molecular characterization of choroid plexus tumours reveals novel clinically relevant subgroups” published in Clinical Cancer Research (Merino et al. 2015) for which I collected, analyzed and interpreted the data, designed and executed the bioinformatics analyses, and wrote the manuscript; (2) “p53 and Hereditary Cancer” published in Subcellular Biochemistry (Merino & Malkin 2014) for which I conducted a comprehensive review and wrote the manuscript; (3) “Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes” published at Cancer Cell (Tong et al. 2015) for which I designed and executed the human choroid plexus carcinoma (CPC) bioinformatics analyses as well as the clinical correlation analyses with patients diagnosed with CPC; (4) “Methylation of the TERT promoter and risk stratification of childhood brain tumours: an integrative genomic and molecular study” published in The Lancet Oncology (Castelo-Branco et al. 2013) for which I extracted and purified high quality nucleic acid material for methylation and PCR analyses and conducted clinical outcomes analyses with all the choroid plexus tumour samples in the study; (5) “Alternative lengthening of telomeres is enriched in, and impacts survival of TP53 mutant paediatric malignant brain tumours” published in Acta neuropathologica (Mangerel et al. 2014) for which I extracted and purified high quality nucleic acid material for C-circle analyses and conducted clinical correlation analyses with all choroid plexus tumour samples; and (6) “Genome Sequencing of Paediatric Medulloblastoma Links Catastrophic DNA Rearrangements with TP53 Mutations” published in Cell (Rausch et al. 2012) for which I coordinated the submission of tumour microarray data of Li-Fraumeni syndrome and choroid plexus carcinoma patients.

2. The choroid plexus The discovery of the choroid plexus is credited to Herophilus (c. 335-280 B.C.), who first observed the lobulated structure on the lateral ventricles. Vesalius later described the gross anatomy of the choroid plexus of the lateral ventricles in his manuscript De Fabrica (1543),

2 which was followed by Willis and Ridley who fully described the choroid plexus of the fourth and third ventricles in 1664 and 1695, respectively (Dohrmann 1970). About 200 years later, Cushing described a key role of the human choroid plexus: the secretion of cerebrospinal fluid (CSF) (Cushing 1914; Lehtinen et al. 2013).

2.1. Function

The choroid plexuses are branched structures composed of villi, which project into the ventricles of the brain from the inner ventricular surface in a cauliflower-like shape (Brown et al. 2004). Each villus is composed of a single layer of cuboidal epithelial cells surrounding a core of connective tissue and numerous fenestrated blood capillaries. Choroid plexus epithelium (CPE) exhibit microvilli and clusters of cilia on its apical border, which are in direct contact with the CSF. The cells of the choroid plexus secrete CSF, and their specialized intracellular structure consisting of many mitochondria and a well-developed endoplasmic reticulum characterizes CPEs as secretory cells (Brown et al. 2004). CPE cells form a barrier between the blood and the CSF and the tight junction complexes formed between cells regulate the exchange of small molecules and ions in a very efficient manner (Wolburg & Paulus 2009). Because of their role as the blood-CSF barrier, CPE cells have a crucial role in the regulation of fluid pressure and balance in the brain ventricles, neuronal function, metabolism of the brain, immunological and inflammatory processes, neurosignalling, neuroprotection and neurodegeneration, among others (Janssen et al. 2013). Recent studies have demonstrated that the CPE cells secrete growth- promoting signals, such as Insulin-like growth factor-2 (Igf2) via the CSF, and that these signals regulate the growth and differentiation of neural stem cells in the cortical ventricular zone (Mairet-Coello et al. 2009), and regulate neurogenesis, formation of the uppermost layers of the cerebral cortex, and brain size in mice (Lehtinen et al. 2013). Moreover, CPE cells have been found to secrete morphogens and proteins that contribute to choroid plexus formation. Several genetic experiments in the mouse embryo have found that CPE cells release Sonic Hedgehog (SHH) via their basal surface to deeper mesenchymal cells, like pericytes in the underlying vascular bed and adjacent CPE progenitor cells, thus coordinating the co-development of vasculature and CPE cells, two distinct cell lineages that together form the choroid plexus (Huang et al. 2009; Nielsen & Dymecki 2010; Lehtinen et al. 2013).

3 2.2. Development

Choroid plexus develops adjacent to the embryonal dorsal midline, which is a well-known CNS patterning center, forming the lateral ventricles, the third ventricle, and the fourth ventricle from the telencephalon, the diencephalon and the hindbrain, respectively (Currle et al. 2005; Lehtinen et al. 2013). CPEs are derived from the neuroectoderm and as non-neuronal cells are defined as a subtype of microglia (Wolburg & Paulus 2009). Lineage tracing analyses have demonstrated CPEs originate from the roof plate and CPE progenitors have been located near the anterior lower rhombic lip (Currle et al. 2005; Awatramani et al. 2003; Hunter & Dymecki 2007; Tong et al. 2015; Lehtinen et al. 2013; Huang et al. 2009). Bone morphogenic proteins (BMPs) have been identified as molecules regulating CPE fate through its communication with the roof plate (Hébert et al. 2002; Lehtinen et al. 2013; Watanabe et al. 2012)

2.3. Disease

Because of its role as the brain-CSF barrier, the choroid plexus and the secretion of CSF are affected by infectious diseases such as tuberculosis, cryptococcosis and nocardosis, as well as parasites, such as Trypanosoma brucei, and viruses, such as human immunodeficiency virus (HIV) and cytomegalovirus (CMV) (Wolburg & Paulus 2009; Johanson et al. 2011). Malfunction of the CPE has also been implicated in several neurological disorders, such as Alzheimer disease (AD), hydrocephalus, ischemic events, and multiple sclerosis (MS) (Lehtinen et al. 2013; Janssen et al. 2013; Wolburg & Paulus 2009). Most of these diseases are associated with aberrant CSF volume or composition, thus affecting CSF homeostasis. Cancer arising in the choroid plexus has also been observed and characterized in detail.

3. Tumours of the choroid plexus Tumours arising in the choroid plexus are collectively called choroid plexus tumours (CPT) and encompass three different subtypes with progressive degrees of malignant transformation. Choroid plexus papillomas (CPPs) are classified by the World Health Organization (WHO) as grade I, a benign, slow growing neoplasm. The WHO recently distinguished atypical choroid plexus papilloma (aCPP) from CPP as grade II, a relatively slow growing neoplasm exhibiting

4 low risk of spread and recurrence. Choroid plexus carcinomas (CPCs) are classified as grade III, a malignant neoplasm exhibiting high risk of spread and dissemination.

3.1. Demographic characteristics of CPT patients

CPTs are rare intracranial neoplasms arising from the neuroepithelial cells lining the cerebral ventricles (Figure 1.1A). With an annual incidence of 0.3 cases per million, CPTs account for 0.4-0.7% of all intracranial neoplasms (Wolff et al. 2002; Hasselblatt et al. 2006; Aguzzi et al. 2000; Cannon et al. 2015). In children, however, the frequency of these tumours is increased (Figure 1.1B), accounting for 2-5% of brain tumours, and reaching up to 20% of brain tumours in infants under 2 years of age (Ogiwara et al. 2012; Addo et al. 2011; Lafay-Cousin et al. 2010; Cannon et al. 2015; Dudley et al. 2015).

CPTs arise in the lateral, third and fourth ventricles; however, they have also been reported to develop extraventricularly (Sameshima et al. 2010; Peyre et al. 2012). The location of these tumours shifts caudally as the age of the patient increases, with younger patients developing tumours more frequently in the lateral ventricles, and older patients in the fourth ventricle and the cerebellopontine angle (Wolff et al. 2002; Tanaka et al. 2009). The median age for supratentorial tumours is 1.5 years, whereas infratentorial tumours occur at a median age of 22 years (Turkoglu et al. 2014).

Somatic mutations in TP53 are commonly observed in CPCs, having been detected in about 50% of CPC samples for which proper Sanger sequencing of the entire TP53 coding and adjacent intronic regions was performed (Tabori et al. 2010). (Figure 1.1C). The frequency of germline TP53 mutations in CPCs varies according to cohort. A small study based at Children’s Hospital Los Angeles reported 36% (4/11) of CPCs carried a germline TP53 mutation (Gozali et al. 2012). A larger cohort consisting of samples from The Hospital for Sick Children, the Collaborative Human Tissue Network and Children’s Hospital Los Angeles demonstrated a greater frequency of germline mutant TP53 in CPC patients (44%, 8/18) (Tabori et al., 2010). In southern Brazil, a region exhibiting enrichment in a particular germline TP53 mutation (R337H), the frequency of germline TP53 mutant CPCs reached up to 63% (14/22) (Custodio et al. 2011)(Refer to chapter 1.5.1).

A! B! Frequency of CPTs by age Incidence'of'CPT'in'children' 20%

15%

10% Frequency 5%

Percentageintracranialtumors of 0% Adult <1 yr <2 yrs <4 yrs <15 yrs Age Range Incidence Age range 0.3 cases per million/year (0.6% intracranial tumors) CFrequency)of)! TP53)mutant)and)wildtype)CPCs) D! 1.00

Frequency of TP53 mutations in CPC 0.75

0.50 OS Probability 0.25

TP53 wt TP53 mut 0.00 p=0.03 49% 0 1 2 3 4 5 6 7 8 9 10 51% Overall Survival (years) Number at risk CPC 35 30 25 21 19 14 12 10 9 7 6 CPP 19 17 12 10 5 4 4 2 2 1 1 aCPP 11 8 5 4 4 3 3 2 2 1 1 CPC CPP aCPP Figure 1.1: Known clinical and molecular characteristics of choroid plexus tumours (A) The choroid plexus is located in the lateral (right and left), third and fourth ventricles (highlighted with a red circle). The incidence of these tumours is of 0.3 cases per million per year, which accounts for 0.6% of all intracranial tumours diagnosed per year (B) The incidence of CPTs is greater in young individuals, with infants under 1 year of age being the most affected (C) TP53 mutations are prevalent in CPCs being found in approximately 50% of tumours (blue) in our cohort (D) Survival of CPTs varies according to tumour subtype. CPPs (yellow), and aCPPs (blue) exhibit favourable long-term overall survival, whereas CPCs (red) exhibit less favourable outcomes

Two recent studies using the US Surveillance Epidemiology and End Results (SEER) database have elucidated the demographic characteristics of a large group of CPTs. The first study focused only on paediatric CPTs, with 107 CPPs, collected between 2004 and 2010, and 95 CPCs, collected between 1978 and 2010 (Dudley et al. 2015). aCPPs were not included in this study. In this cohort, Dudley and colleagues found a trend for increased CPT incidence in males (123/202, 61%) than females (79/202, 39%), finding a slightly higher male to female ratio in CPCs relative to CPPs [65.3% (62/95), versus 57% (61/107)]. This group also found that CPC patients were younger at diagnosis than CPP patients, with 76% (72/95) of patients being diagnosed before the age of 5, compared to 49% (52/107), respectively (Dudley et al. 2015). The incidence of tumours by race (white vs. not white) was not significant in this paediatric cohort.

The second study focused on CPTs observed in patients of all ages, with a total of 349 samples (203 CPPs, 26 aCPPs, and 120 CPCs) (Cannon et al. 2015). In their cohort, Cannon et al. found there was no significant difference in the frequency of tumours in males (51%, 177/349) relative to females (49%, 171/349). However, just like in the paediatric population, there was a trend for increased frequency of CPC in males (59%, 71/120 versus 41%, 49/120). Similar to Dudley et al., this study found that patients with CPC presented at a younger age (median 3 years old, mean 14.8 years old), relative to patients with CPP (median 25 years old, mean 28.4 years old). Although these two reports are based solely on cases occurring in the United States, smaller CPT cohorts in Canada (Lafay-Cousin et al. 2011), Asia (Koh et al. 2014) and the Middle East (Turkoglu et al. 2014) parallel the demographics reported in these larger studies.

3.2. Classification of CPTs

3.2.1. Pathology

Upon pathological inspection, CPP retains most features of normal choroid plexus tissue: papillary structures covered in a single layer of cuboidal to columnar epithelial cells. Yet this benign neoplasm commonly exhibits increased cellular density and atypia. The intermediate aCPP may exhibit higher mitotic activity (≥2 mitosis per 10 high-power fields, hpf), increased nuclear pleomorphism and cellular density, and areas of solid growth, while CPC is characterized by frank signs of malignancy and include at least four of the following features: high mitotic rate, increased cellular density, nuclear pleomorphism, blurring of the papillary pattern, necrosis and brain invasion (Wolff et al. 2002; Jeibmann et al. 2007; Jeibmann et al. 2006; Brassesco, Valera, Neder, et al. 2009). Differentiating aCPPs from CPCs could be a challenge when tumour morphology is not well defined; however, differentiating between these distinct tumour subtypes is crucial as the therapeutic strategies offered and patient outcomes are significantly different. Recent studies investigating the molecular profile of these tumours have reported that CPPs and aCPPs are molecularly similar, but significantly different than CPCs (Merino et al. 2015; Japp et al. 2015), and suggested the use of molecular techniques to complement the pathological classification of CPTs.

3.2.2. Immunohistochemistry

The most sensitive and specific immunohistochemical markers for CPTs are Kir7.1 and stanniocalcin-1. Less specific markers include transthyretin, glial fibrillary acidic protein (GFAP), epithelial membrane antigen (EMA), and synaptophysin (Paulus et al. 1990; Paulus et al. 1999; Hasselblatt et al. 2006; Sarnat 1998). These markers, in addition to the non-diffuse staining of cytokeratin, facilitate the discrimination between choroid plexus tissue and other primary brain tumours or cerebral metastasis (Hasselblatt et al. 2006; Ikota et al. 2011; Rickert 2004). Nonetheless, diagnostic markers that distinguish between CPT subgroups have not been fully characterized, and the accuracy of previously identified CPC markers, such as carcinoembryonic antigen (CEA) and CD44; and CPP markers, S100 protein, transthyretin (TTR), and pre-albumin have not been validated due to inconsistent expression (Wolff et al. 2002) and incongruent findings (Shintaku et al. 2008; Misaki et al. 2011; Rickert & Paulus 2001; Karim et al. 2006).

3.3. Clinical presentation

The central role of the normal choroid plexus is the production of cerebrospinal fluid (CSF) (Segal & Pollay 1977; Strazielle & Ghersi-Egea 2000) and maintenance of the blood-brain barrier (Neumann et al. 1993; Johanson et al. 2011; Cannon et al. 2015). Thus, young CPT patients commonly present with hydrocephalus because of the overproduction of CSF, as well as by local expansion and complications of spontaneous hemorrhage (Rickert & Paulus 2001). Common presenting symptoms are associated with increased intracranial pressure and include headaches, nausea/vomiting, enlarged head circumference, abnormal eye movement, papilledema, seizures, ataxia and dysphasia in older patients (Lafay-Cousin et al. 2011; Turkoglu et al. 2014; Koh et al. 2014; Ogiwara et al. 2012; Bettegowda et al. 2012). Choroid plexus cells are rarely found in the CSF, however their presence in CSF cytology has been associated with neurological disorders such as meningitis and brain stem infarction (de Reuck & Vanderdonckt 1986), and cancer (Savage et al. 2012). Although the accumulation of the CPP cells in the CSF is clinically asymptomatic, the accumulation of CPC cells may be indicative of metastasis (Lafay-Cousin et al. 2010).

8 3.4. Clinical outcomes of CPT patients

The most common treatment approach for CPPs and aCPPs is surgical resection, with adjuvant therapy administered in very few cases of recurrent, metastatic or inoperable lesions (Addo et al. 2011; Safaee et al. 2012; Anderson et al. 2013; Palmer et al. 2010; Takahashi et al. 2009; Wrede et al. 2009). Long-term overall survival for CPP patients is very favourable ranging from 93- 100%, while the overall survival for aCPP patients is slightly lower relative to CPP, yet still favourable ranging from 83-100% (Wrede et al. 2009; Gozali et al. 2012; Merino et al. 2015) (Figure 1.1D). The malignant CPCs, on the other hand, require a much more aggressive treatment approach with surgical resection and combination chemo-and radiation therapy (Lafay-Cousin et al. 2010). A few recent studies have reported the use of high-dose chemotherapy followed by autologous peripheral stem cell transplant (aPSCT) to manage recurrent CPC cases (Lafay-Cousin et al. 2010; Mosleh et al. 2013; Koh et al. 2014) leading to favourable outcomes. CPC patients also have increased risk of tumour progression and metastasis (Rickert & Paulus 2001; Tabori et al. 2010; Gozali et al. 2012). Furthermore, survivors of the malignant neoplasm display significant neurocognitive and developmental deficits (Lafay-Cousin et al. 2010; Lena et al. 1990). Currently, the degree of tumour resection is considered the most important prognostic factor for all CPTs, but most significantly for CPCs (Wolff et al. 2002; Berger et al. 1998; Sun et al. 2014). Despite improvements in surgical techniques and adjuvant therapies, long-term survival outcomes for CPC patients are poor. Five-year overall survival ranges from 31.3%-65.9% (Lafay-Cousin et al. 2011; Gozali et al. 2012; Merino et al. 2015; Wrede et al. 2009; Dudley et al. 2015). Molecular analyses of CPTs have revealed associations between molecular alterations and clinical outcomes. Rickert and colleagues conducted an aCGH study in which they found that CPC patients whose tumours exhibited a gain of chromosome 9p and a loss of 10q exhibited longer survival (Rickert et al. 2002). This finding has not been replicated in other cohorts. More recently, Ruland and colleagues observed, after conducting a multivariate analysis taking into consideration the extent of tumour resection and treatment with chemotherapy, that CPC patients whose tumours harboured a loss in chromosome 12q had reduced survival (Ruland et al. 2014). Similarly, these findings have not been replicated in independent CPC cohorts.

TP53 mutations and immunopositivity have been significantly associated with reduced overall and event-free survival in patients with CPC (Tabori et al. 2010). Moreover, Tabori and colleagues found an association with increased tumour structural variation (TSV), a measure of frequency and magnitude of copy number aberrations, and increased risk of progression in CPCs. Our latest study investigated the allele-specific copy number status of TP53, and has shown that an increasing number of mutant TP53 copies is strongly associated with a reduction in overall and event-free survival in CPC patients. This association may be a reflection of the gain-of- function (GOF) activities observed on mutant TP53, which seem to intensify as the ratio of mutant to wildtype TP53 increases. Thus, loss of wildtype p53’s tumour suppressor function and the accrual of tumour-promoting phenotypic effects of mutant p53, such as enhanced proliferation, and inhibition of apoptosis, may promote tumour aggressiveness and worse patient outcomes (Merino et al. 2015).

4. TP53-more than just a tumour suppressor gene 4.1. TP53: The guardian of the genome

One of the most notable tumour suppressor genes, TP53, encodes the transcription factor protein p53, a ubiquitous protein implicated in the preservation of an intact genome. True to its name, the “guardian of the genome”, p53 is a key suppressor of malignant transformation and somatic alterations commonly observed in numerous cancers (Petitjean et al. 2007). In response to cellular stress signals, p53 activates pathways that regulate the cell cycle, DNA repair, and apoptosis (Diller et al. 1990; Shaw et al. 1992). Additionally, p53 is also involved in regulating cellular senescence and metabolism, which have been shown to contribute to cancer progression (Serrano et al. 1997; Vousden & Ryan 2009; Wang et al. 2013; Wang et al. 1998).

4.2. TP53 and tumourigenesis

Due to the key role of p53 in restricting tumour initiation and progression, it is not surprising that numerous cancers have acquired mechanisms to inactivate p53 and/or its molecular pathway, thus bypassing the cell’s innate tumour suppression system. In vitro studies examining the status of p53 in numerous cancer cell lines and tumours determined that the activity of p53 was

10 commonly lost due to gene mutations or deletions (Baker et al. 1989; Baker et al. 1990; Cheng & Haas 1990; Nigro et al. 1989; Wolf & Rotter 1985). Trp53 knockout mouse models demonstrated that although mice, for the most part, developed normally, they had an increased susceptibility to a variety of cancers, most frequently lymphomas, which developed earlier than in mice harbouring wildtype p53 (Donehower et al. 1992; Harvey et al. 1993; Kemp et al. 1993). Similarly, in humans, germline TP53 mutations are also associated with an increased cancer susceptibility and reduced age of onset than TP53 wildtype carriers. The Li-Fraumeni syndrome (LFS; OMIM 151623), a rare but highly penetrant familial cancer syndrome, is characterized by germline TP53 mutations inherited in an autosomal dominant manner (Malkin et al. 1990). Between 60 and 80% of ‘classic’ LFS families carry a mutant p53. According to the International Agency for Research on Cancer TP53 database (http://www-p53.iarc.fr/), germline mutations in 118 different codons have been reported within the 393 amino acid-long p53 coding region, demonstrating the wide variability of p53 alterations involved in this heritable cancer condition (Figure 1.2).

14%

12%

10%

6% Frequency(%) 4%

0% 1 63 103 125 143 158 174 190 206 222 237 252 271 281 298 321 343 TP53 Codon Number Figure 1.2: Distribution of germline TP53 mutations mapped to the coding sequence of TP53 Sample size n=2,655 (Modified from IARC TP53 database, R17 November 2013) This genotypic variability is similarly reflected by the phenotypic variability LFS patients demonstrate in the number and type of cancer diagnoses (Figure 1.3), as well as their age at presentation. Although some correlations between genotype and phenotype have been suggested

11 in LFS, there are many more genetic modifiers to be identified and their effect on cancer risk remains to be fully defined.

BREAST 27.5% BRAIN 14.0% SOFT TISSUES 11.6% BONES 7.8% ADRENAL GLAND 7.6% HEMATOP. SYSTEM 3.8% LUNG 3.5% COLORECTUM 3.5% STOMACH 3.4% SKIN 2.1% OVARY 1.6% OTHER SITES 1.5% PROSTATE 1.4% HEAD&NECK 1.3% PANCREAS 1.2%

Malignancies Malignancies UTERUS 1.1% LYMPH NODES 1.1% KIDNEY 1.1% LIVER 1.0% OTHER FEMALE GENITAL ORG 1.0% TESTIS 0.5% NERVES 0.5% THYROID 0.5% ESOPHAGUS 0.5% BLADDER 0.4% PERITONEUM 0.3% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% Percentage Figure 1.3: Frequency of tumours associated with germline TP53 mutations Sample size n=2,550. (Modified from IARC TP53 database, R17 November 2013)

4.3. Mutant TP53 and its oncogenic role

Non-synonymous mutations in TP53 abrogate the tumour suppressor function of this master regulator; however, several studies have revealed that mutant TP53 has novel oncogenic properties that promote tumour aggressiveness (Shaulian et al. 1992; Dittmer et al. 1993). TP53 mutants are described as dominant negative (DN) when they inhibit the tumour suppressive function of the wildtype protein. The dominant negative effect is exerted by the heteromerization of mutant p53 to wildtype monomers, changing the conformation of the tetramer, thus rendering it unable to bind DNA (Shaulian et al. 1992; Sigal & Rotter 2000). Other TP53 mutations can also be endowed with novel functions that are associated with increased cell invasion, migration, proliferation, drug resistance, genomic instability and polyploidy, epithelial to mesenchymal transition (EMT), and stem cell dedifferentiation, among others (Muller & Vousden 2014; Oren & Rotter 2010). These are known as gain-of-function (GOF) mutations. Several mutations spanning the coding region of TP53 have been associated with different GOF roles in vitro and in vivo (Olive et al. 2004; Hanel et al. 2013; Muller & Vousden 2014; Lang et al. 2004). The

12 location of these mutations influences the activity endowed to the mutated p53 protein (Goldstein et al. 2010). For instance, DNA contact mutations such as R248Q and R173H exhibit greater expression of an NFkB mediated cancer gene signature (CGS), suggesting that these mutant p53 molecules may interact with this signalling pathway to promote oncogenic GOF activities (Goldstein et al. 2010). Other conformational mutations within the Zn2+ binding domain, such as R175H and H179R exhibit a different CGS profile that has been associated with increased dominant negative activity. Several molecular mechanisms by which mutant p53 exerts its oncogenic functions have been proposed. Recently, Muller and Vousden highlighted four models by which mutant p53 may regulate tumour-promoting gene expression signatures. These include the direct interaction of mutant p53 with DNA, the formation of complexes with transcription factors and transcriptional cofactors to enhance or repress the expression of target genes, and the interaction with other proteins not directly involved in transcriptional regulation, but which may also promote tumourigenesis by enhancing or blocking their function (Muller & Vousden 2013). Understanding the effects exerted by mutant p53 is crucial for designing effective therapeutic strategies, which aim at restoring the wildtype p53 activity of p53 mutant tumours, promoting the degradation of mutant p53 cells and targeting the dysregulated mutant p53 regulated pathways (Muller & Vousden 2014).

5. Choroid plexus tumours and familial cancer syndromes 5.1. CPC and the Li-Fraumeni syndrome

CPC has been identified as a core tumour of the Li-Fraumeni syndrome (LFS) (Garber et al. 1990; Gozali et al. 2012; Yuasa et al. 1992). In LFS families, CPCs have been observed with multiple primary neoplasms, particularly neural tumours at early ages, and in sibling pairs (Garber et al. 1990). LFS-associated CPCs commonly harbour TP53 mutations due to the association between LFS and germline TP53 (Malkin et al. 1990). In southern Brazil, a recurrent non-synonymous germline mutation at codon 337 (c.1010G>A, genomic nucleotide number 17588), mapping to the oligomerization domain of p53, was initially identified in adrenocortical carcinoma (ACC) patients (Ribeiro & Sandrini 2001). The mutation, encoding a R337H amino acid substitution has been identified in a large number of Brazilian families, and tumour

13 frequency and spectrum of affected carriers shows an association with LFS and the Li-Fraumeni Like (LFL) syndrome, a term given to families with incomplete features of the classical LFS definition (Achatz et al. 2007). Haplotyping analysis using 29 single nucleotide polymorphisms (SNPs) in 12 unrelated R337H carriers has revealed that this common mutation exhibits a founder effect (Garritano et al. 2010). Garritano and colleagues further mapped the origin of this founder mutation in which mutation carrier families were distributed along a commercial route commonly used by Portuguese merchants in the eighteenth and nineteenth centuries. R337H families exhibit low penetrance in individuals under the age of 30 (20%), compared to classic LFS families (50%). However, penetrance increases with age and mirrors that of classic LFS families with a lifetime risk of cancer of 90% (Achatz et al. 2009). In this population the incidence of CPCs is greatly increased. (Custodio et al. 2011; Seidinger et al. 2011). In one study, 63.3% (14/22) CPC patients were positive for germline R337H mutation, while none of the 7 CPP patients carried this germline mutation (Custodio et al. 2011). In a second study, 69% (9/13) of CPCs tested carried the germline mutation, while none of the 5 CPPs examined were carriers. Moreover, CPC patients carrying this mutation appeared to have a trend to decreased age of onset and worse prognosis than non-carriers (Seidinger et al. 2011). A report from the Children’s Hospital Los Angeles (CHLA) further supported the correlation between CPCs and germline mutations in TP53 by reporting that 36% (4/11) of CPC patients examined harboured a germline TP53 mutation, while all of the 11 CPPs and 1 aCPP studied did not harbour any TP53 mutations (Gozali et al. 2012). A recent study headed by our research group collected CPC samples from numerous institutions around the world and found that 60% of CPCs (35/58) were mutant for TP53. Fifteen (15/58, 26%) of these samples belonged to 12 patients with LFS carrying a germline mutation in TP53. Furthermore, we observed no significant differences in clinical parameters, such as age of onset and gender, or outcomes between LFS and non-LFS CPCs (Merino et al. 2015).

5.2. CPTs and genetic disease

Both germline and somatic mutations in hSNF5/INI1 (SMARCB1), a member of the SWI/SNF (switch/sucrose nonfermentable) ATP-dependent chromatin-remodeling complex have been associated with CPC (Sévenet et al. 1999; Gessi et al. 2003; Zakrzewska et al. 2005). CPCs have

14 been described as part of the rhabdoid predisposition syndrome, where patients carry a germline mutation of SMARCB1 (Sévenet et al. 1999). Recent reports have shown that the expression of INI1 is a distinguishing feature between atypical teratoid rhabdoid tumours (ATRT) and CPCs, with CPCs retaining immunopositivity of INI1 compared to ATRTs (Judkins et al. 2005). Currently, INI1 staining is performed to exclude the diagnoses of ATRT in CPT cohorts (Tabori et al. 2010; Hasselblatt et al. 2006). Exploring the role of the hSNF5/INI1 gene in the development of CPCs is warranted. CPPs are associated with the Aicardi syndrome, a rare disorder characterized by agenesis of the corpus callosum, lacunar chorioretinopathy and infantile spasms. It affects females (XX) and males diagnosed with Klinefelter syndrome (XXY), while exhibiting lethality in hemizygous males (XY) (Safaee et al. 2012). Several cases of CPPs, as well as cysts of the choroid plexus have been reported in more than 50% of Aicardi syndrome patients (Aicardi 2005). There are currently no known genes associated to this X-linked disorder. CPPs have also been identified in the context of the Von Hippel-Lindau (VHL) disease (Blamires & Maher 1992), an autosomal dominant cancer syndrome in which the tumour suppressor gene VHL is mutated and patients develop highly vascularized tumours in many organs, including the central nervous system and the eye (Haddad et al. 2013).

6. Genomic instability in cancer: aneuploidy and chromothripsis 6.1. Aneuploidy: causes and consequences

Aneuploidy is defined as a karyotype that is not a multiple of the haploid complement of an organism, arising from whole-chromosome gains or losses, and resulting in an unbalanced genome with different number of copies for genes on different chromosomes (Siegel & Amon 2012; Aylon & Oren 2011; Fang & Zhang 2011). Normal cells do not commonly tolerate aneuploidy, as chromosomal imbalances lead to deleterious changes in gene expression that trigger cell death. However, there are cases in which cells exhibiting aneuploid genomes are viable, such as in the congenital condition known as trisomy 21 or Down’s syndrome, and cancer. Aneuploidy is a prevalent characteristic found in cancer cells and may represent a

15 mechanism by which the ideal conditions that drive the selective pressure of highly aggressive cancer clones are generated.

6.1.1. Causes of aneuploidy

Aneuploidy is caused by errors in mitotic or meiotic cell division, which are in part caused by defects in mechanisms that ensure proper chromosomal segregation. Several causes for aneuploidy have been proposed and widely examined, including aberrant kinetochore- microtubule attachments, defects in the spindle assembly checkpoint (SAC), chromosome cohesion defects, aberrant centrosome amplifications, and endocytosis (Zhang et al. 2015; Duensing & Duensing 2005).

6.1.2. Consequences of aneuploidy

The effects of aneuploidy are complex, as the effect of aberrant gene copy number does not directly translate to a change in gene expression or protein levels. Several studies investigating cells with naturally occurring or artificially generated chromosome-specific aneuploidies have found that, in addition to gene expression changes observed on genes located in the aneuploid chromosome, a large percentage of misregulated genes were located in diploid chromosomes (Nicholson & Cimini 2015; Upender et al. 2004; Davidsson et al. 2013). Furthermore, a meta- analysis of gene expression data from aneuploid cells of different origin demonstrated recurrent patterns of upregulated and downregulated genes (Sheltzer et al. 2012) Studies that assessed the transcriptome of trisomic and tetrasomic cell lines have observed expression differences driven by copy number-dependent expression changes of genes on aneuploid chromosomes and recurrent transcriptomic patterns (Dürrbaum et al. 2014; Stingele et al. 2012). Gene pathway analysis of deregulated genes observed in trisomic cells from constitutional trisomy 8 mosaicism (CT8M) patients, who are prone to developing haematologic malignancies, revealed enrichment of pathways involved in cancer, genetic disorders and haematopoiesis pathways (Davidsson et al. 2013). Davidsson and colleagues also observed a unique methylation pattern in trisomy 8 cells with depletion of 5-hydroxymethylcytosine and global hypomethylation of chromosome 8, characteristics that mimic the inactivation of X- chromosomes in females (Davidsson et al. 2013). Aneuploidy appears to induce changes in DNA methylation, and may alter chromatin organization and condensation in an attempt to help cells

16 tolerate the changes associated with altered chromosomal dosage (Chango et al. 2006; Davidsson et al. 2013). Several factors may modify the effect that aneuploidy has on the transcriptome. For instance, not all genes in an aneuploid chromosome may be expressed in a particular tissue, not all cells types may be affected equally because of their own patterns of gene activation, genes may be altered indirectly through aneuploidy’s effect on the epigenome and since genes act in networks, copy number changes may affect gene expression in a pleiomorphic manner, altering the expression of associated genes on or off aneuploid chromosomes (Nicholson & Cimini 2015; Fang & Zhang 2011). The effect of aneuploidy on the function of a cell is further complicated by alterations to the proteome. Similar to gene expression, the over- or under-expression of proteins observed in aneuploid cells is not directly associated with the number of gene copies. Pathway analyses revealed an upregulation of proteins involved in DNA metabolism and cell growth, which were independent of specific aneuploid chromosomes (Upender et al. 2004; Gemoll et al. 2013). Another report investigated trisomic and tetrasomic cell lines and similarly found that overexpressed proteins were not encoded by genes located in aneuploid chromosomes, and that these cells exhibit dowregulation of DNA and RNA metabolism and upregulation of energy metabolism, membrane metabolism and lysososmal/autophagy pathways (Stingele et al. 2012). Unbalanced proteomes cause proteotoxic stress, which promote proteotoxicity, protein quality- control systems overload, and chaperone addiction (Oromendia & Amon 2014). Protein degradation via the proteasome and autophagy are the most common and efficient cellular responses to the proteotoxic effects of aneuploidy. Mutations that accelerate protein degradation, such as loss of function mutations in the deubiquitinating enzyme Ubp6, have been observed in aneuploid yeast cells, thus providing evidence that these mutations may increase tolerance for aneuploidy (Torres et al. 2010). In normal eukaryotic organisms aneuploidy is poorly tolerated, as it results in impaired fitness and abnormal development; however, aneuploidy is also known to drive neoplastic transformation and it is found in nearly all major human cancers (Siegel & Amon 2012; Fang & Zhang 2011; Oromendia & Amon 2014). Thus, cancer cells have evolved to develop increased tolerance for aneuploidy and its detrimental effects on the cell by accumulating aberrations in key genes such as the tumour suppressor, p53.

6.1.3. Aneuploidy and cancer

Aneuploidy is observed in greater than 90% of solid tumours and 75% of blood cancers (Weaver & Cleveland 2006). Studies have found that a higher degree of aneuploidy and karyotypic complexity correlates with increased tumour aggressiveness and metastasis, poor prognosis and drug resistance (Siegel & Amon 2012; Nicholson & Cimini 2015). Due to its dynamic nature, aneuploidy provides cancer cells the opportunity to adapt to a changing microenvironment through the selection of specific phenotypic features that favour growth under particular conditions (Nicholson & Cimini 2015). Tumour evolution due to the acquisition of a heterogeneous chromosomal profile has been reported in several studies that observed increased karyotypic heterogeneity as the cancer progressed and clonal divergence of metastases (Siegel & Amon 2012). No unique patterns of aneuploidy have been observed for different cancer types although a comprehensive study of 15,000 karyotypes and 62 cancer classes from the Mitelman database of cancer chromosomes discovered a high co-occurrence rate of chromosome gains with other gains and losses with other losses, suggesting a non-random pattern of chromosomal gains and losses in cancer (Ozery-Flato et al. 2011). Moreover, this analysis led to the understanding that chromosomal aneuploidy tends to consist of either a pattern of chromosomal gains or a pattern of chromosomal losses. It is not fully understood whether aneuploidy drives tumourigenesis. Several studies have shown that chromosomal instability increases tumourigenesis in mouse models (Weaver et al. 2007; Baker et al. 2009; Li et al. 2010) and inactivating mutations in genes regulating proper chromosomal segregation, such as STAG2, have been identified in a variety of human cancers (Solomon et al. 2011). However, aneuploidy has also been observed to inhibit tumourigenesis in-vitro and in-vivo, where tumour formation has been chemically or genetically induced (Weaver et al. 2007). In many cases, aneuploidy has been observed in the absence of cancer and seems inconsequential to tumour formation (Siegel & Amon 2012; Nicholson & Cimini 2015; Fang & Zhang 2011). In summary, the evidence available to date is equivocal as to whether aneuploidy drives cancer, and it may be that different factors modifying the effect of aneuploidy in living organisms may determine its role in tumour formation (Weaver & Cleveland 2006).

6.1.4. Aneuploidy and TP53

TP53 is a master regulator of cell cycle and apoptosis. A large number of aneuploid tumours harbour mutations in TP53, having lost the ability to regulate cell cycle and apoptosis in the presence of chromosomal instability and cellular stress (Duensing & Duensing 2005; Aylon & Oren 2011; Hanel & Moll 2012). Experimental studies have also shown an association between p53 and aneuploidy. Trp53 knockout mice are highly susceptible to the formation of a wide variety of spontaneous tumours, with lymphomas and sarcomas being the most frequently observed (Donehower et al. 1992). The athymic lymphomas observed were usually aneuploid, and increased chromosomal instability was observed in heterozygous (p53+/-) mice that had lost the remaining wildtype allele in relation to mice that had retained it (Venkatachalam et al. 1998). TP53 mutant tumour cells have also been shown to harbour tetraploid and polyploidy genomes (Galipeau et al. 1996). Transgenic mice harbouring a common missense mutation (R172H, corresponding to human hotspot mutation R175H) frequently develop aneuploid tumours with centrosome anomalies (Murphy et al. 2000). Moreover, this recurrent mutation has been shown to confer tumour- promoting functions, such as cell survival and increased metastatic potential, and is considered a gain-of-function mutation (Liu et al. 2000). In humans, studies of LFS families harbouring germline mutations in p53 show an increase in genomic DNA copy number variation (Shlien et al. 2008). Moreover, mutant p53 CPCs exhibit an elevated number of copy number aberrations relative to p53 wildtype tumours (Tabori et al. 2010). In contrast, several studies have demonstrated that inactivation of p53 function alone is not sufficient to directly cause aneuploidy (Lu et al. 2001; Bunz et al. 2002) and that not all p53 mutated cancers exhibit aneuploidy. Therefore, evidence suggests that although loss of p53 may not directly cause aneuploidy in all tumour cases, lack of p53 function may be required to survive lethal chromosomal imbalances facilitating cellular transformation (Duensing & Duensing 2005). In addition, the gain-of-function activities attributed to missense mutations in TP53 may be perturbing residual p53-independent genome-stabilizing mechanisms, that are otherwise observed in p53-null cells, thus promoting various forms of genomic instability (Hanel & Moll 2012).

19 6.2. Chromothripsis

For many years, cancer was thought to evolve gradually, with individual cells accumulating somatic mutations in oncogenes and tumour suppressor genes, and chromosomal alterations that provide certain subclones the right molecular environment for cancer initiation and progression (Fearon & Vogelstein 1990; Stratton et al. 2009). In 2011, Stephens and colleagues identified a novel mechanism of tumour initiation in which all mutations and alterations were obtained at once, in a singular catastrophic event where chromosomes were shattered and reassembled, introducing massive genomic rearrangements to the genome (Stephens et al. 2011). This phenomenon was named “chromothripsis” (Greek, chromos for chromosome; thripsis, shattering into pieces). The characteristic features of this phenomenon include complex crisscrossing rearrangements involving one or more chromosomes, multiple breakpoints within a chromosome or chromosome arm, and frequent oscillations between two or three copy number states (Stephens et al. 2011; Korbel & Campbell 2013). These rearrangements may promote oncogenic gene fusions and amplifications, disruption of tumor suppressor genes and dysregulation of gene expression, leading to malignant cellular processes without the gradual acquisition of mutations.

6.2.1. Chromothripsis and cancer

Using paired-end next-generation sequencing and genotyping microarrays for thousands of cell lines, Stephens et al. observed chromothripsis in 2-3% of cancers, including melanoma, small cell lung cancer, and glioma among others. They also found a particular enrichment in bone cancers (25%) (Stephens et al. 2011). Another large study examined copy number profiles of more than 8000 cancer genomes and found a lower overall frequency (1-2%) of chromothripsis with an enrichment in prostate cancer (5.6%) (Kim et al. 2013). Whole genome sequencing studies have reported a higher frequency of chromothripsis, with an incidence of up to 39% in glioblastoma (reviewed in (Kloosterman et al. 2014)). The presence of chromothripsis was associated with increased tumour aggressiveness and poor clinical outcomes in acute myeloid leukemia (AML), neuroblastoma, medulloblastoma, melanoma, multiple myeloma, and metastatic colorectal cancer, among others (Rausch et al. 2012; Kloosterman et al. 2014; Magrangeas et al. 2011; Kloosterman et al. 2011). This association appears to be a result of the sudden genomic instability generated by chromosome

20 shattering, of oncogene amplification and tumour suppressor gene deletions, and of the creation of fused genes with deleterious functions.

6.2.2. Chromothripsis and TP53

Chromothripsis has been implicated with paediatric brain tumours harbouring mutant TP53, establishing a link between loss of p53 activity and this tumour-promoting phenomenon. Rausch and colleagues reported high incidence of chromothripsis in patients diagnosed with the Sonic Hedgehog (SHH) subgroup of medulloblastoma (MB) and who concurrently exhibited germline mutations in TP53 (Rausch et al., 2012). Incidentally, the presence of chromothripsis in these tumours identified previously unknown LFS families carrying germline mutations in TP53. Moreover, when a cohort of LFS-associated tumours (n=65) exhibiting mutations or loss of heterozygosity of TP53 were examined, 23% (15/65) were found to be positive for chromothripsis. These included acute myeloid leukemia (AML, 10/23, 43% positive for chromothripsis), adrenocortical carcinoma (ACC, 1/8, 13%), osteosarcoma (1/8, 13%), rhabdomyosarcoma (1/4, 25%), neuroblastoma (1/1, 100%), and glioblastoma (1/1, 100%) (Data not published). The authors suggested that constitutional or early somatic TP53 mutations may predispose cells to chromothripsis, or facilitate this phenomenon, while helping affected cells survive following the massive genomic rearrangements. Lack of functional p53 will promote the maintenance and growth of chromothripsis-exhibiting tumour cells through mutant TP53- mediated mechanisms such as premature chromosome condensation, and failure to induce apoptosis or senescence due to impaired cell cycle control (Rausch et al. 2012).

7. Molecular background of choroid plexus tumours 7.1. TP53 and CPCs

CPCs from LFS families harbour mutations in TP53; however, TP53 mutations have also been observed in CPCs from patients who have no family history and are not associated with LFS. These mutations have been observed in up to 50% of CPCs; thus, TP53 is the most frequently mutated gene in these rare tumours (Tabori et al. 2010; Japp et al. 2015). TP53 mutations in CPC are most frequently missense mutations and found in the DNA binding domain. In our cohort of CPCs, 60% (35/58) harboured mutations in TP53 (Figure 1.4.A).

Fifteen (15/58, 26%) of these samples belonged to 12 patients with LFS carrying a germline mutation in TP53. Similar to other tumour types, TP53 mutant CPCs behave more aggressively than TP53 wildtype tumours and display greater genomic instability as measured by total structural variation (TSV) (Tabori et al. 2010). Patients with mutations in TP53, thus exhibit less favourable overall and event-free survival than patients with no mutations in TP53 (Tabori et al. 2010). A TP53 B

17p13 R248Q R248W 6 (3) (2)

R273C R273S 5 (3) (1) Frequency H193P G245S 4 R175H S241Y A276G 3 R280K V272M V203M E285V

R267W 2 R158H R213X Q331X R333H R337H V97D K132Q Y236H L265P Chr 17 1

TAD PRR DBD NLS TET REG 1 62 102 292 325 356 393

C 13-23 117-142 171-181 234-258 270-286 I II III IV V Conserved domains LSH

C176 C238 Zn D H179 C242 L1 L2 L3 Structural domains

E Residues involved K120 S241 R273 in DNA contact R248 A276 C277 R283 Adapted from Soussi & Beroud, 2001 Figure 1.4: Frequency of TP53 mutations in our cohort of CPCs (A) Location of TP53 on the short arm of chromosome 17 (17p13) (B) Frequency and location of mutations in the coding regions of TP53 in CPCs (n=34) with detailed description of the functional domains of TP53, where TAD: Transactivation domain, PRR: Proline-rich region, DBD: DNA binding domain, NLS: nuclear localization signal, TET: tetramerization domain, REG: regulatory domain (C) Location of the conserved domains of TP53 (D) Location of the structural domains of TP53, where L1, L2 and L3 indicate loops, LSH indicates a loop-sheet-helix structure. Zinc (Zn) is necessary for DNA binding (E) Residues involved in DNA contact are highlighted.

A transgenic mouse model of CPC requires the loss of TP53 in addition to the loss of RB1, for the development of CPC, indicating that TP53 loss is key in the development of these tumours (Tong et al. 2015). However in humans, CPCs also harbour functional wildtype TP53. Thus it

22 appears that loss of TP53 activity is one of many aberrations that may be involved in CPC tumourigenesis.

7.2. Chromosomal aberrations in CPTs

CPTs are characterized by a high degree of chromosomal imbalances. Most of the cytogenetic analyses conducted on CPTs have been reported in the form of single cases in larger series of brain tumours or single case studies (Summarized in Table 1.1) (Brassesco, Valera, Becker, et al. 2009; Neumann et al. 1993; Biegel 1999; Punnett et al. 1994; Donovan et al. 1994; Mertens et al. 1995; Roland & Pinto 1996; Griffin et al. 1992; Agamanolis & Malone 1995; Brassesco, Valera, Neder, et al. 2009; Bhattacharjee et al. 1997; Grill et al. 2002). Nonetheless, a staggering trend has been continuously observed in these tumours: CPPs are characterized by extensive chromosomal gains, and CPCs by chromosomal losses. In 1994, Donovan et al. conducted a retrospective FISH study on 9 CPPs, one the largest studies on CPTs at the time. Although this analysis only included probes to a few chromosomes (7, 11, 12, 15, 16, 17, 18 and 20), the authors observed increased hybridization signals in 6 CPP samples with consistent chromosome gain in chromosome 7 and 12 in CPPs. Using a comparative genomic hybridization (CGH) approach, Rickert and colleagues reported significant differences in the chromosomal imbalances observed between CPTs (Rickert et al. 2002). This impressively large cohort of tumour samples consisted of 34 CPPs and 15 CPCs with detailed clinical and follow-up data. Thirty-two of the 34 CPPs and all 15 CPCs exhibited chromosomal imbalances. CPCs demonstrated greater imbalances per tumour than CPPs (average of 10.7 and 7.3, respectively). When both tumour subgroups were compared, CPPs exhibited increased gains in 5q, 7q, 9q and 15q, as well as losses in 21q. CPCs exhibited increased gains in 1, 4q, 10, 14q, 20q and 21q, as well as losses in 5q and 18q. The authors reported that the number of aberrations overall was not significantly associated with survival, but the gain of 9p, and loss of 10q were significantly associated with longer survival in CPCs only. Not enough data has been collected to identify characteristic chromosomal aberrations in aCPPs; however, Brassesco and colleagues have reported an infant’s aCPP with a normal karyotype (Brassesco, Valera, Becker, et al. 2009), and an aCPP of a young male patient with polyploidy (Valera et al. 2009; Brassesco, Valera, Neder, et al. 2009).

Table 1.1: Choroid plexus tumour chromosomal instability reported in literature Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92), copyright (2015).

The latter study was conducted using FISH and CGH, and identified dicentric chromosomes as well as anaphase bridging, nuclear buds, and a ring chromosome, which suggested telomere alterations were involved in the etiology of this tumour. Recent molecular studies of large CPT cohorts have confirmed findings observed with more elemental karyotyping and CGH approaches. Using a novel genotyping microarray called molecular inversion probe (MIP) which allows for high density genotyping of FFPE-derived DNA, Ruland and colleagues assessed 26 CPCs and found a high frequency of chromosome wide losses on chromosomes 22 (81%), 6 (77%), 5, 16 19, (73%), 11 (69%), 3, 9, 17 (65%), 15, and 18 (62%), whereas frequent gains were observed in chromosomes 1 (31%), 12, 20 (27%), 2, 4, (19%), 7, 14, 19, and 21 (15%) (Ruland et al. 2014). Japp and colleagues also used a MIP approach to assess 62 CPT samples (21 CPPs, 23 aCPPs and 22 CPCs). In CPPs and aCPPs, gains were more prevalent than losses, and CPCs exhibited more recurrent losses but no gains. More than 50% of CPPs exhibited gains of chromosomes 7 (63%), 20 (58%), 9, and 12 (53%), while recurrent losses of chromosomes 2, 10, 13q, and 21q were found in 21% of these benign tumours. aCPPs only exhibited gains of chromosomes 8, 20 (55%), 7, 9, 12 (50%), and 11 (36%). CPCs only exhibited recurrent losses of chromosomes 3 (81%), 11p (76%), 6, 11q (71%), 16p, 22q (67%), 16q, 17p (62%), 5q, 8p, 13q, 15q (57%), 5p, 9, 10p, and 18 (52%). We recently examined a large cohort of 71 CPT samples (25 CPPs, 11 aCPPs and 35 CPCs) for chromosomal aberrations using the MIP genotyping microarray on FFPE-derived DNA, in addition to the SNP 6.0 genome wide array, a platform that has become the gold standard of genotyping microarrays and which only works with high-quality fresh-frozen material (Merino et al. 2015). We found an agreement with previous reports in that CPPs and aCPPs exhibited no recurrent losses but many gains throughout the genome; while CPCs exhibit recurrent losses and gains (refer to Chapter 2.4.2). We further observed distinct CPC subgroups exhibiting either broad chromosomal losses or gains and thus named hypodiploid and hyperdiploid CPCs, respectively (Merino et al. 2015) (Refer to Chapter 2.4.4). Chromosomal aberrations in CPTs are very frequent, although it is not yet clear whether these aberrations are driving tumour development, or appear as a result of tumour growth and progression.

25 7.3. Abnormalities in genes associated with CPT

The molecular genetic characteristics of CPTs have been reported in multiple studies. The most common genetic abnormality associated with CPCs is TP53 alterations. Germline and somatic mutations in TP53, a crucial tumour suppressor gene involved in DNA repair, cellular differentiation, and angiogenesis are associated with CPTs. The frequency of TP53 mutations is much higher in the carcinomas (~50%) than the papillomas (0-5%), which may account for the aggressiveness and poor outcome of these malignant tumours (Tabori et al., 2010, Merino et al., 2015). CPCs have also been associated with mutations in hSNF5/INI1, a member of the SWI/SNF ATP- dependent chromatin-remodeling complex (Kamaly-Asl et al. 2006). Germline mutations in this gene have been described in rhabdoid predisposition syndrome cases with aberrant protein expression. However, studies by Judkins et al. (Judkins et al. 2005) have revealed that these mutations may in fact distinguish CPCs from ATRTs, as most CPCs preserve protein expression of this gene. Tumours believed to be CPCs with hSNF5/INI1 mutations, isolated loss of 22q or lack of expression, are now diagnosed as ATRTs (Ruland et al. 2014; Judkins et al. 2005). Using a microarray-based approach, Hasselblatt et al. identified upregulation in TWIST-1, WIF1, TRPM3, BCLAF1 and AJAP1, as well as downregulation of IL6ST in CPP samples compared to normal choroid plexus epithelium obtained from autopsy tissue. The authors also reported that knockdown of Twist-1 in a rat choroid plexus epithelium cell line reduced proliferation and cell invasion, and led to the upregulation of CDKN1A, CFLAR and SERPINB2 cancer-related genes (Hasselblatt et al. 2006). A study reported the upregulation of NOTCH3 and NOTCH2 in CPPs relative to normal choroid plexus epithelia, however this PCR-based analysis was not able to find any significant associations with the small sample size it used (3 CPPs, 2 CPCs and 2 non-neoplastic choroid plexus samples) (Dang et al. 2005). In a cohort of 26 CPCs, Ruland and colleagues identified 17 genes within 9 regions of significant focal copy number gains on chromosomes 1, 12, 20, 4, 14 and 7, and 96 genes from 14 regions of significant focal copy number loss on chromosomes 6, 15, 16, 11, 19, 5, 9, 3 and 17. The study reported that PI3KR1, a phosphoinositide 3-kinase subunit, was located in a region of loss on 15q31.1, but copy number alterations affecting this gene has not reported in any other studies.

Japp and colleagues investigated genetic abnormalities in a group of 62 CPTs and identified that all subgroups exhibited recurrent (>50% frequency in each subgroup) focal gains of LAMB1 (7q22), TRPM3 (9q21.12), OTX2 (14q21-q22), and PLCB1 (20p12). When comparing the copy number profile between subgroups, ARL4A, a GTP-binding protein was frequently gained in CPPs and aCPPs, relative to CPCs, while losses in RAB6B, C3orf36, SLCO2A1, RYK (3q22), POLH, GTPBP2, MAD2L1BP, RSPH9, MRPS18A, VEGFA (6p21) and RBFOX1 (16p13.3) were more frequent in CPCs relative to CPPs and aCPPs. Our most recent findings, which integrated copy number and gene expression alterations in both human and mouse CPC revealed 3 oncogenes that initiated and maintained CPC development (Tong et al. 2015) (Refer to Chapter 3). These genes were TAF12, NFYC, and RAD54L, which are involved in DNA repair and epigenome remodelling capacity (Tong et al. 2015) (Refer to Chapter 3.4.8).

7.4. Telomere maintenance

Telomeres consist of tandem repeats of a hexameric G-rich sequence (5’-TTAGGG-3’) that cap the ends of eukaryotic chromosomes (Blackburn 2001; Blackburn 2005). Telomeres protect chromosomes from losing genetic information at the ends of each chromosome as the lagging strand shortens during each DNA replication cycle (Blackburn 2001). In protecting the coding regions at the ends of each chromosome, telomeres are shortened with every cell division reaching very small lengths in ageing cells and triggering a DNA damage response that results in cellular senescence. Telomerase is a ribonucleorprotein enzyme that synthesizes telomeres through the addition of hexameric repeats (Blackburn 2005). Telomerase consists of an RNA template sequence and the telomerase reverse transcriptase (TERT) protein, which synthesizes telomere sequences. Telomerase expression is tightly regulated in normal human cells, and its reduced expression in normal adult cells is reflected on the reduced telomere length observed on the chromosomes of ageing individuals (Blackburn 2005). By contrast, telomerase is active in normal human stem cells and germ line cells, while increased expression has been observed in almost all human tissue culture-immortalized cells and in a large number of cancers (Blackburn 2005; Elmore & Holt 2001). Telomerase is the most common way by which cancers maintain telomere length and elevated telomerase expression has been identified in approximately 90% of all human cancers

(Heaphy et al. 2011). Mechanisms of telomerase regulation are poorly understood. Several transcription factors have been observed to bind the TERT promoter, but their role in TERT regulation is unknown (Y. Zhang et al. 2012). The epigenetic regulation of TERT by DNA methylation of the CpG islands overlapping the TERT promoter has been investigated, yet the results yielded have been conflicting. DNA hypermethylation has been observed in the CpG-rich regions of the TERT promoter in both normal cells exhibiting no TERT expression as well as TERT-expressing cancer cells (Shin et al. 2003). A recent study by Castelo-Branco et al., found a 36bp region upstream of the transcription start site (UTSS) of TERT was hypermethylated in malignant tumours, but not in benign tumours or normal tissues (Castelo-Branco et al. 2013). UTSS hypermethylation was associated with increased TERT expression, worse tumour grade, and poor prognosis. This study identified hypermethylation of the UTSS region in CPCs, but not in CPPs or aCPPs. Moreover, this highly specific and sensitive test was able to distinguish a CPC sample that had been misdiagnosed as an aCPP and that exhibited UTSS hypermethylation (Castelo-Branco et al. 2013).

Telomere maintenance is considered one of the hallmarks of cancer (Hanahan & Weinberg 2011). Because of how important it is to maintain telomere length for tumour growth and progression, telomerase-negative tumours must exhibit an alternative, telomerase-independent mechanism. This mechanism is aptly called the Alternative Lengthening of Telomeres (ALT) and it generates telomeres via recombination. ALT has been described in a subset of tumours and some in vitro-immortalized cells (Elmore & Holt 2001; Mangerel et al. 2014; Bryan et al. 1995; Nabetani & Ishikawa 2011; Heaphy et al. 2011). A study assessing 6110 human samples revealed that 3.73% of all tumour specimens exhibited ALT, and the incidence of this phenotype was enriched in adult tumours of epithelial, neuroectodermal and mesenchymal origin, while not observed in adenocarcinomas arising from the prostate, pancreas, small intestine, stomach, or colon (Heaphy et al. 2011). A recent report assessed a large number of paediatric brain tumours (n=517) and found that ALT was highly prevalent in malignant subtypes, such as choroid plexus carcinomas, supratentorial high-grade gliomas, and diffuse intrinsic pontine gliomas, while it was rarely observed in medulloblastomas, atypical teratoid rhabdoid tumours, and the benign choroid plexus papillomas (Mangerel et al. 2014). Moreover, Mangerel and colleagues found that ALT is highly prevalent in TP53 mutant paediatric brain tumours, and that the ALT phenotype is associated with increased survival in patients whose tumours carry TP53 mutations,

28 compared to non-ALT tumours (Mangerel et al. 2014). Several reports have identified significant associations between the ALT phenotype and alterations in chromatin remodeling genes such as ATRX and DAXX, as well as, extensive genomic rearrangements, and an aberrant DNA damage response (Lovejoy et al. 2012; Flynn et al. 2015; Gocha et al. 2013; Wiestler et al. 2013; Marinoni et al. 2014; Liau et al. 2015).

8. Animal Models of CPT For many years, the absence of well-established human cancer cell lines or animal models of CPTs have limited the functional studies necessary to elucidate the mechanisms underlying the development of these brain tumours. An immortalized rat choroid plexus epithelial cell line (Z310) generated by simian virus large T antigen (SV40) transfection has been the only available cell line model to study choroid plexus tissue function to date (Zheng & Zhao 2002). SV40 viral particles sequester and inactivate p53 and Rb1 tumour suppressor genes, thus inducing immortality. SV40 particles have been previously identified in CPCs and other tumours from LFS patients who carried both a wildtype and a mutant copy of TP53 (Malkin et al., 2001). The authors suggested that the expression of the protein product of SV40, the viral T-antigen, may be responsible for the inactivation of the remaining wildtype p53 protein, leading to tumour formation in these patients (Malkin et al. 2001). Since the Z310 cell line was generated using SV40 particles, it does not provide a true molecular representation of normal choroid plexus epithelia and should not be used as a normal reference for candidate gene studies or comparative molecular analyses despite the fact that it its histopathology resembles that of choroid plexus epithelia in vivo.

Growing primary CPT cells in-vitro has proven to be a difficult task as cells stop replicating, achieving a quiescent phenotype after a few passages in various media. The creation of in-vivo models has also been challenging and mostly unsuccessful. Injecting a CPT cell suspension subcutaneously in NSG mice has failed to generate tumours after many months (personal communication with Peter Dirk’s and Richard Gilbertson’s laboratories).

29 8.1. Transgenic mouse models of CPP

Mouse models have been historically used in cancer research to investigate the biology and genetics of cancer in a live organism. More recently, these have been crucial in conducting in- vivo drug screening protocols. The physiological and molecular similarities between mice and humans, in addition to their small size, cost and breeding potential, and an entirely sequenced genome makes them an ideal species on which to model human cancer.

In 2005, Dang and colleagues generated CPP in mice by introducing constitutively active Notch3 into the periventricular cells of embryonic mice. By injecting a retrovirus expressing the intracellular domain of Notch3 into mouse ventricles, they activated Notch3 signalling in proliferating periventricular progenitor cells. CPPs formed in these mice within several weeks of birth, whereas mice injected with control retrovirus did not form any tumours (Dang et al. 2005). These tumours replicated the morphology and the histopathology of human CPPs; however, there have been no further reports validating the oncogenic potential of Notch3 in the development of CPP. Although the development of a mouse model for a benign tumour, such as CPP, is interesting and may elucidate mechanisms of tumour development, there is a greater urgency for the development of models that replicate malignant and highly aggressive tumours, as these will enable and expedite the development of targeted drug therapies potentially leading to increased patient survival.

8.2. Conditional allele CPC mouse model

A recent study has generated the first transgenic mouse model of CPC, which faithfully recapitulates the morphological, histological and molecular features of human CPC (Tong et al. 2015)(Refer to chapter 3). These mice exhibited loss of TP53, RB1 and PTEN and were generated using a Cre Recombinase system. This model has enabled the elucidation of the mechanisms driving CPC initiation and maintenance and identified 3 novel CPC oncogenes promoting tumour formation: RAD54L, NFYC and TAF12 (Refer to chapter 3).

30 8.3. Orthotopic zebrafish model of CPC

There are many advantages in developing and utilizing a zebrafish model of disease to complement current mouse models and human cell culture systems. This vertebrate species is readily genetically modified and the transparency of its embryos and the pigment-deficient adults allow for very rapid visualization of disease progression in a live animal. The use of zebrafish as a rapid primary drug and chemical screening tool has the potential to expedite drug development, while its use as an in-vivo model of disease will complement current mouse models to elucidate tumour biology, initiation and progression (White et al. 2013).

A recent study developed a unique approach to the generation of orthotopic models of paediatric brain tumours in zebrafish. Eden et al. adapted mouse models of brain tumours to grow as orthotopic xenografts in the brains of zebrafish and assessed their use as an intermediary platform for drug screening before starting screening experiments in mice (Eden et al. 2014). A transgenic mouse model of CPC (Tong et al. 2015) was conditioned to grow at 34oC and later transplanted into the brains of zebrafish that had been previously acclimatized to a 34oC environment. Acclimatization led to a successful engraftment of mouse CPC cells in adult zebrafish. The zebrafish CPCs recapitulated the histology and immunophenotype of the parent mouse CPC, which in part recapitulates the human histology and morphology of human CPC (Tong et al. 2015). Moreover, CPC dissemination along the spinal canal was observed in 11% of zebrafish implanted with CPC mouse cells. Histological analyses revealed that tumour masses observed in the spinal axis were independent, rather than direct extensions of the main brain tumour mass (Eden et al. 2014). Gene expression analyses revealed zebrafish CPC xenografts clustered more closely to parent tumours than other paediatric brain tumour types, such as glioblastoma and ependymoma, also grown in zebrafish. Although the degree of similarity between mouse and zebrafish CPC transcriptomes demonstrated a positive correlation, it was not considered significant (Eden et al. 2014). The zebrafish orthotopic model shared morphological and histological features with mouse CPCs, but did not significantly recapitulate the transcriptome profile of the parent tumours, demonstrating that it may not yet be a reliable intermediate platform for drug screening.

31 9. Summary Elucidating the biological mechanisms governing the development of choroid plexus tumours and their unique molecular profile has been a slow, but steady process, spanning many decades of work. Much progress has been achieved within the last decade thanks to the investigation of larger CPT cohorts and the advent of novel techniques, such as high-throughput analyses, which have facilitated the study of these rare tumours. Seeing as how the molecular landscape of CPTs had yet to be characterized, the bulk of this thesis work focused on refining recurrent molecular lesions in all CPT subtypes, thus generating the foundation on which further exploratory analyses revealed novel molecular subgroups with distinct clinical outcomes (Chapter 2). This comprehensive molecular framework has also facilitated the study of CPC ongogenesis in a newly-created mouse model of disease, using a unique cross-species approach that compared the copy number and gene expression profile of CPCs in both species (Chapter 3). Future directions should strive to continue building on our understanding of the molecular foundations of disease using complementary techniques, to better understand the etiology of CPTs using animal models, and focus on identifying how to best use this knowledge to guide the development of more tailored approaches to treat young CPT patients (Chapter 4).

Chapter 2: Defining the genomic landscape of choroid plexus tumours

This chapter has been published and is reproduced with permission from Clinical Cancer Research, The Lancet Oncology and Acta Neuropathologica

1. Acknowledgements This chapter is a modified and reformatted version of the manuscript entitled “Molecular Characterization of Choroid Plexus Tumours Reveals Novel Clinically Relevant Subgroups” published in Clinical Cancer Research (Merino et al. 2015). Under the supervision of Dr. David Malkin, my contribution to this work included designing the study, collecting, analyzing, and interpreting the data, drafting and critically revising the manuscript. I conducted all the bioinformatics analyses with the help of Adam Shlien, Stephen Mack, Vijay Ramaswamy, Marc Remke, and David Shih. Anita Villani ascertained and interpreted clinical records, Ana Novokmet organized the collection of samples and clinical information, Adam Shlien, Ruth Tatevossian, Uri Tabori, Richard Gilbertson and David Malkin provided guidance throughout the analysis, while our international collaborators supplied many of our tumour samples. Findings on CPTs from two other studies that assessed mechanisms of telomere maintenance were also included in this chapter. “Methylation of the TERT promoter and risk stratification of childhood brain tumours: an integrative genomic and molecular study” by Pedro Castelo-Branco et al., published in The Lancet Oncology (Castelo-Branco et al. 2013) and “Alternative lengthening of telomeres is enriched in, and impacts survival of TP53 mutant paediatric malignant brain tumours” by Joshua Mangerel et al., published in Acta Neuropathologica (Mangerel et al. 2014). My contribution to these manuscripts included the collection and extraction of nucleic acids from CPC samples, the annotation of clinical and molecular information for each of the samples, and the analysis and interpretation of clinical correlations to mechanisms of telomere maintenance.

33 2. Abstract Choroid plexus tumours (CPT) are rare intraventricular tumours most commonly affecting children under 2 years of age. The aetiology of these tumours is not fully understood and due to limited tumour samples, the biological mechanisms driving tumour development and progression are largely understudied. Our comprehensive study investigated the molecular alterations present in CPTs using a genome-wide high-throughput approach, in order to identify diagnostic and prognostic signatures that will refine tumour stratification and guide therapeutic options. One hundred CPTs were collected from our CPT cohort comprised of samples from various institutions across the world. Copy number (CN), DNA methylation, and gene expression signatures were assessed for 74, 36 and 40 samples, respectively. Mechanisms of telomere maintenance such as TERT promoter methylation and alternative lengthening of telomeres (ALT) were also investigated in our CPT cohort. Molecular subgroups were correlated with clinical parameters and outcomes. Unique molecular signatures distinguished choroid plexus carcinomas (CPCs) from choroid plexus papillomas (CPPs) and atypical choroid plexus papillomas (aCPPs). No significantly distinct molecular alterations between CPPs and aCPPs were observed. Allele-specific CN analysis revealed two novel subgroups in CPCs: hypodiploid and hyperdiploid CPCs. Hyperdiploid CPCs exhibited recurrent acquired uniparental disomy (aUPD) events. We observed 60% of CPCs harboured TP53 mutations, and we identified a high-risk group of CPC patients carrying 2 mutant TP53 copies. These patients exhibited significantly reduced 5-year event-free (EFS) and overall survival (OS) compared to patients with CPC carrying 1 mutant copy (OS: 14.3%, 95% CI 0.71%-46.5% versus 66.7%, 28.2%-87.8%, respectively, p=0.04; EFS: 0% versus 44.4%, 13.6%-71.9%, respectively, p=0.03). CPPs and aCPPs exhibited favourable survival. TERT promoter methylation was on average higher in CPCs than CPPs and aCPPs. ALT was observed in all tumour subtypes. Our data demonstrates that distinct molecular signatures distinguish CPCs from CPPs and aCPPs; however, molecular similarities among the papillomas suggest these two histological subgroups are indeed a single entity. A greater number of copies of mutated TP53 was significantly associated to increased tumour aggressiveness and a worse survival outcome in CPCs. CPCs were observed to maintain telomere length with different mechanism. Collectively, these findings will facilitate stratified approaches to the clinical management of CPTs.

34 3. Introduction 3.1. Choroid plexus tumours

Choroid plexus tumours are rare intraventricular neoplasms, accounting for up to 20% of brain tumours in children less than 2 years of age (Ogiwara et al. 2012; Addo et al. 2011; Lafay- Cousin et al. 2011). Three histological subgroups have been described according to their histopathological characteristics: choroid plexus papilloma (CPP, WHO grade I), atypical choroid plexus papilloma (aCPP, WHO grade II), choroid plexus carcinoma (CPC, WHO grade III). Long-term survival of CPP patients is favourable with surgical resection alone (>90%)(Addo et al. 2011). Conversely, CPC patients exhibit a dismal prognosis, with an overall survival of ~30% (Rickert & Paulus 2001; Tabori et al. 2010; Lafay-Cousin et al. 2010). Despite aggressive treatment protocols, including surgical resection and combination chemo- and radiation therapy (Lafay-Cousin et al. 2010; Greenberg 1999), clinical behaviour is variable and most of the few CPC survivors exhibit life-altering long-term cognitive and developmental deficits (Lafay-Cousin et al. 2010). aCPP, a recently described histological subtype, exhibits an intermediate degree of mitotic activity and outcome (Jeibmann et al. 2006; Wrede et al. 2009), however, some cases may be difficult to distinguish from CPC by histology alone.

3.2. Mutant TP53 and its association with choroid plexus carcinomas

Over 50% of CPC tumours carry somatic TP53 mutations, and TP53 mutant CPCs have been associated with increased genetic tumour instability and worse prognosis (Tabori et al. 2010). Germline TP53 mutations have also been observed in CPC patients. CPC is one of the hallmark cancers of the Li-Fraumeni syndrome (LFS), a familial cancer syndrome in which affected family members harbour a mutant copy of the TP53 tumour suppressor gene. Patients diagnosed with LFS associated CPCs (LFS-CPCs) do not exhibit worse survival outcomes than somatic TP53 mutant CPCs.

35 3.3. Telomere maintenance in choroid plexus tumours

One of the many functions of wildtype p53 is the regulation of cellular senescence by responding to telomeric erosion-induced DNA damage commonly observed during telomere crisis (Serrano et al. 1997; Junttila & Evan 2009). In the absence of functional p53, a permissive environment is formed where cancer cells that have acquired mechanisms of telomere elongation are selected. The most common mechanism of telomere elongation in cancer cells is telomerase activation. More than 90% of tumours exhibit telomerase activation and an overexpression of the gene telomerase reverse transcriptase (TERT), the catalytic unit of the telomerase complex (Castelo- Branco et al. 2013). A less common mechanism of telomere maintenance is the homologous recombination-dependent process known as alternative lengthening of telomeres (ALT), which has been observed in less than 4% of cancers (Mangerel et al. 2014). To date, the prevalence of telomerase-dependent or ALT processes in CPTs, and its effect in tumour prognosis has been poorly understood.

3.4. Chromosomal instability is prevalent in choroid plexus tumours

Cytogenetic studies of central nervous system (CNS) tumours using fluorescence in-situ hybridization (FISH) and comparative genomic hybridization (CGH) have revealed high chromosomal instability in more than 90% of CPTs analyzed (Table 1.1). Multiple gross chromosomal aberrations were observed to affect every single chromosome in the genome; with a trend exhibiting increased chromosomal gains in CPPs and aCPPs. Because of limited sample size and differences in methodology, previous studies have not been conclusive in identifying recurrent aberrant regions or patterns of chromosomal gains or losses characterizing each tumour type.

36 4. Results 4.1. Unsupervised clustering reveals significant segregation of CPCs from CPPs and aCPPs

Unsupervised clustering analyses performed with gene expression and methylation data revealed clear segregation of CPCs from CPPs and aCPPs (Figure 2.1). Non-negative matrix factorization (NMF) analysis of gene expression (Figure 2.1A) and methylation (Figure 2.1B) data demonstrated greatest difference between two subgroups (FDR-corrected p=2.54x10-7 and p=1.02x10-34, respectively), segregating CPCs from CPPs and aCPPs.

A Gene expression B DNA Methylation

Diagnosis Diagnosis TP53 status TP53 status 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 CP62.cpc.wt CP103.cpp.m CP47.cpp.wt CP104.cpp.wt CP4.cpp.wt CP55.acpp.wt CP125.cpp.wt CP121.acpp.w CP142.cpp.wt CP148.cpp.wt CP122B.acpp. CP122A.cpc.m CP138.acpp.w CP117.cpp.wt CP107.cpp.m CP136.acpp.w CP118.cpp.wt CP132.acpp.w CP114.cpp.wt CP110.cpp.wt CP108.cpp.wt CP57.cpp.wt CP65.cpc.wt CP133.cpp.wt CP146.cpc.wt CP41.cpcL.m CP113cpc.wt CP56.cpc.wt CP43.cpcL.m CP5B.cpcL.m CP3.cpc.m CP1.cpc.m CP127.cpp.wt CP139B.cpc.m CP141.cpc.m CP115.cpc.m CP58.cpc.wt CP52.cpc.m CP46.cpcL.m CP51.cpc.m CP51.T.cpc.m CP3.T.cpc.mu CP114.T.cpp. CP140.TB.cpc CP159.T.acpp CP115.T.cpc. CP108.T.cpp. CP139.TB.cpc CP141.TB.cpc CP66.T.cpc.w CP133.T.cpp. CP146.T.cpc. CP157.T.cpc. CP50.T.cpc.m CP59.T.cpc.w CP143.TB.cpc CP151.T.acpp CP107.TA.cpp CP136.T.acpp CP62.T.cpp.w CP57.T.cpp.w CP153.T.cpc. CP56.T.cpc.w CP148.T.cpp. CP119.T.cpp. CP55.T.acpp. CP110.T.cpp. CP106.T.cpp. CP138.T.acpp CP107.T.cpp. CP152.T.cpp. CP53.T.cpp.w CP104.T.cpp. CP111.T.cpp. CP47.T.cpp.w CP112.T.cpc. NMF group NMF group CP62.cpc.wt 2 CP51.T.cpc.m 1 CP103.cpp.m 2 CP3.T.cpc.mu 1 CP47.cpp.wt 2 CP114.T.cpp. 1 CP104.cpp.wt 2 CP140.TB.cpc 1 CP4.cpp.wt 2 CP159.T.acpp 1 CP55.acpp.wt 2 CP125.cpp.wt 2 CP115.T.cpc. 1 CP121.acpp.w 2 CP108.T.cpp. 1 CP142.cpp.wt 2 CP139.TB.cpc 1 CP148.cpp.wt 2 CP141.TB.cpc 1 CP122B.acpp. 2 CP66.T.cpc.w 1 CP122A.cpc.m 2 CP133.T.cpp. 1 CP138.acpp.w 2 CP146.T.cpc. 1 CP117.cpp.wt 2 CP157.T.cpc. 1 CP107.cpp.m 2 CP50.T.cpc.m 1 CP136.acpp.w 2 CP118.cpp.wt 2 CP59.T.cpc.w 1 CP132.acpp.w 2 CP143.TB.cpc 1 CP114.cpp.wt 2 CP151.T.acpp 2

k= 2 k= 3 k= 4 k= 5 CP110.cpp.wt 2 CP107.TA.cpp 2 k= 2 k= 3 k= 4 k= 5 CP108.cpp.wt 2 CP136.T.acpp 2 CP57.cpp.wt 2 CP62.T.cpp.w 2 CP65.cpc.wt 2 CP57.T.cpp.w 2 CP133.cpp.wt 1 CP153.T.cpc. 2 CP146.cpc.wt 1 CP56.T.cpc.w CP41.cpcL.m 1 2 CP113cpc.wt 1 CP148.T.cpp. 2 CP56.cpc.wt 1 CP119.T.cpp. 2 CP43.cpcL.m 1 CP55.T.acpp. 2 CP5B.cpcL.m 1 CP110.T.cpp. 2 CP3.cpc.m 1 CP106.T.cpp. 2 CP1.cpc.m 1 CP138.T.acpp 2 CP127.cpp.wt 1 samples samples CP107.T.cpp. 2 samples samples CP139B.cpc.m 1 samples samples samples samples CP152.T.cpp. 2 CP141.cpc.m 1 CP115.cpc.m 1 CP53.T.cpp.w 2 CP58.cpc.wt 1 CP104.T.cpp. 2 CP52.cpc.m 1 CP111.T.cpp. 2 CP46.cpcL.m 1 CP47.T.cpp.w 2 CP51.cpc.m 1 CP112.T.cpc. 2 CPC CPP aCPP TP53 mutant TP53 wildtype CPC CPP aCPP TP53 mutant TP53 wildtype samples samples samples samples samples samples samples samples Consensus matrix k=2; dataset= /xchip/gpprod/servers/genepattern/temp/Diana Merino_run9049295342431439568.tmp/CorrHuEx-7 −Var5K.gct -34 Cophenetic coef.= 0.9964p=2.54x10 CopheneticConsensus coef.= matrix k=2; 0.979 dataset= /xchip/gpprod−upload/servers/genepattern/users/DianaCophenetic coef.=Cophenetic Merino/uploads/tmp/run403039836211570815.tmp/dataset.filename/1/HuMe450 0.9684 coef.= 1 p=1.02x10Cophenetic Copheneticcoef.= 0.9382−ALLCPTs coef.=−5Kvar.gct 1 Cophenetic coef.= 0.9776 Cophenetic coef.= 0.9959

Cophenetic Coefficient Cophenetic Coefficient 1.00 1.00 0.98 0.94 0.96 0.88 Cophenetic correlation Cophenetic correlation 2.0 3.0 4.0 5.0 2.0 3.0 4.0 5.0

k k Cophenetic Coefficient= 9.96x10-1 Cophenetic Coefficient= 1.0

Figure 2.1: Unsupervised clustering of choroid plexus tumours (A) gene expression normalized intensities and (B) methylation Beta-values by non-negative matrix factorization (NMF) demonstrate significant segregation of CPCs (red) from CPPs (yellow) and aCPPs (light blue). No segregation was observed between CPPs and aCPPs. NMF was conducted using 5000 probesets with the largest median absolute deviation (MAD). This clustering algorithm identified the most significant measures of similarity (cophenetic coefficient) when the data were at k=2 (2 clusters). In the matrix, red represents the highest measure of similarity (1), while blue/purple represents the lowest measure of similarity (0). Any other colors within the matrix represent a spectrum of changing measures of similarity, from red to blue/purple. Colors: TP53 mutation status: Black: TP53 mutant, White: TP53 wildtype. Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

This significant molecular stratification was also observed using a smaller number of probesets for gene expression and methylation differences analyzed by PVCLUST (Figure 2.2). The

37 concordance between tumours stratified by gene expression and methylation was significant (Rand index=0.73, p<1.0x10-4) and revealed consistent molecular segregation of CPTs into unique molecular subgroups.

Figure 2.2: Unsupervised clustering analysis of CPTs showed significant segregation of CPCs from CPPs with aCPPs Subgroups delineated by red rectangles (p=0·05). Clustering was performed using PVCLUST and 1000 gene expression normalized intensities and methylation Beta values with the largest MAD. Colors: Red: CPC, Yellow: CPP, Light blue: aCPP. Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

4.2. CPCs are characterized by frequent chromosome-wide aberrations and frequent uniparental disomy events

Copy number analysis revealed widespread chromosomal instability in all tumour subgroups with multiple chromosome-wide gains and losses per tumour (Figure 2.3). A distinct signature characterized by increased frequency of chromosome-wide gains and losses was observed in CPCs (average 5.43 chromosomes gained and 5.65 lost per CPC), compared to increased

38 frequency of chromosome-wide gains but very few losses in CPPs and aCPPs (average 6.68 chromosomes gained and 0.32 lost per CPP, and 10.00 versus 0.25 per aCPP) (Figure 2.3).

CPC CPP aCPP Frequency of chromosome-wideCPC Frequency of chromosome-wideCPCCPP Frequency of chromosome-wideaCPPCPCCPP Frequency of chromosome-wide Frequencychromosome-wide of ofFrequency chromosome-wideFrequency of chromosome-wide FrequencyFrequency of chromosome-wide FrequencyFrequency ofof chromosome-widechromosome-wide $!!"# $!!"# $!!"# $!!"# $!!"# $!!"# $!!"# $!!"# $!!"# $!!"# CPCs in losses gains in CPCs $!!"# $!!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# 100% ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# 100% LossesLosses Gains Gains Losseschromosome-wide LossesLossesof Frequency Gains FrequencyGains of chromosome-wide Losseschromosome-wide LossesLossesof Frequency GainsGains FrequencyGains of chromosome-wide !"# !"# 10% 20% 30% 40% 50% 60% 70% 80% 90% 90% 80% 70% 60% 50% 40% 30% 20% 10% !"# !"# !"# !"# !"# !"# !"# !"# !"# !"# 0% 0% CPPs in losses gains in CPPs aCPPs in losses gains in aCPPs $!!"# $!!"# $!!"# $!!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# ,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"# $!"# %!"# &!"# '!"# (!"# )!"# *!"# +!"# ,!"# !"# !"# !"# %100 80 60 40 20 0 0 20 40 60 80 100% %100 8080 6060 4040 2020 00 0 20 40 60 80 100%100% %100 8080 6060 4040 2020 00 0!"# 0 2020 4040 6060 8080 100%100% 1 1 1 1 11 1 1 11 1 1 2 2 2 2 22 2 2 22 2 2 3 3 3 3 33 3 3 33 3 3 4 4 4 4 44 4 4 44 4 4 5 5 5 5 55 5 5 55 5 5 6 6 6 6 66 6 6 66 6 6 7 7 7 7 77 7 7 77 7 7 8 8 8 8 88 8 8 88 8 8 9 9 9 9 99 9 9 99 9 9 !"##$ !"##$ !"##$ !""# !""# !""# !""# !""# !"!# !"!# !"!# !"!# 1010 10 10 !"!# 10 10 10 10 10 !"!# 10 10 10 10 10 10 11 1 1 1 1 11 1 11 1 1 1 1 1 1 1 11 1 1 1 1 1 1 1212 12 12 12 12 12 12 12 12 12 12 12 12 12 1313 13 13 13 13 13 13 13 13 13 13 13 13 13 1414 14 14 14 14 14 14 14 14 14 14 14 14 14 15 15 1515 15 15 15 15 15 15 15 15 15 15 15 15 16 16 16 16 16 1616 16 16 16 16 16 16 16 16 17 17 17 17 17 1717 17 17 17 17 17 17 17 17 18 18 18 18 18 1818 18 18 18 18 18 18 18 18 19 19 19 19 19 1919 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 20 20 2020 20 20 20 20 20 21 21 2121 21 21 2121 21 21 21 21 21 21 21 21 21 22 22 2222 22 22 2222 22 22 22 22 22 22 22 22 22 X XX XXX X X XX X X X XX X XX n=37 n=37n=25 n=37n=n=2511 n=35 n=25 n=11 Losses GainsLosses GainsLosses

Lossesp-value Gains p-valuep-value p-value p-value CPC vs. CPP 1.82x10-11 CPCCPC vs. CP vs.P CP P 1.82x100.06 -11 CPCCPC vs. CP vs.P CP P CPC vs. CPP <1.0E-4 CPC vs. CPP 0.29 CPCCPC vs. vs.aCPP aCPP 3.64x10<1.0E-4-12 CPCCPCCPC vs. aCPP vsvs.. aCPP aCPP 6.083.64x10 0.03x10-3 -12 CPCCPC vs. aCPP vs. aCPP CPCPP vsP. vs.aCPP aCPP 0.590.49 CPCPPCP vsP. aCPP vsvs.. aCPP aCPP 0.150.590.32 CPPCP vsP. aCPP vs. aCPP

Figure 2.3: Characterization of recurrent chromosome-wide gains and losses for each tumour subtype (CPC: n=37, CPP: n=25, aCPP: n=11). P-values were calculated using the frequency of chromosome-wide copy number alterations per subgroup and the non-parametric Mann-Whitney test. Reprinted (adapted) by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

The frequency of chromosome-wide losses in CPCs was significantly greater than in CPPs and aCPPs (Two-tailed t-test p<1.00 x10-4). Allele-specific copy number analysis allowed us to investigate the allelic ratios in our samples. This technique revealed a striking pattern of copy number-neutral loss of heterozygosity. This phenomenon is commonly observed in cancer cells and may also be referred to as acquired uniparental disomy (aUPD), wherein a chromosome pair is homozygous, thus having two copies of the same allele (Tuna et al. 2009). CPCs exhibited frequent aUPD events with an average of 2.31 aUPD events per sample while the phenomenon occurred less frequently in CPPs and aCPPs (average 0.32 and 1 events per sample, respectively).

4.3. CPPs and aCPPs share similar molecular signatures which correlate with favourable survival outcomes

Analyzing CPPs and aCPPs independently from CPCs revealed a striking molecular similarity between the papilloma subgroups (Figure 2.4). Unsupervised clustering analysis demonstrated that CPPs and aCPP did not segregate according to differences in gene expression or methylation (Figure 2.4 A&B). Supervised analysis using the Wilcoxon rank-sum (WRS) test between CPPs and aCPPs revealed no significant differences in gene expression or methylation (Figure 2.4 C&D). Additionally, signatures of chromosomal instability characterized by recurrent chromosome-wide gains and very few losses were observed in both CPPs and aCPPs; no significant differences in the frequency of chromosome-wide gains and losses were observed (Two-tailed t-test p=0.32 and p=0.49, respectively) (Figure 2.4). There were no differences in age at diagnosis (Mann-Whitney test p=0.45) or ratio of males to females (Fisher’s Exact test p=0.31) between CPPs and aCPPs. Moreover, survival outcomes for CPP and aCPP patients were not significantly different (Refer to section 4.10.1, Figure 2.11).

Cluster dendrogramGene Expression with AU/BP values (%) ClusterDNA dendrogram Methylation with AU/BP values (%)

A B6 2.5 5 2.0 4 3 1.5 2 1.0 1 Height Height 0.5 0 0.0 Diagnosis CPP aCPP C D

Distance: correlation Distance: correlation Cluster method: ward Cluster method: ward

q=0.05 Significance (FDR-corrected p-value) 10 -log

Increased methylation Increased methylation in aCPP in CPP Difference (CPP-aCPP) E Diagnosis CPP ACPP Gender 1 2 2 2 2 2 2 2 2 2 2 2 2 2 G 3 2 2 2 2 4 2 2 2 2 2 5 2 2 2 2 2 2 2 2 2 2 2 2 6 2 2,1 7 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2,3 2 G G 8 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2,3,2 2 2 2 2 2,3 2 G 9 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 G 10 11 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 G 12 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 G G 13 2 2 2 2 2 2 2 2 2

GAINS 14 2 2 2 2 2 2 2 2 G 15 2 2 2 2 2 2 2 2 2 2 2 2 2 2 16 2 2 17 2 2 2 2 2 2 2 2 2 2 2 2 2 18 2,3 2 2 2 2 2 2 2 2 2 2 2 2,3 2 19 2 2 2 2 2 2 2 2 2 2 2,3 2 20 2 2 2 2,1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 G G 21 2 2 2 2 2 2,3 3,2 22 2 2 2 2 2 2 2 2 X 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2,3 1 2 3 4 5 6 7 8 9 CHROMOSOMAL ALTERATIONS CHROMOSOMAL 10 1 1 1 1 11 12 13 14 LOSSES 15 16 17 18 19 20 21 1 1 1 1 22 X Alteration No alteration Male Female

Figure 2.4: Unsupervised clustering of CPPs and aCPPs reveal no significant molecular differences between the two tumour subgroups (A) gene expression normalized intensities and (B) methylation Beta values demonstrate no segregation between CPPs (yellow) and aCPPs (light blue). Volcano plot comparing the number of significant differentially expressed genes (C) and significantly methylated regions (D) reveals no significant gene expression or methylation differences between the two subgroups (after FDR adjustment). Signatures of chromosomal instability (E), as measured by the number of chromosome-wide gains and losses in CPPs and aCPPs, were similar between the two subgroups and characterized by extensive chromosomal gains and very few

41 losses. Black squares represent chromosome-wide aberrations, while white squares represent unchanged chromosome copy number. Gender was also depicted (pink: female, purple: male). Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

4.4. Ploidy analysis reveals novel CPC subgroups with unique molecular alterations

Ploidy analysis revealed the presence of aneuploidy in 87% of tumours (Figure 2.5). CPPs and aCPPs exhibited ploidy greater than 2 (hyperdiploidy); however, CPCs exhibited a wide distribution of ploidy values, with two significantly distinct subgroups observed: hyperdiploid CPCs (average ploidy 2.76, range: 2.21-3.34) and hypodiploid CPCs (average ploidy 1.45, range: 1.25-1.71) (Mann-Whitney test, p<1.00x10-4). Ploidy in Choroid Plexus Tumors

3 ****

2 ! Ploidy

****p=0.0001 0 CPC aCPP CPP CPC CPP aCPP

CPC CPP TumoraCPP subtypeHyperdiploid Hypodiploid

Figure 2.5: Copy number analysis reveals aneuploidy on three CPT subgroups with CPCs demonstrating a binary distribution of ploidy values Scatter plot of CPCs (red), CPPs (yellow) and aCPPs (light blue) showing two distinct groups in CPCs: hypodiploid CPCs (dark blue), and hyperdiploid CPCs (green) exhibiting significant differences (Mann-Whitney test p=1.00x10-4). Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

Only three of 36 CPCs were diploid. Hypodiploid CPCs exhibited recurrent chromosome-wide losses and very few gains with an average of 12.70 chromosomes lost and 0.10 gained per tumour. Chromosome 3 was lost in all hypodiploid CPCs, with loss of chromosomes 6, 9, and 22 observed in 90% of tumours (Figure 2.6). Hyperdiploid CPCs exhibited a high frequency of chromosome-wide gains and almost no losses (average 12.22 and 0.22 chromosomes, respectively). Chromosomes 12, 7 and 1 were gained in more than 80% of hyperdiploid CPCs (Figure 2.6). In addition to a high frequency of chromosomal gains, hyperdiploid CPCs also exhibited aUPD more frequently than hypodiploid CPCs (average of 4.93 affected chromosomes

42 per tumour compared to 0.33 chromosomes in hypodiploid CPCs) (Fisher’s Exact test p<0.0001) (Figure 2.6). Moreover, significant enrichment in aUPD was observed in TP53 mutant hyperdiploid CPCs compared to hyperdiploid CPCs with wild-type TP53 (Fisher’s Exact test p<0.0001). aUPD was most frequently observed in chromosome 17, affecting 30% (10/33) of CPCs.

TP532 Chromosome 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Ploidy status CP118%T A CP133%T B CP110%T CP109%T CP106%T 1.00 CP107%TA CP107%T CP111%T 0.75 CP108%T CP148%T CP62%T 0.50 CP47%T CP57%T

CP103%T OS Probability

CPP CP120%T 0.25 CP125%T CP152%T CP45%T 0.00 p=0.03 CP53%T CP119%T 0 1 2 3 4 5 6 7 8 9 10 CP104%T Overall Survival (years) CP60%T Number at risk CP65%T CPC 35 30 25 21 19 14 12 10 9 7 6 CP142%TB CPP 19 17 12 10 5 4 4 2 2 1 1 CP116%T CP122%TA aCPP 11 8 5 4 4 3 3 2 2 1 1 CP122%TB CP137%T CPC CPP aCPP CP4%T CP132%T CP136%T CP121%T CP138%T aCPP CP55%T CP151%T1 CP159%T CP41-TA C CP51-T CP115-T CP141-TB 1.00 CP143-TB CP128-T CP157-T 0.75 CP66-T CP56-T CP146-T CP153-T 0.50 Copy Number Status CP158-T CP22-TA Whole-chromosome CP22-TB EFS Probability 0.25 CP39-T gain CP52-T Whole-chromosome CP54-T 0.00 p=4.90x10-3 loss CP1-T CP50-T 0 1 2 3 4 5 6 7 8 9 10 Acquired Uniparental CPC CP3-T Event-Free Survival (years) Disomy (aUPD) CP43-T CP46-T Number at risk CP144-TB CPC 35 21 13 12 10 8 7 4 3 3 2 Ploidy CP10-T CPP 19 16 12 10 5 4 4 2 2 1 1 Hyperdiploid CP147-TA aCPP 11 7 4 3 3 2 2 2 2 1 1 CP150-TB Hypodiploid CP2-T CP5-TC CPC CPP aCPP Diploid CP6-T CP64-T TP53 status CP139-TB CP140-TB Mutant CP63-T Wild-type CP59-T CP113-T

Figure 2.6: Genome-wide characterization of chromosome-wide aberrations and allelic abnormalities in 71 unique CPC samples Red squares represent chromosome-wide gains, blue losses, and teal acquired uniparental disomy (aUPD). White squares represent unchanged chromosome-wide copy number status. Chromosome-wide aberrations were defined as aberrations that encompass more than 75% of the chromosome. Ploidy: green: hyperdiploid, blue: hypodiploid, tan: diploid. TP53 status: black: mutant, white: wildtype. Reprinted (adapted) by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

43 We conducted GSEA to identify biological pathways and processes that are differentially expressed between hypodiploid and hyperdiploid CPCs. GSEA revealed enrichment in RNA processing, DNA replication and repair, and chromosome segregation in hyperdiploid CPCs. Hypodiploid CPCs exhibited enrichment in cellular metabolism, signalling and cell migration pathways, as well as leukocyte activation and proliferation (5% FDR, p<0.05) (Figure 2.7).

Figure 2.7: Gene set enrichment analysis (GSEA) comparing the expression of different biological pathways and processes between hypodiploid (red) and hyperdiploid (blue) CPCs (5% FDR, p<0·05). Visualization was obtained by Cytoscape and Enrichment Map, where each colored circle represents a node, and each line an edge. Nodes represent enriched gene sets annotated by similarity in biological pathway or process. Node size varies according to the number of genes within each gene set. This network has been curated to group similar sub-networks into larger and related networks, and to remove uninformative sub-networks. Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92) . Copyright 2015.

The patterns of enrichment observed suggest hyperdiploids are more proliferative than hypodiploid CPCs, and that hypodiploids appear to be undergoing a significant immune response. A greater understanding of these distinct enrichment patterns will elucidate the mechanisms underlying the progression of these molecularly distinct CPC subgroups (Figure 2.7). There were no significant differences in DNA methylation between subgroups, although this may be due to the low number of samples in the comparison groups (hypodiploid n=4, hyperdiploid n=9).

44 4.5. Transcriptomic and epigenomic signatures differ in CPCs according to TP53 status

Unsupervised clustering using gene expression and methylation data segregated CPCs into two significantly distinct clusters (p=0.05) (Figure 2.8).

!"#$%&'()&*)'+,'-.(/0%1(23456(7-"#&$(89: !"#$%&'()&*)'+,'-.(/0%1(23456(7-"#&$(89: A B !"#$%&'()&*)'+,'-.(/0%1(23456(7-"#&$(89: Gene Expression %"! DNA Methylation &"% %"! &"$ $"# &"# $"# &"' $"! &"! 9+6:/* $"! !"% ;-8<1, 9+6:/* !"# !"$ !"# !"# * # Diagnosis * TP53 status Ploidy status

p=0.05 p=0.05 TP53 mutant TP53 wild-type Hyperdiploid Hypodiploid Diploid Unknown 56)*478+2--80,,+'4*607 &'()*+,-.+*/012-34,1 78+,69:-4//:2..-)6,829 Figure 2.8: Unsupervised hierarchical clustering of gene expression and 1.00DNA methylation values demonstrate molecular C 1.00 ()*+,-./0-,1234/56.3 D 56)*478+2--80,,+'4*607 heterogeneity among CPCs associated with TP53 status &'()*+,-.+*/012-34,1

(A) Gene expression 0.75 normalized intensities and (B) DNA methylation Beta0.75 values using the PVCLUST algorithm, where subgroups are characterized by differences in TP53 status (mutant=black, wildtype=white). Red rectangles delineate statistically significant different groups0.50 (p=0·05). PVCLUST algorithm was conducted using0.50 1000 probesets with the largest median absolute OS Probability deviation (MAD). * shows sample with an uncharacterized intronic alteration,EFS Probability # shows samples with mutation in SH3- like/Proline-rich TP53 0.25domain. Colors: Diagnosis: Red: CPC, Yellow: CPP, Light0.25 blue: aCPP, TP53 status: Black: TP53 mutant,

White: TP53 wild-type, Ploidy: Green: hyperdiploid, Blue: hypodiploid,-3 Tan: Diploid, Grey: unknown. Reprinted by permission 0.00 p=3.80x10 0.00 p=0.07 from Clinical Cancer Research0 1 (Merino2 3 4et al.,5 2015;6 7 21(1):1848 9 10 -92) . Copyright 2015.0 1 2 3 4 5 6 7 8 9 10 Overall Survival (years) Event-Free Survival (years) Number at risk Number at risk TP53 mut CPC 21 19 14 10 9 6 4 3 3 1 1 TP53 mut CPC 21 13 6 5 4 3 2 1 1 1 1 TP53 wt CPC 14 11 11 11 10 8 8 7 6 6 5 TP53 wt CPC 14 8 7 7 6 5 5 3 2 2 1 Although CPCs did not segregateTP53 mut CPC accordingTP53 wt CPC to ploidy status, we TP53observed mut CPC twoTP53 wtclust CPC ers, which were significantly distinct according to TP53 status using DNA methylation data (Fisher’s Exact test, p=0.007), but did not reach significance using gene expression data (Fisher’s Exact test, p=0.089). LFS-CPCs did not segregate from the spontaneous CPCs, suggesting no significant transcriptomic or epigenomic differences were present in tumours arising from patients with an inherited TP53 mutation. This evidence also provides support to the theory that p53 alterations are an early event in spontaneous CPC development, resulting in similar transcriptomic and epigenomic profiles than LFS-CPCs. An in-depth analysis of the type of TP53 mutations revealed that a few samples, which appeared to be miscategorized by unsupervised clustering, had an uncharacterized intronic alteration (c.28-28G>A/wt) and mutations outside the DNA- binding domain (ie. c.290T>A in the SH3-like/Proline-rich domain), which may account for

45 differences in the transcriptomic (Figure 2.8A) and epigenomic (Figure 2.8B) signature of these samples.

4.6. Allele-specific copy number analysis determines number of mutant p53 molecules in CPCs

Mutations in TP53 were assessed in our cohort by Sanger sequencing. Sixty percent of CPCs (35/58) were mutant for TP53. Fifteen (15/58, 26%) of these samples belonged to 12 LFS patients carrying a germline mutation in TP53. Mutations in TP53 were observed in both hypodiploid and hyperdiploid CPCs, however, in our cohort, the frequency of TP53 mutations was significantly greater in hypodiploid CPCs (16/18, 89%) than hyperdiploid CPCs (7/15, 47%) (Fisher’s Exact test p=0.02). Diploid CPCs (n=3) were TP53 wildtype (Table 2.1).

No significant enrichment for LFS patients was observed in either hypo- or hyperdiploid subgroups. Combining TP53 sequencing results with allele-specific copy number status of chromosome 17 in 33 CPCs, we estimated the number of mutated copies of TP53. We found that 36.4% of CPCs (12/33) had 2 copies of mutant p53, 30.3% (10/33) had 1 copy of mutant p53 and 33.3% (11/33) had zero copies of mutant p53 (wildtype). CPCs with 2 mutant copies of p53 exhibited a homozygous TP53 mutation status, according to Sanger sequencing, in all but one tumour sample with a low aberrant cell fraction (46%), suggesting this sample was largely contaminated with normal cells. Seventy-five percent of samples with 2 copies of mutated p53 (9/12) exhibited aUPD in chromosome 17. Eighty-three percent of CPCs with 2 copies of mutated p53 (10/12) had missense mutations in the DNA binding domain, while 1 sample had a

46 missense mutation in the SH3-like/Proline-rich domain and the other sample, a splicing mutation. CPCs with 1 copy of mutated p53 had missense mutations in the DNA binding (9/10) and tetramerization (1/10) domains, and carried a single copy of chromosome 17, exhibiting LOH of the entire chromosome. Three samples exhibited a heterozygous TP53 mutation status by sequencing, which may be a result of normal cell contamination. Gene expression and methylation analyses revealed no significant differences among CPCs carrying 1 or 2 mutated copies of p53 because of the limited sample sizes (gene expression: 3 and 6 samples, respectively; methylation: 1 and 6 samples, respectively).

4.7. Methylation analysis of TERT upstream of transcription start site (UTSS) in CPTs reveals increased expression of TERT in a subset of CPCs.

Telomere maintenance is crucial for tumour growth and proliferation. The methylation of the promoter of TERT regulates the gene’s expression and has been associated with unfavourable patient outcomes (Castelo-Branco et al. 2013). Castelo-Branco and colleagues conducted a nested methylation approach where a region within the promoter of TERT that was significantly differentially methylated between highly malignant tumours and low-grade and normal tissue was identified. A pyrosequencing analysis of 36 base pairs in the region, containing 5 CpG sites that exhibited the highest specificity for malignant tumours was conducted with 190 samples. This region was called the Upstream of the Transcription Start Site (UTSS) and was found to be hypermethylated in tumour samples, but not hypermethylated in all normal tissues tested (p<0.0001; Figure 2.9A). The UTSS methylation status for 34 CPTs (14 CPCs, 12 CPPs and 8 aCPPs) was examined. All CPPs and most aCPPs had low UTSS methylation, with 2 aCPPs displaying high methylation levels (Figure 2.9B). One of these 2 samples was an aCPP that recurred after debulking surgery and treatment with chemotherapy, while the other was re- diagnosed as CPC after a thorough pathological examination. This discovery further supports the reliability of using UTSS methylation as a complementary diagnostic test accurately distinguishing between CPTs. CPCs displayed a binary distribution in UTSS methylation levels, with 36% (5/14) exhibiting high methylation, and 64% (9/14) exhibiting low methylation of UTSS. Interestingly, the two distinctly methylated groups did not exhibit any significant

47 difference in ploidy or TP53 mutation status, suggesting that methylation differences are independent on TP53 mutation status and dosage. A B

Figure 2.9: UTSS methylation status and survival in CPTs (A) Pyrosequencing analysis of the UTSS region in normal tissues and patient-derived tumour samples in the validation cohort Horizontal lines indicate means. (B) UTSS methylation in 34 choroid plexus tumours of different grade (C) Overall and progression-free survival of 45 patients with ependymoma. Vertical lines indicate censoring. UTSS=upstream of the transcription start site. Reprinted (adapted) by permission from The Lancet Oncology (Castelo-Branco et al., 2013; 14(6):534-42). Copyright 2013.

Alternative lengthening of telomeres (ALT) was assessed for a subset of this CPC cohort (Mangerel et al. 2014). Two of the 5 CPCs (40%) with hypermethylated UTSS were assessed for ALT and were found to be ALT negative. Eight out of the 9 CPCs with low UTSS methylation were assessed for ALT and 38% (3/8) were ALT positive, while 62% (5/8) were ALT negative. These findings suggest that ALT and UTSS hypermethylation are mutually exclusive events and that ALT appears to be a “last chance” mechanism for cell self-renewal and tumour cell survival. Favourable survival outcomes were observed in ependymoma patients whose tumours exhibited non-hypermethylated UTSS tumours compared to those whose tumours had hypermethylated UTSS (Figure 2.9C). Due to the limited number of CPC samples available for this study, survival

48 differences were not observed between patients whose CPCs exhibited hypermethylated UTSS and those whose tumours did not.

4.8. ALT is a recurrent feature in CPCs and not observed in CPPs

In a larger effort to study the prevalence and significance of ALT in childhood brain tumours, a cohort of 49 CPTs (25 CPCs, 18 CPPs and 6 aCPPs) were assessed for ALT with a C-circle analysis (CCA). CCA is a simple, reliable, and inexpensive method to test for ALT as it measures the presence of extra-chromosomal circular telomeric repeats that are tightly associated with ALT using very low quantities of DNA. After establishing the accuracy of CCA to detect ALT, dot blot assays were performed on CPTs, among other paediatric brain tumours, such as ependymomas, medulloblastomas, and gliomas (Mangerel et al., 2014). ALT was identified in 28% of CPCs (7/25), but not observed in any of the CPPs or aCPPs analyzed (0/24, 0%) (Figure 2.10A). CPCs were the malignant paediatric brain tumours with the highest incidence of ALT followed by supratentorial high-grade gliomas (11/50, 22%) and diffuse intrinsic pontine gliomas (9/48, 18.8%). None of the low-grade paediatric brain tumours exhibited ALT. A B

Figure 2.10: Frequency of alternative lengthening of telomeres (ALT) in CPTs and its association to CPC survival (A) Prevalence of ALT in different paediatric brain tumours and Li-Fraumeni associated cancers. ALT is prevalent in brain and non-brain cancers common in patients with Li-Fraumeni syndrome (B) Kaplan–Meier estimates of overall survival for TP53 mutant CPCs stratified by ALT status. Reprinted (adapted) by permission from Acta Neuropathologica (Mangerel et al., 2014; 128(6):853-62). Copyright 2014.

49 4.9. ALT is associated with TP53 mutations in paediatric brain tumours

Tumours that exhibited the greatest frequency of ALT, such as osteosarcomas, adrenocortical carcinomas, CPCs, and supratentorial high grade gliomas, are known to harbour somatic TP53 mutations. These brain tumours are commonly observed in families affected by the Li-Fraumeni syndrome; however, no association was found between ALT and LFS-CPCs harbouring germline TP53 mutations in our cohort. ALT was significantly associated with somatic TP53 mutations across the entire spectrum of paediatric tumours studied as 77% of ALT tumours also carried mutant TP53 (p = 7.32 × 10−8). One hundred percent (100%, 5/5) of ALT positive CPCs harboured a functionally deleterious TP53 mutation.

4.10. Patient outcomes

4.10.1. Patient outcomes for CPT histological subgroups

Analysis of clinical variables between all three CPT histological subgroups revealed no significant difference of age at diagnosis (Kruskal-Wallis test p=0.30), or ratio of males to females (Two-way ANOVA p=0.26). Similarly, there were no significant differences in the age of diagnosis (Mann-Whitney t-test p=0.80, and p=0.59) or the ratio of males to females (Fisher’s Exact test p=1.0, and p=1.0) according either to ploidy or p53 status in CPCs, respectively (Table 2.2).

CP103%T OS Probability

CPP CP120%T 0.25 CP125%T for CPCs was 56.3% (95% CI 36.5%-72.0%), compared to 92.9% (59.1%-99.0%) and 100% for CP152%T CP45%T 0.00 p=0.03 CP53%T CPPs and aCPPs, respectively (p=0.03) (Figure 2.11A). Only one patient with CPP died due to CP119%T 0 1 2 3 4 5 6 7 8 9 10 CP104%T Overall Survival (years) CP60%T Number at risk CP65%T complications from a concurrent diagnosis of ependymoma.CPC 35 Five30 -25year21 EFS19 for14 12CPCs10 was9 739.7%6 CP142%TB CPP 19 17 12 10 5 4 4 2 2 1 1 CP116%T CP122%TA aCPP 11 8 5 4 4 3 3 2 2 1 1 CP122%TB (95% CI 22.8%-56.2%), compared to 87.4% (58.1%-91.9%) and 70.0% (22.5%-91.8%) for CP137%T CPC CPP aCPP CP4%T -3 CP132%T CPPs and aCPPs, respectively (p=4.90x10 ) (Figure 2.11B) CP136%T CP121%T CP138%T aCPP CP55%TTP532 Chromosome 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Ploidy CP151%T1status CP118%T CP159%T A CP133%T CP41-TA B A C B CP110%T CP51-T CP109%T CP115-T CP106%T CP141-TB 1.00 1.00 CP107%TA CP143-TB CP107%T CP128-T CP111%T CP157-T 0.75 0.75 CP108%T CP66-T CP148%T CP56-T CP146-T CP62%T 0.50 CP47%T CP153-T 0.50 CP57%T Copy Number Status CP158-T CP22-TA

CP103%T OS Probability EFS Probability CPP CP120%T Whole-chromosome CP22-TB 0.25 0.25 CP125%T CP39-T CP152%T gain CP52-T CP45%T Whole-chromosome CP54-T 0.00 p=0.03 0.00 p=4.90x10-3 CP53%T loss CP1-T CP119%T CP50-T 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Acquired Uniparental CPC CP3-T CP104%T Overall Survival (years) Event-Free Survival (years) Disomy (aUPD) CP43-T CP60%T Number at risk Number at risk CP65%T CP46-T CP144-TB CPC 35 30 25 21 19 14 12 10 9 7 6 CPC 35 21 13 12 10 8 7 4 3 3 2 CP142%TB Ploidy CPP 19 17 12 10 5 4 4 2 2 1 1 CP116%T CP10-T CPP 19 16 12 10 5 4 4 2 2 1 1 CP147-TA aCPP 11 8 5 4 4 3 3 2 2 1 1 CP122%TA Hyperdiploid aCPP 11 7 4 3 3 2 2 2 2 1 1 CP122%TB CP150-TB Hypodiploid CP2-T CP137%T CPC CPP aCPP CPC CPP aCPP CP4%T Diploid CP5-TC CP132%T CP6-T Figure 2.11: Kaplan-Meier curves depicting overall and event-free survival estimates of CPT patients by diagnosis CP136%T CP64-T CP121%T TP53 status CP139-TB (A) Overall and (B) event-free survival estimates. Statistical values were obtained with the Log-rank (Mantel-Cox) test. Red: CP138%T CP140-TB aCPP Mutant CP55%T CP63-T CPC, yellow: CPP, light blue: aCPP. Reprinted (adapted) by permission from Clinical Cancer Research (Merino et al., 2015; CP151%T1 Wild-type CP59-T CP159%T CP113-T 21(1):184-92. Copyright 2015. CP41-TA C CP51-T CP115-T CP141-TB 1.00 CP143-TB CP128-T CP157-T 0.75 CP66-T CP56-T CP146-T CP153-T 0.50 Copy Number Status CP158-T CP22-TA Whole-chromosome CP22-TB EFS Probability 0.25 CP39-T gain CP52-T Whole-chromosome CP54-T 0.00 p=4.90x10-3 loss CP1-T CP50-T 0 1 2 3 4 5 6 7 8 9 10 Acquired Uniparental CPC CP3-T Event-Free Survival (years) Disomy (aUPD) CP43-T CP46-T Number at risk CP144-TB CPC 35 21 13 12 10 8 7 4 3 3 2 Ploidy CP10-T CPP 19 16 12 10 5 4 4 2 2 1 1 Hyperdiploid CP147-TA aCPP 11 7 4 3 3 2 2 2 2 1 1 CP150-TB Hypodiploid CP2-T CP5-TC CPC CPP aCPP Diploid CP6-T CP64-T TP53 status CP139-TB CP140-TB Mutant CP63-T Wild-type CP59-T CP113-T 51 4.10.2. TP53 status has a significant effect on the overall survival of CPC patients

OS or EFS estimates were not significantly different between CPC patients exhibiting a hyper- or hypodiploid tumour genome (p=0.82, p=0.94, respectively) (Figure 2.12 A&B). TP53 status had a significant effect on the OS of our CPC cohort (Log-rank test p=3.8x10-3) with TP53 wildtype CPC patients exhibiting more favourable outcomes than TP53 mutant patients. EFS estimates were not significantly different between TP53 mutant and wildtype CPC patients (p=0.07). (Figure 2.13 A &B).

A 1.00 B 1.00

0.75 0.75

0.50 0.50 OS Probability EFS Probability 0.25 0.25

0.00 p=0.68 0.00 p=0.94 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Overall Survival (years) Event-Free Survival (years) Number at risk Number at risk Hypodiploid CPC 13 12 10 8 7 5 4 3 3 1 1 Hypodiploid CPC 13 10 4 4 3 2 1 0 0 0 0 Hyperdiploid CPC 12 9 7 5 4 4 3 2 2 2 1 Hyperdiploid CPC 12 4 4 3 2 2 2 1 1 1 0

Hypodiploid CPC Hyperdiploid CPC Hypodiploid CPC Hyperdiploid CPC

Figure 2.12: Kaplan-Meier curves depicting overall survival and event-free survival estimates of CPCs by ploidy (A) Overall and (B) event-free survival estimates. Statistical values were obtained with the Log-rank (Mantel-Cox) test. Green: hyperdiploid, blue: hypodiploid. Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92. Copyright 2015.

A B

Figure 2.13: Kaplan-Meier curves depicting overall survival and event-free survival estimates of CPCs by TP53 mutation status

52 (A) Overall and (B) event-free survival estimates. Statistical values were obtained with the Log-rank (Mantel-Cox) test. Dashed line: TP53 mutant CPC, solid line: TP53 wildtype CPC. Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92, copyright (2015).

4.10.3. Number of mutant copies of TP53 predicts overall and event- free survival outcomes in CPC patients

Investigating survival differences according to the number of mutant copies of TP53 in CPCs revealed a significant reduction in OS (Log-rank test 2 copies vs. 1 copy, p=0.04; 2 copies vs. 0 copies, p<1.0x10-4), and EFS (Log-rank test 2 copies vs. 1 copy p=0.03; 2 copies vs. 0 copies p=0.003) in patients harbouring a greater number of mutant TP53 copies. The estimated OS of patients with CPCs harbouring wildtype TP53 (zero copies) was 88.9% (95% CI 43.3%-98.4%) and EFS, 66.5% (32.9%-86.1%). Patients with CPCs harbouring a single copy of mutant TP53 exhibited an estimated OS of 66.7% (28.2%-87.8%) and EFS of 44.4% (13.6%-71.9%), while patients with CPCs harbouring two copies of mutant TP53 showed an OS of 14.3% (0.71%- 46.5%) and EFS of 0% (Figure 2.14). A B

Figure 2.14: Kaplan-Meier curves depicting overall and event-free survival estimates of CPC patients by number of mutated copies of TP53, as estimated by Sanger sequencing and allele-specific copy number analysis (A) Overall and (B) event-free survival estimates. Statistical values were obtained with the Log-rank (Mantel-Cox) test. Red: CPCs with no mutant TP53, green: CPCs with 1 mutant TP53 copy, blue: CPCs with 2 mutant TP53 copies. Reprinted (adapted) by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92). Copyright 2015.

4.10.4. Mutant TP53 CPC patients exhibit no significant differences in survival outcomes according to ALT status

The clinical outcome of CPC patients carrying a mutant TP53 was compared according to ALT status. Although we had a limited number of clinical information, we were able to observe a non-

53 significant trend for longer overall survival in TP53 mutant CPC patients exhibiting ALT. Five year overall survival was 67 ± 19 % for ALT positive tumours and 27 ± 13 % for ALT negative tumours (p=0.07, Figure 2.10B). Seventy-five percent (75%, 3/4) of ALT positive, mutant TP53 samples for which we had sufficient clinical information, carried 2 copies of mutant TP53, which may be associated with reduced survival.

5. Discussion Our study is the first and largest comprehensive investigation of the molecular alterations found in CPTs, demonstrating that the molecular profile of CPCs is significantly distinct from that of CPPs and aCPPs, and that the papillomas are not significantly distinct from each other. In addition, using an innovative allele-specific approach in combination with TP53 sequencing, we identified a subgroup with a particular poor prognosis in TP53 mutant CPC patients exhibiting aUPD in chromosome 17, and who as a result had an elevated number of mutated copies of p53. This study provides evidence for the crucial role of molecular stratification as a tool to improve the clinical management of patients with CPT.

Atypical CPPs are currently distinguished from CPPs by histopathology, where aCPPs exhibit increased mitotic activity (Louis et al. 2007); yet, survival outcomes for both CPPs and aCPPs are comparably favourable. Standard of care for these tumours consists of surgical resection with very few aCPP cases requiring adjuvant chemotherapy. In our cohort, all aCPP patients for which we had clinical history (6/11), were treated with surgical resection alone, yet demonstrated favourable survival comparable to CPPs. We suggest that the benign phenotype of aCPPs may reflect the molecular characteristics it shares with CPPs, including very few chromosome-wide losses, and similar gene enrichment patterns and methylation signatures. The data lend support for the conservative management of aCPP patients with surgical resection followed by observation. Our findings also revealed that copy number, gene expression and methylation profiles were significantly distinct between the papillomas and CPCs, indicating the unlikelihood that CPPs or aCPPs progress onto CPCs by the acquisition of a few additional aberrations. Although a few studies have reported on progression from papillomas to CPCs (Jeibmann et al. 2007; Dhillon et al. 2013), we believe this unlikely scenario may have been the result of a heterogeneous tumour sample harbouring co-existing CPC and papilloma cells. Analyzing tumour heterogeneity in

54 CPTs, will be necessary in order to identify benign tumours more likely to recur with an aggressive phenotype.

Wide variability in clinical outcome has been observed among CPC patients despite the use of similar treatment protocols (Lafay-Cousin et al. 2011; Gozali et al. 2012; Bettegowda et al. 2012). Our findings demonstrate that the molecular heterogeneity of CPCs may be driving this clinical variability. Extensive chromosomal alterations were recurrent in CPCs. An allele-specific copy number approach allowed us to identify aneuploidy in 91% of CPCs, and distinguish between hypodiploid CPCs, exhibiting numerous chromosomal losses, and hyperdiploid CPCs, exhibiting numerous chromosomal gains and concurrent aUPD. Our findings uncovered that chromosomal instability is a common mechanism involved in CPC development; however, further examination of the molecular differences driving hypo- and hyperdiploid development will identify distinct mechanisms responsible for tumour progression. Ploidy was not significantly associated with age at diagnosis, or patient survival; nonetheless, we identified that hyperdiploid CPCs were significantly enriched in chromosomes exhibiting aUPD. A recent study reported similar subgroupings in CPCs, where a higher frequency of chromosomal losses were observed in younger children and chromosomal gains in older children, and loss of 12q was associated with shorter survival (Ruland et al. 2014). In our cohort we found no significant correlations between patient age and CPC subgroups or TP53 status. Moreover, survival differences were identified only when TP53 copy number and mutation status were examined concomitantly.

Arising from somatic recombination errors during mitosis, aUPD is an important mechanism leading to loss of heterozygosity with an unaffected copy number, and is therefore associated with the enrichment of chromosomes or regions harbouring pre-existing mutations, specific promoter methylation patterns, and focal deletion of genes (Tuna et al. 2009). In our study, we observed aUPD affect entire chromosomes in all tumour subgroups, however aUPD was most frequent in CPCs harbouring TP53 mutations. We identified chromosome 17 to be the most frequent site affected by aUPD, and that 90% of CPCs exhibiting aUPD of chromosome 17 harboured a mutation in TP53, increasing the number of mutant p53 copies to 2 in these tumours. Our findings suggest aUPD is a mechanism by which CPCs accumulate deleterious aberrations,

55 such as TP53 mutations, while retaining the normal function of other genes due to an unaffected chromosome copy number.

Focusing on the known association between p53 mutations and CPCs, we identified that the number of mutated copies of p53 was significantly associated with patient survival. Our findings support the concept that in addition to the loss of tumour-suppressive activity of p53, mutant TP53 also acquires oncogenic activities that promote CPC development. The gain of function (GOF) properties of mutant p53 include cellular invasion, proliferation, genomic instability, and polyploidy, among others (Reviewed in (Muller & Vousden 2014)). Because of increased GOF activity, in addition to a complete loss of the tumour suppressor functions of p53, an elevated number of mutant p53 copies could result in an aggressive phenotype associated with decreased survival as we observed in our high risk CPC patient cohort. Since we did not assess the mutation status of other cancer genes in chromosome 17, we cannot infer that the number of mutant copies of p53 is the only aberration at this locus driving CPC development and tumour aggressiveness in the high risk CPC patient cohort. However, the significant dose-dependent correlation observed with overall and event-free survival in CPC patients indicates a role for p53 GOF activity in CPC aggressiveness.

TP53 mutations alone do not drive chromosomal instability in CPCs as TP53 wild-type tumours also exhibit high levels of chromosome-wide gains and losses. However, we have demonstrated that TP53 mutations are associated with changes in gene expression and methylation patterns that may result in increased tumour aggressiveness and may elucidate the different clinical outcomes observed. Alterations affecting the p53 pathway, either upstream or downstream of p53, may generate the molecular background necessary for CPC development, and should be investigated further.

Recurrent lesions, such as the chromosome-wide gains of chromosome 1, which was not only recurrently gained but also the least frequently lost in CPCs, chromosome 12, and chromosome- wide loss of chromosome 3, may also be contributing to CPC’s unique genotype and would need to be further investigated in order to identify unique targets for effective therapies.

56 Methylation of TERT promoter and ALT analyses revealed that CPC development is not reigned by a single telomere maintenance mechanism. Both methylation of the UTSS of TERT and ALT were observed in CPCs and these mechanisms were mutually exclusive. Determining which maintenance mechanism is active in CPCs will also guide the creation of tailored therapeutic options that target telomere maintenance in CPCs.

6. Conclusion Our study demonstrates that investigating the molecular characteristics of CPTs is crucial to further refine the molecular stratification of patients in order to improve patient care. We suggest that the prognostic significance of TP53 mutation and copy number status in CPCs be validated prospectively in future cooperative clinical trials. Validation of these data in future prospective studies will inform risk stratification of CPC patients, and set the framework for future treatment intensification for high-risk patients. In this study, we used an integrative molecular approach to characterize the genomic, transcriptomic and epigenomic landscape of the largest cohort of CPTs to date. The information derived from these analyses creates a molecular foundation on which to develop approaches to improve the clinical management of this devastating disease.

7. Materials and Methods 7.1. Patients and sample preparation for microarray study

CPT samples and/or clinical data were collected from several national and international institutions (see appendix for a list of participating institutions) in accordance with each institution’s Research Ethics Board. Informed consent was obtained from the parents/legal guardians of all patients.

We studied 100 unique tumour samples (58 CPC, 30 CPP and 12 aCPP) from 91 paediatric patients (ages 0.03-16.50 years old) for which TP53 sequence data were available (Table 2.3). Pathological review of CPTs was conducted by Cynthia Hawkins, David W. Ellison, and Brent T. Harris when samples were available. In all other institutions, expert neuropathologists critically examined each case. In fifteen CPC cases, immunohistochemical analysis of hHF5/INI1 was conducted and revealed immunopositivity excluding the diagnosis of atypical

57 teratoid/rhabdoid tumours (ATRT). Nucleic acids were derived from fresh frozen (n=75), optimal cutting temperature compound (OCT) (n=9), and formalin-fixed paraffin-embedded (FFPE) (n=12) samples. We received isolated tumour DNA from 4 samples. Twenty-two nucleic acid samples exhibited suboptimal quality and/or quantity, leaving 78 high-quality samples from 73 patients for analysis (Figure 2.15).

Detailed clinical data were obtained for 68 patients. Tumour DNA was extracted using standard phenol-chloroform extraction from fresh-frozen samples, and the RecoverAll™ Total Nucleic Acid Isolation Kit for FFPE (Ambion, Carlsbad, USA) from FFPE samples. Total RNA was isolated from fresh-frozen samples using the TRIzol method (Invitrogen, Carlsbad, USA) according to the manufacturer’s instructions.

CPT Samples (with TP53 status) n=100 CPC=58 CPP=30 aCPP=12

Clinical Data & Biological Specimens

Clinical data Survival Optimal Non-optimal not available (OS & EFS) quality & quality or amount amount

n=32 n=68 n=78 n=22

Gene Expression Methylation Genotyping Genotyping Affymetrix Human Exon Illumina Infinium Human Affymetrix SNP6.0 Affymetrix OncoScan 1.0ST Methylation 450K Genome-wide FFPE Express 2.0 Beadchip CPC=16 CPC=15 CPC=23 CPC=15 CPP=16 CPP=16 CPP=23 CPP=3 aCPP=8 aCPP=5 aCPP=9 aCPP=2

Figure 2.15: Sample flow-chart for CPT study Details on number of biological samples and clinical data obtained from numerous international institutions and the distribution of samples in each microarray. Reprinted by permission from Clinical Cancer Research (Merino et al., 2015; 21(1):184-92). Copyright 2015.

58 7.2. Patients and sample preparation for TERT methylation study

For the promoter of TERT methylation study, DNA material from 34 CPTs (19 CPCs, 12 CPPs and 8 aCPPs) was used for hybridization with the Human Methylation450 BeadChip (Illumina, San Diego, CA, USA). Sequenom methylation analysis of the TERT promoter was used at McGill University and Génome Québec Innovation Centre (Montreal, QC, Canada), as previously described. Tert expression was assessed by real-time PCR in 40 CPT samples for which DNA methylation was performed (Castelo-Branco et al. 2013).

7.3. Patients and sample preparation for ALT Study

DNA material from 49 CPTs (25 CPCs, 18 CPPs and 6 aCPPs) was used for the determination of ALT status using the C-circle assay (CCA). CCA was performed using a dot blot assay and qPCR, and when available, ALT detection was validated using the telomere restriction fragment (TRF) assay (Mangerel et al. 2014).

7.4. TP53 sequencing

Sequencing of the coding region of TP53 (exons 2-11) was performed in the molecular diagnostic laboratory at The Hospital for Sick Children (Toronto) by direct Sanger sequencing of whole genome DNA as previously described (Tabori et al. 2010).

7.5. Microarray processing & bioinformatics analysis

Forty RNA samples were hybridized to GeneChip® Human Exon 1.0ST gene expression microarrays (Affymetrix, Santa Clara, USA), and 36 DNA samples were hybridized to Illumina® 450K Infinium methylation bead arrays (Illumina, San Diego, USA) as per manufacturer’s instructions. An initial set of 55 tumour DNA samples was hybridized to Genome-Wide Human SNP Array 6.0 (Affymetrix), while an independent set of 20 tumour DNA samples was hybridized to Affymetrix OncoScan™ FFPE Express 2.0 arrays. One technical replicate was included in both genotyping platforms and analyzed for copy number call consistency.

59 Partek® Genomics Suite™ 6.5 (PGS) (Partek Inc. St. Louis, USA) and BioDiscovery Nexus Copy Number software (Discovery Edition 7.0, BioDiscovery, Hawthorne, USA) were used for copy number analysis as previously described (Tabori et al. 2010) (see Appendix-Supplemental Experimental Procedures). Copy number changes encompassing more than 75% of the chromosome were called whole-chromosome aberrations. Allele-specific copy number analysis of tumours (ASCAT) was performed in R as previously described (Van Loo et al. 2010) and verified by the BioDiscovery Nexus software. Tumour ploidy, where hypodiploid <1.90 and hyperdiploid >2.10, heterogeneity, and allelic imbalances were inferred from the output. ASCAT failed to resolve the ploidy of two samples with very low aberrant fraction, so these were excluded from further analysis. Clustering of gene expression and methylation data were investigated in R (version 3.0.1) by unsupervised hierarchical clustering (uHCL), non-negative matrix factorization (NMF), and PVCLUST algorithms. Differential gene expression analysis was conducted with PGS 6.5 (see Appendix-Supplemental Experimental Procedures). Gene set enrichment analysis (GSEA) was performed as previously described (Subramanian et al. 2005) and its visualization was obtained by Cytoscape and Enrichment Map using an in-house curated database containing freely available NCI, KEGG, PFAM, Biocarta and GO databases as described in Witt et al. (Witt et al. 2011) Differences in DNA methylation status were analyzed with the Illumina® GenomeStudio software (see Appendix-Supplemental Experimental Procedures). All probesets were annotated according to the human genome build hg19 (GRCh37). Microarray data can be accessed from GEO (GSE61044)

7.6. Statistical analysis

Statistical analyses of copy number and gene expression were performed in PGS 6.5, whereas methylation was analyzed in Genome Studio. Associations between TERT promoter methylation and cancer were calculated with Fisher’s exact test and a two-tailed Student’s t test. Patient survival was calculated in StataSE (version 12), while other statistical analyses were conducted in R (version 3.0.1) and SPSS (version 20) (see Appendix-Supplemental Experimental Procedures). Survival estimates for tumour subgroups, and for CPCs by TP53 and ploidy status were generated using the Kaplan-Meier method and curves were compared using a log-rank test. Overall survival (OS) measured time from initial diagnosis to death from any cause or last follow-up as of December 1, 2013. Event-free survival (EFS) measured time from initial

60 diagnosis to tumour progression, recurrence or death from any cause. p values <0.05 were considered statistically significant.

Chapter 3: Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes

This chapter has been published and is reproduced with permission from Cancer Cell

1. Acknowledgements This chapter is a modified and reformatted version of the manuscript entitled “Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes” published in Cancer Cell (Tong et al. 2015). Under the supervision of David Malkin, my contribution to this work included designing the cross species analysis approach used in this study, executing the bioinformatics copy number and gene expression analyses of the entire human CPC data, and investigating clinical correlations according to the presence or absence of the novel oncogenes in the tumours of our patient population. Richard Gilbertson provided key guidance and mentorship throughout this project and during my time working in his laboratory at St. Jude Children’s Research Hospital. Dr. Yiai Tong’s arduous work led to the creation of the first CPC mouse model. He conducted all experiments in mice. Birgit Nimmervoll conducted work with CPC mouse cells and drug screening, while David Finkelstein and Yong-Dong Wang provided the bioinformatics support. James Dalton and David W. Ellison conducted the pathology and FISH experiments. Margaret Pienkowska provided training support and Stephen Mack, Marc Remke and Vijay Ramaswamy analytical assistance.

2. Abstract Choroid plexus carcinomas (CPCs) are poorly understood and frequently lethal brain tumours with few treatment options. Few recurrent point mutations or focal amplifications have been identified in whole genome sequencing studies of paediatric cancers. Rather, these tumours contain large chromosomal alterations encompassing tens to thousands of genes, rendering the identification of oncogenes difficult. CPCs frequently gain chromosome 1 that encodes ~2,100

62 genes. Using copy number and gene expression data from a newly generated mouse model of the disease and 24 human CPCs, we performed a cross-species, genome-wide search for novel oncogenes within syntenic regions of chromosome gain (Figure 3.1). TAF12, NFYC and RAD54L, co-located on human chromosome 1p32-35.3 and mouse chromosome 4qD1-D3, were identified as oncogenes that are gained in tumours in both species and required for disease initiation and progression in mice. TAF12 and NFYC are transcription factors that regulate the epigenome, while RAD54L plays a central role in DNA repair. Our data identify a group of concurrently gained, novel oncogenes that cooperate in the formation of CPC and reveal potential new avenues for therapy.

Figure 2.1: Graphical abstract of cross-species choroid plexus carcinoma study Summary of cross-species approach taken to identify and validate oncogenes associated with choroid plexus carcinoma development. Reprinted by permission from Cancer Cell (Tong et al., 2015; 27:712–727). Copyright 2015.

3. Introduction 3.1. Choroid plexus carcinomas

Choroid plexus carcinomas (CPCs) are malignant and highly vascularized intraventricular tumours observed in infants and small children. The great majority of CPCs are diagnosed in children aged less than three years, of whom two thirds die within five years (Wrede et al. 2009; Tabori et al. 2010). The biological and clinical behaviour of CPCs is highly variable as these malignant tumours respond differently to current treatment protocols, which include surgical resection (Lafay-Cousin et al. 2010). Efforts to identify more effective treatments of CPC have been hindered by poor understanding of its pathogenesis.

63 Recent molecular studies have defined the molecular variability of CPCs, which parallel the phenotypic variability observed in clinic (Tabori et al. 2010; Merino et al. 2015; Ruland et al. 2014). CPCs exhibit large chromosomal instability that may enrich cells with the right combination of alterations driving tumour growth. These malignant tumours are characterized by gains of chromosomes 1, 2, 4, 7, 12, 14, 19, 20 and 21, and numerous autosomal losses (Paulus et al. 1999; Rickert et al. 2002; Ruland et al. 2014). Moreover, approximately 50% of CPCs exhibit mutations in TP53. Mutations in this tumour suppressor gene modify gene expression and DNA methylation patterns of CPCs (Merino et al. 2015) and are associated with poor survival outcomes (Tabori et al. 2010; Merino et al. 2015).

3.2. Cross-species approach to identify novel oncogenes

Various genetic alterations activate oncogenes or delete tumour suppressor genes (TSGs) in cancer. Recurrent mutations that disrupt the same gene can be highly informative, pinpointing oncogenic alterations that may serve as therapeutic targets (Baselga et al. 1996; Druker et al. 2001; Flaherty et al. 2010). But focal alterations are relatively infrequent in many cancers, particularly those arising in children (Alexandrov et al. 2013; J. Zhang, Ding, et al. 2012). Rather, these tumours contain large DNA copy number alterations (CNAs) that presumably drive the overexpression of oncogenes or delete TSGs (Chen et al. 2014; Downing et al. 2012; Johnson et al. 2010; Wu et al. 2012). These CNAs are often chromosomal in scale, making it difficult to identify which genes are driving transformation. RNA silencing technologies have discovered novel TSGs within large deletions (Scuoppo et al. 2012; Xue et al. 2012; Zender et al. 2015); but approaches to screen the oncogenic capacity of genes located within large regions of gain are less well developed. A comprehensive cross-species analysis of syntenic chromosomal CNAs between human and animal models of the disease would identify novel oncogenes located within regions exhibiting copy number gain. This data could be further filtered by integrating copy number to gene expression data, thus identifying copy number-driven changes in gene expression that are present in both human and mouse tumours. The data obtained with this integrative type of analysis will pinpoint at oncogenes required for disease initiation and progression.

64 3.3. CPC development in mouse

Research in mouse has identified that ablation of Tp53 and/or Rb function drive CPC formation (Brinster et al. 2015; Sáenz Robles et al. 1994). Deletion of PTEN has also been implicated in CPC, but activated oncogenes have not been described (Morigaki et al. 2012; Rickert et al. 2002; Ruland et al. 2014). To date, a mouse model of CPC has not been generated.

4. Results 4.1. Deletion of Tp53, Rb and Pten generates CPC in mouse

To better understand the cellular and molecular origin of CPC, the Gilbertson laboratory developed the first mouse model of the disease using in utero electroporation of Cre Recombinase into the hindbrain choroid plexus epithelium (CPE) of embryonic day 12.5 Tp53flx/flx; Rbflx/flx; Ptenflx/flx; ROSAYFP mice. CPCs were observed in the ventricles of treated mice within 220 days of postnatal life with a penetrance of 38% (26/69). All tumours were YFP+, confirming their origin from in utero Cre- recombined cells. Mouse CPCs recapitulated the morphology, differentiation state, proliferation, and ultrastructural features of the human disease (Supplemental Figure 3.1A-M).

The need for losses in Tp53, Rb and Pten in the development of CPCs was investigated. These studies revealed that losses of both Tp53 and Rb are required to generate CPCs, and neither of these deletions can be substituted by loss of Pten. However, deletion of Pten together with loss of Tp53 and Rb significantly increased tumour penetrance (Supplemental Figure 3.1O, refer to Tong et al., 2015 for a detailed overview of the experiments conducted).

The transcriptomes of Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP CPCs with those of other mouse hindbrain tumours and tissues were compared in order to assess differences in gene expression profiles. CPCs and normal adult and embryonic choroid plexuses co-clustered separately from those of medulloblastoma and normal embryonic and/or adult brainstem and cerebellum (Supplemental Figure 3.1P) (Gibson et al. 2010; Uziel et al. 2005). Furthermore, gene set enrichment analysis (GSEA) identified ‘Markers of Choroid Plexus’ as the most

65 enriched of 4,293 gene sets in CPCs relative to mouse medulloblastoma, brainstem and cerebellum (Supplemental Figure 3.1Q). Among the choroid plexus derived tissues, CPCs were more closely related to embryonic than adult choroid. Thus, the newly generated CPC mouse model recapitulates the histological and ultrastructural features of the human disease and express an embryonic choroid plexus-like transcriptome.

4.2. Cross species analysis reveals candidate CPC oncogenes encoded in human chromosome 1p31.3-ter (mouse chromosome 4qC6-qE2)

To identify genetic alterations that drive CPC, Affymetrix 6.0 DNA Single Nucleotide Polymorphism (SNP) microarrays were used to catalogue CNAs in human choroid plexus papillomas (CPPs, n=32) and human CPCs (n=23). We also performed microarray comparative genomic hybridization (aCGH) of 47 Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP mouse CPCs, including 12 primary tumours and serial secondary (n=20) and tertiary (n=15) orthotopic transplants of these tumours. Additionally, the whole genomes of four human CPCs and matched normal blood were sequenced (WGS) with greater than 98% of the tumour genome and 91% of the normal genome showing 20-fold coverage with high-quality sequence reads (see Supplementary Experimental Procedures). No recurrent, single nucleotide variations, insertion/deletions, or focal CNAs (<5 genes) were identified in the four human CPCs subject to WGS, although one of these cases had extensive chromothripsis of chromosomes 1 and 19 in the absence of TP53 mutations. However, as expected, human CPCs contained numerous, recurrent, chromosomal gains and losses (Figure 3.2A). Similar non-random chromosomal alterations were observed in mouse tumours, suggesting large CNAs are a primary oncogenic driver of CPC (Figure 3.2B). Since both human and mouse CPCs contained recurrent chromosomal CNAs, we examined syntenic chromosome fragments that were gained in tumours in both species since these might be enriched for oncogenes. While 61% (n=14/23) of human CPCs gained at least one whole copy of chromosome 1 (encoding ~2,100 genes), only one syntenic fragment of chromosome 1 was gained in mouse CPCs (encoding 671 genes; mouse chromosome 4qC6-qE2 [94,608,732-155,608,945]; syntenic with human chromosome 1p31.3-ter [895,967- 67,594,220]; Figure 3.2C). Gain of 4qC6-qE2 was seen in 50% (n=6/12) of mouse CPCs and was retained

66 through secondary and tertiary serial tumour transplants (Figure 3.2C). Only 3% (n=1/32) of the more benign human CPPs gained chromosome 1. Thus human chromosome 1p31.3-ter / mouse chromosome 4qC6-qE2 (hereon, chr1p31.3-ter/4qC6-qE2) may harbour oncogenes that drive aggressive choroid plexus tumours.

Figure 3.2: Cross-species analysis of syntenic chromosomal gains in human and mouse CPC

67 Heatmaps of genomewide DNA copy number alterations in human CPPs and CPCs (A) and mouse CPCs (B). Heatmaps of the copy number of chromosome 1 (C), 7 (D) and 12 (E) in human CPPs and CPCs (left in each figure) and the corresponding syntenic regions in mouse CPCs (right in each figure). Reprinted by permission from Cancer Cell (Tong et al., 2015; 27:712– 727). Copyright 2015.

Notably, gain of chr1p31.3-ter/4qC6-qE2 may be associated with dysfunctional TP53 since our mouse model is Tp53 null and chromosome 1 gain was significantly associated with mutant TP53 in human CPCs (Fisher’s Exact, p<0.05; Figure 3.3). Similar analyses of other autosomal gains in human CPCs including chromosomes 7 and 12 that are among the most commonly gained chromosomes in the disease (Rickert et al. 2002; Ruland et al. 2014), failed to identify additional syntenic gains (Figure 3.2D,E). To further resolve which of the 671 genes on chr1p31.3-ter/4qC6-qE2 might be oncogenes, copy number and expression of this region in human and mouse tumours were integrated to identify genes that were both gained and overexpressed (Figure 3.3). Twenty-six percent of genes (n=176/671) on human chromosome 1p31.3-ter were significantly overexpressed in human tumours that gained this region (n=9), relative to those tumours in which this region was balanced or deleted (n=25; log ratio >2, p<0.05 with Bonferroni correction). Mouse CPCs that gained 4qC6-qE2 (n=15) overexpressed 11% of genes (n=64/579 Affymetrix 430 microarray) in this region relative mouse tumours exhibiting a balanced region (n=5) and normal mouse choroid plexus (n=7; log ratio >2, p<0.05 with Bonferroni correction). Comparison of these human and mouse data revealed a common set of 21 genes on chr1p31.3-ter/4qC6-qE2 that were both gained and overexpressed in human and mouse CPCs, pinpointing these as ‘lead candidate’ oncogenes (Figure 3.3).

Figure 3.3: A common set of 21 syntenic genes gained and overexpressed on 1p31.3-ter in human and 4qC6-qE2 in mouse, CPC Top: heat maps of human chromosome 1 copy number in 34 human CPCs and CPPs (left), and mouse chromosome 4qC6-qE2 in 27 mouse CPCs and normal choroid (right). Middle: TP53 status and tissue type of each sample. Bottom: heat maps reporting the expression of 21 copy-number driven orthologs located on human chromosome 1p31.3-ter in the 34 human tumours (left) and 27 mouse CPCs and normal choroid (right). Dotted lines demarcate the results of independent microarray probes for each human and mouse ortholog. Gene symbols shown right. Reprinted by permission from Cancer Cell (Tong et al., 2015; 27:712–727). Copyright 2015.

69 4.3. Gain of chromosome 4qC6-qE2 is an early event in CPC tumourigenesis

Cancers accumulate genetic alterations sequentially, suggesting these defects play temporally distinct roles during transformation. Therefore, it is imperative to identify the timing when the region of interest (4qC6-qE2) is gained during CPC development (Supplemental Figure 3.2). To investigate this, in-utero electroporated mice were sacrificed at P0, P21 or P35 two hours following injection with BrdU and their hindbrains were subjected to histological study (>4 mice per time point). At one week following electroporation (P0), small areas of YFP+ dysplasia where the epithelium lost its monolayer organization and decreased transthyretin (Ttr) expression was observed in the mice CPE. However, proliferation (BrdU incorporation), DNA double strand breaks (gH2AX staining), and gains of 4qC6-qE2 (fluorescence in situ hybridization [FISH]) were not detected (Supplemental Figure 3.2A-D). At three weeks post-electroporation (P21), the amount of dysplastic YFP+/Ttr- CPE had clearly increased and significant numbers of aberrantly proliferating YFP+/BrdU+ CPE were now detected relative to control mice (p<0.05 Mann-Whitney; Supplemental Figures 3.2A, B); however, neither DNA double stand breaks nor 4qC6-qE2 gain were apparent. By six weeks post-electroporation (P35) the choroid plexus had undergone a dramatic change: large regions of hyperplastic CPE were now visible, in which 20%+2.4SE of YFP+ CPE were proliferating (BrdU+, p<0.0005 Mann Whitney, relative to controls; Supplemental Figures 3.2A,B). In addition, significant DNA DSBs and gain of 4qC6-qE2 were now detected in YFP+ CPE (both p<0.0005, Mann-Whitney; Supplemental Figure 3.2 A,C,D). In fully formed CPCs, high levels of aberrant proliferation and gain of 4qC6-qE2 persisted; however, DSBs were not detected in these tumours (Supplemental Figures 3.2 B-D). Together, these data confirm that mouse CPCs develop from mutated CPE, that early DNA DSBs are associated with initiation of chromosomal instability, and that gain of 4qC6-qE2 is a relatively early event in the disease process, coinciding with choroid plexus hyperplasia. These findings suggest this region may contain oncogenes that play a role in tumour initiation.

70 4.4. Taf12, Nfyc and Rad54l promote aberrant proliferation of the developing choroid plexus epithelia

Since gain of 4qC6-qE2 is an early event in CPC development, we reasoned that developing choroid plexus would be an appropriate context in which to screen the transforming potential of our 21 lead candidate oncogenes. To do this, the capacity of each candidate to drive dysplasia and proliferation of early postnatal CPE was tested. Plasmids expressing green fluorescence protein (GFP) and a single lead candidate oncogene each were co-electroporated into the hindbrain CPE of E12.5 embryos (>5 embryos/candidate, total n=107 embryos; Supplemental Figure 3.3A). Five control mice were electroporated with empty vector-GFP. Proliferating CPE cells were labelled at E19.5 by maternal injection of BrdU, and the proportion of targeted and proliferating (GFP+/BrdU+), relative to targeted but non- proliferating (GFP+/BrdU-) CPE cells was calculated at P0. GFP+ cells were detected in all electroporated mice. Fewer than 0.5% of electroporated (GFP+) CPE cells were proliferating (BrdU+) in embryos that were electroporated with control plasmid or 13 of 21 lead candidate oncogenes (Supplemental Figure 3.3A-C). Expression of five other candidates (Cdc20, Stmn1, Atp6v0b, Clspn, Mier1) was associated with proliferation in 0.5-2% of GFP+ cells but did not alter choroid plexus morphology (Supplemental Figure 3.3B,C). In marked contrast, Taf12, Nfyc and Rad54l induced CPE dysplasia and proliferation similar to that observed following 4qC6- qE2 gain in Cre-electroporated Tp53flx/flx; Rbflx/flx; Ptenflx/flx; ROSAYFP mice (compare Supplemental Figures 3.2A and 3B). Notably, when correlating these findings to human CPC patients, we observed a significant reduction in overall survival among patients whose CPCs expressed relatively high levels of RAD54L (p=0.0036, Log-rank test), and a trend for reduced survival in patients whose tumours expressed relatively high levels of TAF12 and NFYC (Figure 3.4). Therefore, in light of these data we selected Taf12, Nfyc and Rad54l for further study as potential CPC oncogenes.

Overall Survival of CPCs by RAD54L Expression Overall Survival of CPC by TAF12 Expression

Increased RAD54L Increased TAF12 100 Not increased RAD54L 100 Not increased TAF12

80 80

60 60

40 40 Percent survival 20 Percent survival 20

p=0.0036 0 0 p=0.1805 0 2 4 6 8 10 0 2 4 6 8 10

Overall Survival Overall Survival

Overall survival of CPC by NFYC expression

100 Increased NFYC 80 Not increased NFYC

Percent survival 20

0 p=0.2012 0 2 4 6 8 10

Overall Survival Figure 3.4: Prognostic significance of RAD54L, TAF12 and NFYC expression in human CPC We first calculated the average expression level of each gene in our cohort of relatively benign CPPs using Affymetrix arrays. CPCs were then segregated into tumours that expressed each gene at levels greater than the CPP average (relatively increased, red lines) or less than the average (relatively decreased, blue lines). Reprinted (adapted) by permission from Cancer Cell (Tong et al., 2015; 27:712–727). Copyright 2015.

4.5. Expression of Taf12, Nfyc and Rad54l is required to maintain CPC in vivo

Next, the expression of Taf12, Nfyc and/or Rad54l was knocked down in vitro and in vivo to identify whether they are required to maintain CPC cells. Three shRNA-green fluorescence protein (GFP) lentiviruses against each candidate were generated. shRNAs targeting two ‘control’ genes (Pqlc2 and Mier1) on chr1p31.3-ter/4qC6-qE2 that did not induce dysplasia or aberrant proliferation of developing CPE were also generated (Supplemental Figure 3.3C). Greater than 90% of mouse primary CPC cells were transduced following exposure to lentiviruses in vitro, resulting in >50% knockdown of the corresponding target gene relative to control shRNA transduced cells (Supplemental Figure 3.4A). Ablation of either Rad54lshRNA or Taf12shRNA induced significant Caspase 3/7 activity in mouse CPC cells and dramatically reduced the survival of these cells in culture relative to control shRNA, Pqlc2shRNA or Mier1sh

72 RNA transduced cells (Supplemental Figure 3.4B, C). In contrast, knockdown of Nfyc did not affect apoptosis or survival of CPC cells in vitro. To test if Taf12, Nfyc and Rad54l maintain CPC cells in vivo, we transduced primary CPC cells with Taf12shRNA, NfycshRNA, Rad54lshRNA or control shRNA and transplanted them orthotopically into the hindbrains of immunocompromised mice (Supplemental Figure 3.4D). In agreement with the in vitro data, the median survival of mice harbouring Taf12shRNA (n=6, median survival 34 days) or Rad54lshRNA (n=8, median survival 55.5 days) transduced CPC cells was significantly longer than mice injected with control shRNA transduced CPC cells (n=19, median survival 19 days; P<0.005 Log Rank; Supplemental Figure 3.4D). Mice harbouring Nfyc shRNA transduced CPC cells also survived longer than control animals (n=5, median survival 29 days, p=0.05 Log Rank). Thus expression of Taf12, Nfyc and Rad54l appears to support optimal growth of CPC cells, supporting the notion that they are oncogenes.

4.6. Taf12, Nfyc and Rad54l are required to initiate CPC

Since gain of 4qC6-qE2 coincided with the onset of severe hyperplasia in our mouse model (Supplemental Figure 3.6) we next tested whether the expression of our candidates is required to initiate CPC. Cre Recombinase was electroporated into the hindbrain choroid plexus of E12.5 Tp53flx/flx; Rbflx/flx; Ptenflx/flx; ROSAYFP mouse embryos exactly as described above, but this time, embryos were simultaneously injected with either Taf12shRNA (n=6 mice) NfycshRNA (n=13 mice), Rad54lshRNA (n=10 mice) or control shRNA (n=12 mice) lentivirus. Remarkably, while 42% (n=5/12) of Cre-electroporated mice injected with control shRNA developed CPCs, injection of either Taf12shRNA or NfycshRNA lentiviruses completely abolished CPC development and only 10% (n=1/10) of mice receiving Rad54lshRNA lentiviruses developed tumours (P<0.05 Log Rank; Supplemental Figure 3.4E). To confirm that the choroid plexus of mice injected with active shRNAs underwent Cre- recombination and viral transduction, the hindbrains of these were reviewed after >100 days. Several of these mice contained hyperplastic but histologically benign choroid plexus masses containing numerous YFP+ (Cre-recombined) and GFP+ (shRNA transduced) cells (Supplemental Figure 3.4F), suggesting that expression of shRNAs had arrested CPC development.

73 These data provide compelling evidence that expression of Taf12, Nfyc or Rad54l is critical for the initiation of CPC.

4.7. Upregulation of Taf12, Nfyc and Rad54l drives CPC

Having shown that loss of Taf12, Nfyc and Rad54l expression markedly impairs CPC initiation and maintenance, we conducted two sets of experiments to test whether overexpression of these genes might promote disease initiation and/or progression. First, we tested if IV ventricular co-injections of Taf12-cop(c)GFP, Nfyc-cGFP and Rad54l- cGFP expressing lentiviruses (1.67x105 lentiviral particles each per mouse, total=5x105 particles) alters CPC disease course in Cre-electroporated Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP E12.5 embryos. The tumour free survival of Cre-electroporated and lentivirus injected mice was significantly reduced relative to that of mice that were Cre-electroporated alone (Supplemental 3.5A; Log Rank p<0.05). Mice that were injected with Taf12-cGFP, Nfyc-cGFP and Rad54l-cGFP lentiviruses but were not Cre-electroporated contained numerous cGFP+ CPE cells but did not develop CPCs (total mice n=7, median follow up 374 days; Supplemental Figure 3.5A). Thus, increased expression of Taf12, Nfyc and Rad54l significantly accelerates CPC development in Tp53, Rb, Pten deleted CPE, but is not sufficient to initiate the disease. These data are compatible with the observation that 4qC6-qE2 is gained following deletion of Tp53, Rb, Pten, but precedes CPE transformation (Supplemental Figure 3.2A,D).

As a further test of whether upregulation of Taf12, Nfyc and Rad54l can cooperate with TSG deletion to drive CPC, we generated a variant of our mouse model in which we titrated the amount of Tp53, Rb and Pten deletion using a Cre Recombinase-cGFP (Cre-cGFP) lentivirus. The lentiviral delivery of Cre-recombinase serves as a titratable alternate to electroporation for inducing CPCs in our model system. The injection of just 1.25x105 Cre-cGFP lentiviral particles yielded no tumours, however, when Taf12-cGFP, Nfyc-cGFP and Rad54l-cGFP lentiviruses were concurrently delivered in this model, YFP+ (recombined) /cGFP+ (lentiviral transduced) CPCs formed (Supplemental Figure 3.5B, C). When single candidates were transduced, the mice still developed CPC but at significant lower penetrance than animals receiving all three oncogene candidates (Supplemental Figure 3.5C).

74 Together with our cross-species genomic and in vivo gene knockdown studies, these data comprehensively validate Taf12, Nfyc and Rad54l as novel CPC oncogenes and support the notion that concurrent gain of these three genes on chr1p31.3-ter/4qC6-qE2 cooperates with deletion of Tp53, Rb and Pten in the initiation and progression of CPC.

4.8. Aberrant DNA metabolism is a significant feature of CPC

TAF12, NFYC and RAD54L are key regulators of DNA metabolism, suggesting that dysregulation of DNA maintenance and/or repair are necessary for CPC formation (Benatti et al. 2011; Nardini et al. 2013; Schmitz et al. 2009; Wright & Heyer 2014). As a first step to test this, GSEA was used to compare the transcriptomes of mouse E12.5 CPE with those of daughter CPCs. Moreover, we conducted an experiment on which the sensitivity of CPC cells to two separate inhibitors of the ataxia telangiectasia and Rad3-related protein (ATR) kinase, a key gene involved in DNA repair, including DSBs (Somyajit et al. 2013) was measured. In keeping with our hypothesis, numerous gene sets that maintain the integrity of the genome and epigenome were significantly enriched in CPC (Figure 3.5A, B). As expected, Rad54l was the most upregulated homologous repair gene (Figure 3.5B). Notably, DNA replication and repair gene sets were also enriched in human CPCs that are diploid for chromosome 1 relative to CPPs, suggesting dysregulation of these cell functions are a general feature of CPC (Figure 3.5C). Finally, CPC cells proved remarkably sensitive to ATR inhibitors in vitro relative to cells from an unrelated mouse ependymoma (Figure 3.5D). Exposure of CPC cells to 0.4mM AZ-20 that inhibited their growth by 90%, increased the induction of DNA double stranded breaks in these cells >3-fold after only six hours; however, this concentration of AZ-20 had no impact on ependymoma cell proliferation or DSB formation (Figure 3.5D,E).

A B

C D Ribosome processing Viral processes mRNA editing

Neurotransmitter activity DNA replication

Cell cycle regulation Transcription & DNA repair translation Mitosis

Chromosome Response to condensation apoptosis Chromatid segregation Cell cycle regulation via APC/C:CDC20 Proteasome mediated Cellular motility Chromosome degradation & transport packing Red: Enriched in diploid CPC Blue: Enriched in diploid CPP

Figure 3.5: DNA repair is upregulated in CPC (A) Gene set enrichment analysis (GSEA) of ‘Kauffmann DNA Repair Gene’ set in CPC versus E12.5 CPE. (B) Heat map reporting the GSEA of the ‘Kegg Homologous Repair Gene’ set in CPC versus E12.5 CPE. (C) GSEA plot depicting different biological pathways and processes enriched in diploid CPCs (red) and diploid CPPs (blue), (D) Growth inhibition assays of mouse CPC and ependymoma (Johnson et al., 2010) cells following 72 hour exposure to the ATR inhibitors AZ-20 and VE-821. (E) Quantification of DNA DSBs in the CPC and ependymoma cells shown in D, both before (time 0) and at the indicated times following exposure to 0.4mM AZ-20. Reprinted (adapted) by permission from Cancer Cell (Tong et al., 2015; 27:712–727). Copyright 2015.

Importantly, while sensitivity to ATR inhibitors may vary with basal levels of replicative stress; CPC cells proliferated slightly slower that ependymoma cells (doubling time 25.7 hours +1.2SD CPC vs. 19.6 hours +1.7SD ependymoma) and no significant difference in replication stress was detected between untreated CPC and ependymoma cells (Figure 3.5E). These data support the hypothesis that gain of TAF12, NFYC and RAD54L on chr1p31.3-ter/4qC6-qE2 promotes aberrant DNA maintenance and repair during CPC formation and suggest that inhibition of this activity may have therapeutic potential.

5. Discussion CPCs are frequently lethal paediatric brain tumours for which there are few treatment options. Efforts to develop new therapies for these tumours have been limited by a lack of understanding of the molecular alterations that drive the disease. Here, using a combination of cross-species genomics and in vivo mouse modelling, we describe a new mouse model of CPC; demonstrate that postnatal CPE from which TSGs have been deleted in utero can generate these tumours; identify three novel oncogenes – TAF12, NFYC and RAD54L – that are required to initiate and maintain the disease; and implicate upregulation of DNA maintenance and repair as a critical requirement for CPC formation.

How might concurrent gain of TAF12, NFYC and RAD54L contribute to CPC formation? Both TAF12 and NFYC are histone-fold domain containing transcription factors (Bieniossek et al. 2013; Nardini et al. 2013). TAF12 recruits GADD45, and thereby the nucleotide excision repair (NER) machinery, to the promoter of active genes, removing methylated cytosines and maintaining a hypomethylated active state (Schmitz et al. 2009). Transcriptional regulation by TAF12 is thought to contribute to transformation, possibly by promoting an invasive phenotype (Voulgari et al. 2008). NFYC is also engaged in chromatin remodelling, establishing permissive chromatin modifications at CCAAT promoters, including those of cell cycle regulators (Benatti et al. 2011; Nardini et al. 2013). Deletion of NFYC or the related gene NFYB, halts cell cycle progression, predominantly by causing a G2/M arrest (Benatti et al. 2011). Thus, upregulation of TAF12 and NFYC might promote an aberrant epigenome during CPE transformation, maintaining or re-activating the expression of proto-oncogenes that are normally silenced in

77 post-mitotic, postnatal choroid plexus. This notion is supported by our observation that the CPC transcriptome most closely matches embryonic CPE. RAD54L plays a central role in homologous recombination in which damaged DNA – particularly DSBs – are repaired by copying an intact homologous sequence (Wright & Heyer 2014). Loss of Rad54l from Tp53 and LigIV null cells results in the accumulation of DSBs, severe chromosomal defects, and failed cell proliferation (Mills et al. 2004). Therefore, upregulation of Rad54l may be required for CPE to tolerate the severe genotoxic stress that is associated with rampant proliferation in the context of accumulating chromosomal insults. Our observations that Tp53/Rb/Pten null CPE accumulate DSBs that apparently resolve following the gain of 4qC6-qE2, and that CPC transcriptomes are markedly enriched for DNA repair genes, supports this concept. In light of our data, we propose a new hypothesis for the development of CPC; particularly in patients with germline mutations in TP53 (Figure 3.6). We propose that the loss of TP53 (and potentially RB and PTEN) from CPE results in the accumulation of genetic and epigenetic changes that disrupt terminal differentiation and senescence. Early in postnatal life, these aberrantly proliferating CPE cells develop increasingly abnormal genomes leading to genotoxic crisis. While most of these cells undergo cell death, presumably through TP53-independent mechanisms, a fraction acquire aberrant DNA repair and epigenome remodelling capacity, including gain of TAF12, NFYC and RAD54L. This enables them to tolerate an aberrant but stable genome and drive further transformation. The advantage afforded by these genes provides a selective pressure for retention of chr1p31.3-ter/4qC6-qE2 gain. These mutant CPE cells go on to form CPCs that can tolerate the recurrent and extensive chromosome changes associated with the disease (Merino et al. 2015). Thus, the concurrent gain of TAF12, NFYC and RAD54L may serve to model an aberrant epigenome that promotes a proliferative, and relatively undifferentiated state (TAF12 and NFYC), while tolerating genotoxic stress (RAD54L). Interestingly, recent data indicate that CPCs that do not gain chromosome 1 retain both copies of this chromosome in an otherwise hypodiploid genome (Merino et al. 2015). Thus absolute or relative gain of this chromosome might be required for CPC development.

Our demonstration that increased expression of Taf12, Nfyc and Rad54l induce dysplasia and aberrant proliferation in otherwise wild-type CPE, support the notion that these are CPC oncogenes. However, the dependency of transforming cells on aberrant DNA repair may also occur in the context of non-oncogene addiction (Luo et al. 2009).

Figure 3.6: Model proposed for the development of CPC from normal choroid plexus epithelium This model includes the early deletion of TSGs such as Tp53, followed by the accumulation of additional genetic and epigenetic alterations driving the cell into genotoxic stress. Those cells that have gained oncogenes involved in the activation of aberrant DNA repair (RAD54L), and/or an aberrant epigenome (NFYC and TAF12) are able to survive the chromosomal instability observed in these tumours and eventually transform into CPC with a stable aberrant genome. Reprinted (adapted) by permission from Cancer Cell (Tong et al., 2015; 27:712–727). Copyright 2015.

79 Thus, CPCs might be ‘addicted’ to RAD54L, and/or the targets of TAF12 and NFYC chromatin remodelling, as non-oncogenes. Regardless of the underlying mechanism, validation of aberrant DNA repair as a requirement for CPC formation and maintenance could lead to new therapies, since inhibition of DNA repair enzymes should result in intolerable DNA damage. Precedent for this approach exists in the addiction of BRCA2 mutant cells to the DNA repair enzyme PARP1 and our demonstration that CPC cells are sensitive to ATR inhibitors (Sakai et al. 2008). As well as suggesting potential drug targets within DNA repair pathways, our data show that inhibitors of phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K) signalling (suppressed by PTEN) might also serve as new CPC treatments. But translating these and other potential new therapies to the clinic for CPC has proved immensely challenging since fewer than 50 children are diagnosed with this disease in the United States each year (Louis et al. 2007).

6. Conclusion Our mouse model is the first to be validated as a faithful replicate of human CPC. Therefore, rigorous preclinical studies using this model could help discover, triage and prioritize potential new therapies for clinical trial. This model will be a highly beneficial tool in the study of mechanisms of tumour development and functional analysis as there are currently no other biological models of this disease.

Our syntenic mapping approach holds promise to identify oncogenes located within large regions of chromosomal gain in other cancers. Indeed, emerging evidence suggests that large CNAs characteristic of other paediatric solid tumours are replicated when these tumours develop in other species (Chen et al. 2013; Frappart et al. 2009). Similar studies of syntenic regions of loss coupled with RNA silencing technologies could be applied to discover novel TSGs (Scuoppo et al. 2012; Xue et al. 2012; Zender et al. 2015).

7. Methods and Materials 7.1. Patients and sample preparation for microarray study

Fresh frozen and FFPE preserved human CPCs (n=30) and CPPs (n=33), as well as associated clinical data were collected from institutions in Canada, USA, Brazil, Israel, and Germany in

80 accordance with each institution’s Research Ethics Board. Informed consent was obtained from the parents/legal guardians of all patients. Tumour DNA was extracted using standard phenol- chloroform extraction from fresh-frozen samples, and the RecoverAll™ Total Nucleic Acid Isolation Kit for FFPE (Ambion, Carlsbad, USA) from FFPE samples. Total RNA was isolated from fresh-frozen samples using the TRIzol method (Invitrogen, Carlsbad, USA) according to the manufacturer’s instructions.

7.2. TP53 sequencing

7.3. RNA and DNA microarray analysis

Gene expression profiles of human and mouse RNAs were generated using GeneChip® Human Exon 1.0ST and 430v2 microarrays, respectively (Affymetrix, Santa Clara, USA). Gene expression data were normalized and analyzed to detect significantly differentially expressed genes exactly as described previously (Parker et al. 2014). Mouse CPE samples were used as reference controls for expression analysis. Human CPPs were used as reference due to the lack of normal human CPE. Human and mouse DNA samples were hybridized to Genome-Wide Human SNP Array 6.0 (Affymetrix, Santa Clara, USA) and the Mouse Genome CGH 244K (Agilent), respectively. Human gene copy number was inferred using Partek® Genomics Suite™ 6.5 and genomic segmentation algorithm (Partek, St Louis, MO). Segmented regions were those significantly different from neighbouring regions (p<0.001) and contained 10 or more probes, with a signal to noise ratio of 0.3. Mouse aCGH tumour data were compared to a common control as reference DNA and the circular binary segmentation algorithm implemented in the DNAcopy package from Bioconductor was used to identify copy number alterations for each tumour sample. Aberrant segments contained at least 5 probes with a mean log2 ratio greater than 0.3 or less than -0.3.

81 7.4. Whole genome sequencing data analysis

Detection of somatic single nucleotide variations (SNVs), indels, structural variations and focal copy number changes from whole genome sequencing (WGS) was performed as previously described and are described in detail in Supplementary Experimental Procedures (Parker et al. 2014).

7.5. Statistical analysis

Patient survival was calculated in StataSE (version 12), while other statistical analyses were conducted in R (version 3.0.1) and SPSS (version 20)(see Appendix). Survival estimates were generated using the Kaplan-Meier method and curves were compared using a log-rank test. Overall survival (OS) measured time from initial diagnosis to death from any cause or last follow-up as of December 1, 2013. Event-free survival (EFS) measured time from initial diagnosis to tumour progression, recurrence or death from any cause. p values <0.05 were considered statistically significant.

Chapter 4: Thesis main findings and future directions in choroid plexus tumour research

1. Acknowledgements This chapter starts with a quick summary of the salient findings achieved as part of my thesis project. It is followed by a compilation of several studies conducted throughout the course of my PhD degree. Some of these studies are part of collaborations on which I am currently working and wish to publish in the near future. Thanks to Dr. Jan Korbel and his research group at the European Molecular Biology Laboratories (EMBL) for their help and expertise in the chromothripsis analysis of CPCs and for training me in such analyses. Thanks to Dr. Richard Gilbertson and his research group for funding and leading the next generation sequencing efforts for CPCs, to Dr. Xiaotu Ma and Yongjin Li for their bioinformatics expertise in the analysis of whole genome sequencing and RNA sequencing of CPCs. Thanks to Drs. Robert Wechsler-Reya and Martine Roussel for the opportunity to work with the Myc-driven CPC mouse model, and to Drs. Jun Wang and Brian Murphy for providing me a more in-depth understanding of the creation of Myc-driven CPC mouse models.

2. Salient findings obtained from thesis project The comprehensive molecular analysis of choroid plexus tumours and the use of a novel mouse models of CPC have provided key evidence that will serve as the framework for future research, aiming at improving our understanding of this disease and how to best manage it in young patients. Our findings demonstrated the molecular similarities of CPPs and aCPPs, which exhibited no significant differences in gene expression, DNA methylation, or copy number profiles. These benign tumour subtypes exhibited a very different molecular profile relative to CPCs, which paralleled the clinical differences observed in CPPs and aCPPs compared to CPCs. Because the molecular signature of aCPPs is comparable to CPPs, we suggest a similar approach to the clinical management of aCPPs, opting for a post-operative wait-and-see approach, rather than the administration of adjuvant chemo- or radiotherapy. CPCs, on the other hand, exhibited a heterogeneous molecular profile. Allele specific copy number analysis revealed differences in chromosome-wide gains and losses distinguishing between a hyperdiploid CPC subgroup,

83 characterized by chromosome-wide gains and acquired uniparental disomy, and a hypodiploid CPC subgroup characterized by chromosome-wide losses. Unsupervised clustering analyses of gene expression and DNA methylation demonstrated that CPCs segregate into subgroups that are differentiated by TP53 status, demonstrating that the activity of this tumour suppressor gene strongly modifies the molecular signature of CPCs. Gene expression and DNA methylation did not distinguish CPCs by ploidy status. We identified that the number of mutant TP53 copies was associated with reduced survival in CPCs in a dose-dependent manner. This finding should be validated in a prospective study in order to assess its prognostic capacity and clinical relevance. Lastly, through an integrative cross-species study, we identified three novel oncogenes involved in CPC development. Copy number increase and overexpression of RAD54L, NFYC and TAF12 alter DNA metabolism, promoting tumour development in the presence of genotoxic stress. These findings have refined the molecular landscape of CPTs and elucidated the mechanisms involved in CPC development driving distinct clinical outcomes. However, our findings have also revealed novel research avenues to pursue in the field of choroid plexus tumour biology. The following chapter sections describe potential areas for further investigation, which will build upon the molecular framework created by the findings generated by this thesis project.

3. High-throughput genomics and CPT research The advent of a new generation of genomics research, characterized by “big data” derived from high-throughout analyses, has driven a greater understanding of the molecular underpinnings of cancer. Conducting high-density genomic analyses on large tumour cohorts has exposed not only the blueprints of disease, unraveling its origin, foundation, and organization, but also its weaknesses—those that can be targeted for an effective and swift destruction.

CPTs have been largely understudied due to very low tumour incidence rates and sample availability. This limitation has prevented the in-depth study of CPT’s genomic code and the biological mechanisms underlying CPT development. For decades the only evidence available was that of a few isolated case studies reporting chromosomal alterations (Reviewed in Merino et al., 2015), patient response to treatment alternatives (Dickens et al. 2005; Addo et al. 2011) or its association with TP53 and the Li-Fraumeni (Rieber et al. 2009; Gonzalez et al. 2009; Malkin et al. 1990; Krutilkova et al. 2005; Seidinger et al. 2011) or Aicardi (Taggard & Menezes 2000; Pianetti Filho et al. 2002; Uchiyama et al. 1997) syndromes. It is only recently that multi-

84 institutional collaborations have enabled the study of large enough cohorts of high-quality tumour samples (Ruland et al. 2014; Merino et al. 2015; Japp et al. 2015). Refining the biological underpinnings of CPTs has led to the development of recommendations for improving the clinical approach to these tumours.

3.1. Clinical implications of the molecular substratification of CPTs

3.1.1. CPPs and aCPPs are molecularly similar

Our comprehensive study found that CPPs and aCPPs are genetically, epigenetically and transcriptionally alike, and different than CPCs (Merino et al. 2015). Clinical outcomes for patients suffering with CPPs and aCPPs were also very favourable. In both tumour subtypes, surgical resection was sufficient to achieve complete remission with no evidence of disease in the majority of patients. In our cohort, there were no deaths due to disease in CPP and aCPP patients, and event free survival was high (CPP: 87.4% and aCPP: 70%) demonstrating that these tumours exhibit low rates of recurrence. The significant differences between the copy number profile and clinical outcomes of CPPs and aCPPs relative to CPCs were also observed in an independent cohort of 66 patients for which FFPE-derived CPT tissue was available (Japp et al. 2015). These two independent studies have provided significant evidence to recommend that CPPs and aCPPs be treated using similar less aggressive therapeutic approaches, characterized by surgical resection followed by observation. Chemotherapy should be discouraged for aCPPs as a primary therapeutic option. Furthermore, molecular analyses assessing copy number and gene expression patterns differentiating aCPPs from CPCs (Figure 4.1) should complement histopathology when a sample cannot be conclusively diagnosed as aCPP or CPC. Prospective studies should assess the use of these recommendations in the care of patients affected by CPTs.

A CPC aCPP Frequency of chromosome-wide Frequency of chromosome-wide 100%# 100%# 100%# 100%# 90%# 80%# 70%# 60%# 50%# 40%# 30%# 20%# 10%# 10%# 20%# 30%# 40%# 50%# 60%# 70%# 80%# 90%# 10%# 20%# 30%# 40%# 50%# 60%# 70%# 80%# 90%# 90%# 80%# 70%# 60%# 50%# 40%# 30%# 20%# 10%# Losses Gains 0%# Losses Gains 0%# 0%# 0%# %100 80 60 40 20 0 0 20 40 60 80 100% %100 80 60 40 20 0 0 20 40 60 80 100% 1 11 11 2 22 22 3 33 33 4 44 44 5 55 55 6 66 66 7 77 77 8 88 88 9 99 99 aCPP$ aCPP$ CPC# CPC# 10 10 10 10 11 11 11 11 12 12 12 12 13 13 13 13 14 14 14 14 15 15 15 15 16 16 16 16 17 17 17 17 18 18 18 18 19 19 19 19 20 20 20 20 21 21 21 21 22 22 22 22 X XX XX Cluster dendrogram with AU/BP values (%) B 5 4 3 2 Height 1 0

Regulation of the anaphase

EXTRACELLULAR BASEMENT MATRIX MEMBRANE ORGANIZATION

EXTRACELLULAR promoting complex (APC/C) MATRIX PART

INTEGRIN BINDING

ANAPHASE-PROMOTING COMPLEX-DEPENDENT PROTEASOMAL UBIQUITIN-DEPENDENT PROTEIN CATABOLIC PROCESS

INACTIVATION OF APC/C VIA DIRECT INHIBITION OF THE APC/C COMPLEX

ACTIVATION OF APC/C AND APC/C:CDC20 CILIUM MEDIATED DEGRADATION OF MITOTIC PROTEINS

APC/C-MEDIATED DEGRADATION OF CELL CYCLE PROTEINS

AMPLIFICATION OF SIGNAL FROM THE KINETOCHORES ECM and membrane AXONEME

AMPLIFICATION OF SIGNAL FROM UNATTACHED DYNEIN COMPLEX KINETOCHORES VIA A MAD2 INHIBITORY INITIATION OF SIGNAL CHECKPOINT SIGNAL FROM DEFECTIVE KINETOCHORES Distance: correlation

APC-CDC20 MEDIATED DEGRADATION OF NEK2A INHIBITION OF THE PROTEOLYTIC ACTIVITY OF APC/C REQUIRED FOR THE ONSET OF ANAPHASE BY MITOTIC SPINDLE CHECKPOINT COMPONENTS

REGULATION OF UBIQUITIN-PROTEIN LIGASE ACTIVITY DURING MITOTIC CELL CYCLE organization

CELL CYCLE CHECKPOINTS Cluster method: ward.D

APC/C:CDC20 MEDIATED DEGRADATION OF SECURIN

REGULATION OF APC/C ACTIVATORS BETWEEN G1/S AND MITOTIC CELL EARLY ANAPHASE CYCLE CHECKPOINT

G2/M DNA REPLICATION CHECKPOINT

CDC20:PHOSPHO-APC/C MEDIATED DEGRADATION OF CYCLIN A

ORGANELLE FISSION G2/M CHECKPOINTS

APC/C:CDC20 MEDIATED DEGRADATION OF MITOTIC PROTEINS

CELL DIVISION

NEGATIVE REGULATION OF CELL GROWTH Cell movement

NEGATIVE REGULATION OF CELL SIZE

MITOTIC SPINDLE CHECKPOINT

NUCLEAR DIVISION

REMOVAL OF LICENSING FACTORS FROM ORIGINS

SIGNALING BY AURORA KINASES

ACTIVATION OF ATR IN RESPONSE TO REPLICATION M PHASE Cell cycle STRESS

CELL_CYCLE_KEGG

DNA REPLICATION PRE-INITIATION

CHK1/CHK2(CDS1) MEDIATED INACTIVATION OF CYCLIN B:CDK1 COMPLEX

REGULATION OF DNA checkpoint DNA REPLICATION REPLICATION Nuclear and cell

HSA04110_CELL_ CYCLE

POLO-LIKE KINASE MEDIATED EVENTS

"CELL CYCLE, MITOTIC" division

REGULATION OF MITOTIC CELL CYCLE

MITOTIC CELL CYCLE REGULATION OF MITOSIS CELL CYCLE CHECKPOINT

POSITIVE REGULATION OF UBIQUITIN-PROTEIN LIGASE ACTIVITY DURING MITOTIC CELL CYCLE

NEGATIVE REGULATION OF UBIQUITIN-PROTEIN LIGASE ACTIVITY DURING MITOTIC Cell cycle CELL CYCLE

SPINDLE ORGANIZATION

M PHASE OF MITOTIC regulation CELL CYCLE

MITOSIS

"CHROMOSOME, CENTROMERIC REGION"

POSITIVE REGULATION OF CELL MIGRATION "CONDENSED CHROMOSOME, CENTROMERIC REGION"

MITOTIC SISTER CHROMATID SEGREGATION Chromosome

REGULATION OF LOCOMOTION

MITOTIC PROMETAPHASE

CHROMOSOME Mitosis SEGREGATION

NEGATIVE REGULATION OF CELL MOTION CONDENSED CHROMOSOME KINETOCHORE segregation

CONDENSED CHROMOSOME

SISTER CHROMATID Chromosome SEGREGATION

REGULATION OF CELL MIGRATION ENDOTHELIAL CELL MIGRATION

VASCULATURE DEVELOPMENT

POSITIVE REGULATION OF condensation CELL MOTION

CHROMOSOMAL PART

REGULATION OF ENDOTHELIAL CELL MIGRATION

REGULATION OF CELL MOTION

CHROMOSOME

CHROMATIN ASSEMBLY

PROTEIN-DNA COMPLEX ASSEMBLY

DNA PACKAGING

BLOOD VESSEL ENDOTHELIAL CELL MIGRATION

NUCLEOSOME

CHROMATIN ASSEMBLY OR Regulation of DISASSEMBLY

NEUROPEPTIDE NEUROPEPTIDE RECEPTOR BINDING ACTIVITY

METABOLISM OF WATER-SOLUBLE VITAMINS AND COFACTORS

REGULATION OF COAGULATION

BLOOD VESSEL MORPHOGENESIS

CHROMATIN NUCLEOSOME ORGANIZATION

PROTEIN-DNA cell migration COMPLEX IONOTROPIC GLUTAMATE RECEPTOR GLUTAMATE ACTIVITY SIGNALING PATHWAY

POLYSACCHARIDE BINDING

NUCLEOSOME ASSEMBLY

MULTICELLULAR ORGANISMAL MACROMOLECULE METABOLIC COLLAGEN PROCESS METABOLIC PROCESS

CALMODULIN-DEPENDENT PROTEIN KINASE ACTIVITY

BLOOD VESSEL DEVELOPMENT

SMOOTH MUSCLE CONTRACTION

REGULATION OF SYNAPTIC PLASTICITY

CELLULAR RESPONSE TO EXTRACELLULAR STIMULUS

HEPARIN BINDING "OXIDOREDUCTASE ACTIVITY, ACTING ON THE CH-OH GROUP OF DONORS, NAD OR NADP AS ACCEPTOR"

GLYCOSAMINOGLYCAN BINDING Chromatin assembly Vasculature development and endothelial migration Cellular metabolism Figure 4.1: CPCs and CPPs are molecularly distinct (A) aCPPs are characterized by a high frequency of chromosome-wide gains—especially in chromosomes 7, 11, 12 and 20, whereas CPCs are characterized by frequent chromosome-wide gains and losses (B) Unsupervised gene expression analysis significantly segregates aCPPs (blue) from CPCs (red), and (C) GSEA analysis demonstrating these two tumour subtypes are driven by different gene expression patterns and pathways.

86 3.1.2. Molecular prognostic factors for CPC

CPCs exhibit clinical heterogeneity as observed by a wide range of responses to treatment, disease dissemination and survival outcomes. Recent evidence has demonstrated that clinical heterogeneity is driven by the molecular heterogeneity of these tumours. Our study has demonstrated the presence of distinct subgroups of CPCs with unique copy number profiles. Hyperdiploids were characterized by numerous chromosome-wide gains and acquired uniparental disomy (aUPD), while hypodiploids by numerous chromosome-wide losses. Two other reports have confirmed this distinction between hyperdiploid and hypodiploid CPCs using an independent tumour cohort and a different genotyping platform (Ruland et al. 2014; Japp et al. 2015). However, no significant correlation was found between ploidy status and survival outcomes in either tumour cohort. Our study also assessed the gene expression and epigenome signatures of CPCs and found that these clustered according to p53 status. When survival outcomes were compared, we found significant survival reduction in patients whose tumours had increased mutant p53 copies—indicating an increase in mutant p53 dosage. Although these findings have not as yet been replicated in a larger independent CPC cohort, it is crucial to determine whether mutant p53 dosage is an accurate prognostic factor for CPCs as this will enable the creation of tailored therapeutic strategies for high-risk patients. Other molecular abnormalities have been associated to changes in CPC patient survival; however, these have not been validated in independent cohorts. Ruland et al. found an association between loss of 12q and shorter survival, whereas Rickert et al. found a correlation between gain of 9p and loss of 10q and longer survival (Ruland et al. 2014; Rickert et al. 2002). The formation of large collaborative networks that encourage tumour sample accrual and the development of standard analytical platforms and methods to assess genomic abnormalities will promote a more unified understanding of common pathways involved in CPC development and patient prognosis. It is necessary to assess the effectiveness of potential novel markers of CPC prognosis in a prospective analysis using a multi-institutional framework in order to make significant recommendations that will improve the outcomes of CPC patients with refractory disease.

87 3.2. The use of next-generation sequencing to investigate CPC aetiology

Matched tumour-blood CPC pairs with very high quality are scarce. This limitation has impeded the analysis of these rare tumours using next generation sequencing (NGS) tools, which provide a more detailed overview of recurrent alterations at base pair resolution.

3.2.1. Whole genome sequencing reveals very low mutation rates in CPCs

A recent study on which we collaborated, described the use of whole genome sequencing to assess 4 tumour-blood CPC pairs (Tong et al. 2015). Despite the generation of high-quality sequence reads and a coverage of 20x, we found no recurrent point mutations, insertion/deletions or focal CNAs. Paediatric cancers are known to exhibit a very low rate of non-silent point mutations (Downing et al. 2012; Huether et al. 2014; Lee et al. 2012; J. Zhang, Benavente, et al. 2012; Zhang et al. 2013). CPCs exhibited an average of 4.75 non-silent mutations per sample (range: 1-11 mutations/CPC) and an average mutation rate of 0.0013 mutations/Mb, a very low rate even compared to other paediatric tumour types (Lawrence 2011) (Figure 4.2). A Sample Mutations Mutation Rate SJCPC001011 4 0.001/Mb SJCPC001012Spectrum of base1 substitutions0.0003/Mb SJCPC001013observed11 in CPCs0.003/Mb SJCPC001015Spectrum of base3 substitutions0.001/Mb Mean 4.75 0.001/Mb observed in CPCs B 5%

32% 21% 5%

21% 32% 5%

11% 5% 16% 5% 5% 11% 16% [G/A] [C/G] [T/A] [A/G] [A/T] [A/C] [C/T] [G/C] 5% Figure 4.2: Whole genome sequencing reveals that CPCs 5% exhibit very low mutation rates and no recurrent point mutations [G/A] [C/G] [T/A] [A/G] [A/T] [A/C] [C/T] [G/C]

88 (A) Number of non-silent mutation and mutation rates observed by CPC sample. (B) Frequency of base substitutions observed in CPCs.

We found no significant pattern of base pair substitutions, but a slight enrichment in G>A substitutions in CPC samples (Figure 4.2).

3.2.2. Chromothripsis as a novel mechanism in CPC development

Rausch et al. reported an association with constitutional or early TP53 mutations and chromothripsis, generating a link between LFS and this rare phenomenon (Refer to chapter 1.6.2). These findings led us to investigate the copy number profile of 20 CPC samples for this phenomenon. Twelve samples harboured mutant TP53 (12/20, 60%), while eight carried wildtype TP53 (8/20, 40%). Five of the TP53 mutant CPCs (5/12, 42%) were from LFS patients. No CPC, regardless of TP53 status or association with LFS, exhibited the characteristic copy number oscillations that are unique to chromothripsis (data not published). Next generation sequencing of a small and independent cohort of CPCs, followed by validation using a genotyping microarray revealed that the recurrent CPC sample CP22-TA and a further recurrent sample from the same patient (CP22-TB) exhibited chromothripsis affecting chromosomes 1 and 19 (Figures 4.3 and 4.4). Tong et al. reported a single CPC exhibiting extensive chromosomal rearrangements characteristic of chromothripsis between chromosomes 1 and 19 (Figure 4.3) (Korbel & Campbell 2013)(Refer to chapter 1.6.2 for a more detailed description of chromothripsis). To our knowledge, this is the first time that chromothripsis has been reported in CPCs.

Chr19

CP22-TA

Chr1

Figure 4.3: Whole genome sequencing identifies evidence of chromothripsis event in CPC Recurrent CPC sample (CP22-TA) exhibited massive chromosomal rearrangements between chromosomes 1 and 19, evidence of chromothripsis (Figure obtained from Dr. Xiaotu Ma).

The presence of chromothripsis in these recurrent CPC samples arising from the same patient suggests that this phenomenon may be an important prognostic factor for recalcitrant disease or a mechanism that endows highly aggressive characteristics to these recurrent tumours. The in- depth study of the damaging consequences of shattered chromosomes 1 and 19 in these samples will highlight important genes involved in CPC tumourigenesis that may become altered via this chromosome-shattering event. Chromosome 1, which is recurrently gained in CPCs (Merino et al. 2015)(Refer to chapter 2) has already been involved in CPC tumourigenesis (Tong et al. 2015) (Refer to chapter 3), yet the role of chromosome 19 in this process has not been examined yet. Examining larger CPC cohorts using FFPE-derived DNA and fresh frozen tumour DNA samples for chromothripsis is an ongoing project that will help elucidate the frequency and extent of this event in CPCs.

90 Both CPCs with chromothripsis harboured a wildtype TP53. Although mutations in TP53 have been associated with this phenomenon, finding chromothripsis in wildtype TP53 samples suggests that other TP53-independent mechanisms may be associated with creating an environment of tolerance and increased survival in cells following catastrophic DNA rearrangements. Moreover, the TP53 mutation status of a sample may not accurately represent the status of TP53 pathway activity as other upstream or downstream proteins may still be misregulated in wildtype TP53 samples. Lastly, examining the timing of this phenomenon in a patient’s tumour evolution is another interesting question that can be addressed by examining patient CP22’s primary neoplasm. Preliminary analysis for this sample has not been successful due to low quality of material available. We hope that as novel protocols requiring a lower quality threshold arise, and more effective extraction methods become available, we will be able to decipher the timing of chromothripsis in this patient, and thus propose a model defining the timing of chromothripsis in CPC evolution.

CP22-TA CP22-TB

-1 0 1 -1 0 1 Log-R ratio Log-R ratio

-1 0 1 -1 0 1 Log-R ratio Log-R ratio Figure 4.4: Copy number profile confirms chromothripsis event in CPC affecting chromosomes 1 and 19 Copy number analysis of recurrence samples CP22-TA and CP22-TB by genotyping microarray reveals a distinct oscillating copy number pattern between the short arm of chromosome 1 and chromosome 19, characteristic of chromothripsis

91 3.2.3. Identifying fusion lesions in CPC by RNA sequencing

Recent studies have also shown that paediatric tumours may be affected by epigenetic aberrations and oncogenic fusion events (Parker et al. 2014; J. Zhang, Benavente, et al. 2012; Jones et al. 2009). To understand oncogenic fusion events involved in CPC development, we collaborated with St. Jude Children’s Research Hospital–Washington University Pediatric Cancer Genome Project (PCGP) to perform RNA sequencing (RNAseq) on 26 CPC samples. Nine samples (35%) exhibited at least one fusion event, while 17 (65%) exhibited at least one in- frame fusion event affecting the coding region of target genes. The presence of gene fusions was not dependent on whether the sample was a primary or recurrent lesion. Recurrent lesions had a larger number of fusion events, but identical fusion breakpoints were observed in shared events, suggesting recurrent samples may have been given rise from a therapy-resistant clone found in the primary lesion. After validating the fusion calls with an orthogonal approach (RT-PCR), a single fusion event affecting RAF1 and TMEM40 was observed in two independent CPC samples (Figure 4.5).

Wildtype RAF1

457-648 Deleted part (“tail”)

Fused protein

Figure 4.5: Diagram of RAF1 and RAF1-TMEM40 fused protein found CPC This diagram demonstrates the wildtype structure of RAF1 and the fused protein created by RAF1 and TMEM40 in two independent samples of CPC (Figure modified from Dr. Yongjin Li).

92 Interestingly, the breakpoints in the two samples were very similar. The type of variant generated by this fusion was an in-frame interstitial duplication, which resulted in TMEM40 overexpression (Figure 4.6). To our knowledge, this is the only recurrent fusion observed in CPCs.

CP146-T

CP43-T

CN#gain#

CP43%T'

CP146%T'

Upregula)on, Figure 4.6: Copy number and gene expression microarray analysis validates the molecular changes associated with the RAF1-TMEM40 fusion (A) Copy number analysis demonstrates a duplication in the regions affecting the fusion between RAF1 (interrupted at exon 11) and TMEM40 (B) Gene expression analysis of this region in chromosome 3 reveals overexpression of TMEM40 but not RAF1.

RAF1 (also known as CRAF) is a mitogen-activated protein (MAP) kinase (MAP3K) that regulates cell division cycle, apoptosis, cell differentiation and cell migration by the phosphorylation of MEK1 and MEK2 (Matallanas et al. 2011) . Mutations in RAF1 are very rare.

93 A study reported that only four RAF1 mutations were observed in 545 human cancer cell lines (0.7%), with two cell lines retaining activity levels similar to wildtype RAF1 (Emuss et al. 2005). Mutations were found in cell lines derived from patients with fibrosarcoma, lung carcinoma, lung adenocarcinoma and colorectal carcinoma. Moreover, mutations were observed in a mouse model of chemically induced lung cancer (Storm & Rapp 1993) and in the germline of 2 patients with therapy-related acute myeloid leukemia (t-AML) (ages 43 and 72), suggesting that germline alterations in this gene could predispose individuals to cancer (Zebisch et al. 2006). In these patients, somatic loss of the tumour suppressor RAF kinase inhibitor protein (RKIP) was found to be the second hit necessary for the development of t-AML; however, the expression of RKIP was not affected in the patients’ primary tumour, indicating this is a t-AML specific event (Zebisch et al. 2009). Constitutional mutations in RAF1 account for 8 and 33% of Noonan and LEOPARD syndrome cases for which patients have no other known mutations in the Ras signalling pathway, respectively (Pandit et al. 2007). Rearrangements affecting RAF1 have also been observed in prostate cancer (Palanisamy et al., 2010), and pilocytic astrocytoma (Jones et al. 2009; Tatevossian et al. 2010; Cin et al. 2011; Forshew et al. 2009). Although these fusions were very infrequent, these findings demonstrate that alterations in RAF1, most of which affect the kinase activity of this protein, promote tumourigenesis. To date, there is no evidence that suggests that TMEM40 is involved in cancer development, but its location approximately 7kb upstream of RAF1 may have been responsible for its association with RAF1 in this fusion (Figure 4.7).

Figure 4.7: Physical location of RAF1 and TMEM40 on chromosome 3 Figure obtained from UCSC Genome Browser on Human Feb. 2009 (GRCh37/hg19) Assembly

Although the role of TMEM40 as part of the fused protein needs to be further examined, it is hypothesized that the unaffected transmembrane domain, which appears to be under the regulation of the RAF1 regulatory domain, may endow cells carrying this fusion a growth advantage. Investigating the incidence of this fusion protein in a larger cohort of CPCs is

94 necessary, as these data will elucidate the involvement of the Ras signalling pathway in CPC development and guide the use of known kinase inhibitors as therapeutic options for CPC patients. Although there was only one recurrent fusion observed in our CPC cohort, we examined the pathways affected by the rearrangements observed in our tumours. We identified that the most recurrent pathways affected by rearrangements in CPCs were the kinase, cell proliferation and apoptosis pathways, highlighting the crucial role of these pathways for tumour development (Figure 4.8).

CP1-T CP3-T CP43-T CP5-TC CP52-T CP141-T CP41-T CP50-T CP41-TA CP143-T CP54-T CP144-T CP5-TA CP5-TB CP51-T CP22-TA CP22-TB CP66-T CP56-T CP139-T CP146-T CP158-T CP15-T CP140-T CP171-T CP58-T NGS platform Details TP53 status x x x x x x x x x x x x x x x Structural Chromosome 1 Variations Chromosome 19 Kinases x x x x Cell proliferation x x Apoptosis x x x Transmembrane x x NFkB x x (pathways)

Fusion targets Fusion DNA damage x Lysosomes x Figure 4.8: Fusion profile of CPCs and affected pathways found by whole genome sequencing (WGS) and RNA sequencing analysis (RNAseq) NGS platform, pink: RNAseq, purple: RNAseq and whole genome sequencing performed. TP53 status, light green: TP53 mutant CPC, white: TP53 wildtype CPC. Black squares represent positive hits for the affected pathways and chromosome alterations listed on the left, while white squares represent the lack of positive hits.

Future sequencing studies of CPTs should focus on the elucidation of transcript rearrangements as these will help identify transcriptomic aberrations generated by the intrinsic chromosomal instability present in these tumours and help refine the aetiology of CPC. Sequencing will provide a more detailed layer of genomic information that will improve the biological understanding of these tumours and generate a more refined list of gene candidates for further targeted drug studies.

4. Mutant p53 & CPC prognosis 4.1. Mutant p53 dosage

Mutations in TP53 are highly prevalent in CPC and have been associated with a reduction in patient survival (Tabori et al. 2010); however, TP53 mutation status alone does not explain the phenotype differences and clinical variability observed in CPC patients. Desiring to understand

95 these discrepancies, we investigated the allele-specific copy number of TP53 in addition to its mutation status. We found that by inferring the copy number of chromosome 17, identifying regions exhibiting loss of heterozygosity and assessing the mutation status of TP53 we were able to quantify the number of mutated copies of TP53 in a sample. Although our CPC cohort size was small, we found a significant reduction in survival in patients whose tumours harboured a greater number of mutant TP53 molecules (Merino et al. 2015). In our cohort, CPC patient outcomes, as measured by overall and event-free survival, had a dosage-dependent association with the number of mutant p53 copies in a tumour. Our study was the first to report a copy number-dependent effect of mutant p53 on patient survival, more specifically in paediatric cancer. We suggested that the devastating effects of greater amounts of mutant p53 in the tumour may be attributed to a greater burden generated by the oncogenic gain-of-function (GOF) activities of mutant p53. Work in mice conducted in Dr. Lozano’s laboratory has investigated the GOF activities of mutant p53. They found that the stabilization of mutant p53 was necessary for the manifestation of GOF phenotypes, which included increased tumour incidence and metastasis. In vivo mutant p53 stabilization was observed with the deletion of murine double minute 2 (Mdm2)and p16INK4a, and achieved in the presence of cellular stress pathways that regulate wildtype p53 activity. All of these have shown to contribute to exacerbated tumour phenotypes (Suh et al. 2011). Stabilization of mutant p53 by exogenous stress was also observed in breast cancer cell lines and patient samples with poor outcomes (Bouchalova et al. 2014). The increased stability achieved by mutant p53 is key to its ability to exhibit GOF properties and accumulate to high levels (Frum & Grossman 2014). It is thus expected that when stability is achieved, elevated levels of mutant p53 may lead to an even more aggressive tumour phenotype and worse patient outcomes. Prospective studies should assess the validity of using the number of mutant p53 copies as a prognostic marker in CPCs, and other tumours associated with mutant p53. In addition, investigating the functional status of MDM2 and CDKN2A may further refine the prognostic value of the number of mutant p53 copies, as tumours harbouring a loss of any of these two genes may exhibit an even more aggressive tumour phenotype. Further studies should also focus on the mechanisms by which elevated levels of mutant p53 lead to aggressive disease. In vitro and in vivo experiments should be performed to assess whether an increase in mutant p53 GOF activity is correlated to the number of mutated p53 copies observed in silico in our study. This

96 would validate our initial findings. As well, assessing what other factors contribute to the increased stabilization of mutant p53 will help identify therapeutic targets for these tumours.

4.2. Mutant p53 function

Transcriptomic and epigenomic data clustered CPCs into significantly distinct subgroups according to TP53 status (Merino et al. 2015). TP53 mutant CPCs harbouring missense mutations in the DNA binding domain shared similar transcriptomic and epigenomic profiles. However, one sample with a mutation in the SH3-like Proline-rich domain (c.290T>A; p.V97D) exhibited a different transcriptomic profile that clustered with TP53 wildtype CPCs. Although protein sequence prediction models considered this mutation as deleterious and its transactivation activity as non-functional; prediction of structure function according to the 3D structure of the mutant p53 protein suggests this mutant form of p53 has retained structural functionality (www.p53.iarc.fr). This finding demonstrates how crucial it is to investigate the functional effects generated by different mutation types in TP53. Since we know that missense mutations that abrogate p53 function are associated with poor patient outcomes (Tabori et al. 2010; Merino et al. 2015), it is imperative to not only define CPCs according to a binary TP53 mutation state (mutant vs. wildtype), but also according to the functional capacity of the mutated p53 protein in order to guide the creation of more accurate stratification strategies. Moreover, knowing that some of these recurring mutations exhibit known deleterious GOF activities (Muller & Vousden 2014), more comprehensive classification criteria based on functional changes attributed to a tumour’s particular TP53 mutation need to be defined to better stratify patients and more accurately identify high-risk groups.

5. Mouse models The newly generated transgenic mouse model of CPC (Tong et al. 2015) is a powerful tool available to conduct functional studies that will help elucidate the biological mechanisms underlying these insidious tumours. This model has already revealed potential mechanisms driving tumourigenesis, such as aberrant DNA repair and chromatin remodelling, as well as novel oncogenes associated with CPC development and maintenance. Animal models are also helpful in drug screening studies, and drive the development of more effective and less toxic agents to be implemented in the clinic. Richard Gilbertson’s group at St. Jude Children’s

97 Research Hospital is conducting a drug screening study where thousands of drug compounds are tested to assess which drugs CPC responds to best. Using the new transgenic CPC mouse model and derived cell lines, drug kinetics are tested in vivo to identify drugs that are most effective at reducing tumour burden. This ongoing study is only possible due to the availability of the recently generated transgenic CPC mouse model. Two new genetically engineered mouse models of CPC have been concurrently reported from the laboratories of Robert Wechsler-Reya at Sanford-Burnham Medical Research Institute in La Jolla, CA, and Martine Roussel at St. Jude Children’s Research Hospital in Memphis, TN (data not published). Crossing mice carrying a Cre-driven constitutively active mutation in Myc (T58A) and a deletion in Tp53, the two research groups found the formation of CPCs in the hindbrains of mice approximately 11 weeks after birth. While Roussel’s group used the promoter of the Brain Lipid Binding Protein (Blbp), a gene expressed during development in radial glia, to drive Cre-recombinase expression, the Wechsler-Reya’s group used the Atonal Homolog 1 (Atoh1) promoter, a gene associated with neuronal cell lineage, suggesting that CPC may originate from any of these cell lineages during development. Wechsler-Reya’s group also observed the formation of CPPs when a wildtype Tp53 background was used. Were this model found to be a faithful replicate of its human counterpart, it will be an interesting tool that will help us understand CPP development and elucidate the molecular differences and similarities between CPP and CPC formation. It will also help understand whether CPC is a more aggressive form of CPP that progresses from a benign CPP state, although our initial high-throughput analyses refutes this hypothesis (Merino et al. 2015). Ongoing work is comparing these two Myc-driven mouse models to human CPTs in order to assess whether they represent an accurate model of human disease. We are conducting comparative copy number and gene expression cross-species analyses (Figure 4.9) to examine whether these two new CPC mouse models faithfully recapitulate the molecular aberrations observed in human disease, and to understand the role of Myc activation on CPT development.

Human CPC

Myc-Mouse CPC

Figure 4.9: Copy number profile of human and Myc-driven mouse CPC Mouse genome is plotted on syntenic regions of human genome (“humanized mouse genome”). Copy number profiles for CPC samples from both species reveal broad copy number aberrations throughout the genome.

The prospective projects suggested in this thesis will promote an even better understanding of the biological mechanisms governing the growth and development of CPCs, a disease for which there are no tailored therapeutic approaches, and for which mortality rates remain high.

Chapter 5: Appendix 1- Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers

This chapter has been reproduced with permission from Nature Genetics

Adam Shlien, Brittany B. Campbell, Richard de Borja, Ludmil B. Alexandrov, Daniele Merico, David Wedge, Peter Van Loo, Patrick S. Tarpey, Paul Coupland, Sam Behjati, Aaron Pollett, Tatiana Lipman, Abolfazi Heidari, Shriya Deshmukh, Naama Avitzur, Bettina Meier, Moritz Gerstung, Ye Hong, Diana M. Merino, Manasa Ramakrishna, Marc Remke, Roland Arnold, Gagan B. Panigrahi, Neha P. Thakkar, Karl P. Hodel, Erin E. Henninger, A. Yasemin Göksenin, Doua Bakry, George S Charames, Harriet Druker, Jordan Lerner-Ellis, Matthew Mistry, Rina Dvir, Ronald Grant, Ronit Elhasid, Roula Farah, Glenn P. Taylor, Paul C. Nathan, Sarah Alexander, Shay Ben Shachar, Simon C Ling, Steven Gallinger, Shlomi Constantini, Peter Dirks, Annie Huang, Stephen W. Scherer, Richard G Grundy, Carol Durno, Melyssa Aronson, Anton Gartner, M Stephen Meyn, Michael D Taylor, Zachary F Pursell, Christopher E Pearson, David Malkin, P Andrew Futreal, Michael R Stratton, Eric Bouffet, Cynthia Hawkins, Peter J Campbell and Uri Tabori. On behalf of the Biallelic Mismatch Repair Deficiency Repair Consortium. Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers. Nature Genetics 2015, 47: 257-62

My role in this study was to assess the copy number (CN) profile of tumour samples from bMMRD patients, sporadic tumour samples, and controls in order to quantify the percentage of CN changes per tumour sample. Although this publication does not contribute to the theme of this thesis directly, it still demonstrates my productivity and one of the many collaborations formed during my PhD training.

1. Abstract DNA replication−associated mutations are repaired by two components: polymerase proofreading and mismatch repair. The mutation consequences of disruption to both repair components in humans are not well studied. We sequenced cancer genomes from children with inherited biallelic mismatch repair deficiency (bMMRD). High-grade bMMRD brain tumors exhibited massive numbers of substitution mutations (>250/Mb), which was greater than all childhood and most cancers (>7,000 analyzed). All ultra-hypermutated bMMRD cancers acquired early somatic driver mutations in DNA polymerase ε or δ. The ensuing mutation

100 signatures and numbers are unique and diagnostic of childhood germ-line bMMRD (p < 10−13). Sequential tumor biopsy analysis revealed that bMMRD/polymerase-mutant cancers rapidly amass an excess of simultaneous mutations (~600 mutations/cell division), reaching but not exceeding ~20,000 exonic mutations in <6 months. This implies a threshold compatible with cancer-cell survival. We suggest a new mechanism of cancer progression in which mutations develop in a rapid burst after ablation of replication repair.

2. Introduction Genetic changes underlie the development of neoplasia and can take many forms, including point mutations, copy number alterations and rearrangements. Irrespective of their type, somatic changes are caused, or allowed to persist, because of deficiencies in DNA repair. However, our understanding of the relationship between specific DNA repair defects and the resultant mutation type is limited. This is primarily because sporadic cancers are heterogeneous and involve dysfunction in multiple DNA-repair defects and types of mutation that accumulate over many years. In contrast, early-onset cancers from patients with inherited DNA-repair deficiency can offer an unobstructed view of the mutation types and secondary pathways that drive carcinogenesis. bMMRD is a childhood cancer syndrome characterized by early-onset cancers in various organs caused by biallelic mutations in the mismatch repair pathway1. This is one of two components that prevent point mutations during replication. The second safeguard resides within the intrinsic proofreading ability of the DNA polymerases (ε and δ). Although correction of replication errors has been studied in model systems, the consequences of its complete absence have not been investigated in humans. To study the secondary alterations and mutation types that lead to bMMRD cancer, we analyzed genomes of 17 inherited cancers (from 12 patients), using genome and exome sequencing and microarrays. Additionally, we sequenced non-neoplastic tissues from patients for which matched tumor was not available (total of 16 exomes and 1 genome from 18 patients). We compared the mutational landscape of bMMRD tumors to a reference data set of >7,000 cancers2.

3. Results Of the 17 bMMRD cancers, all 10 malignant brain tumors exhibited an extremely large number of point mutations (average, 7,911 coding mutations; 249 mutations/Mb). This mutation

101 frequency is in stark contrast to that in other pediatric cancers (0.61 mutations/Mb) and in all other sequenced cancers, irrespective of age of onset (Fig. 1a), and we therefore refer to these cancers as ‘ultra-hypermutated cancers’. Ultra-hypermutated bMMRD cancers contained an even distribution of mutations throughout the genome (Fig. 1b) and displayed other features distinct from other sequenced tumors: they were almost completely devoid of the copy number alterations typically observed in childhood brain cancers (Fig. 1c,d and Figure S.5.1) and were microsatellite stable (unlike mismatch repair (MMR)-mutated sporadic cancers3).

102 Figure 5.1: Somatic mutation frequency in bMMRD ultra-hypermutated cancers (a) Mutation frequencies in bMMRD ultra-hypermutated malignant brain tumors (mean = 249 mutations/Mb) compared to a diverse cohort of other childhood brain cancers (<1 mutation/Mb), childhood cancers (<1 mutation/Mb) and adult cancers (<10 mutations/Mb). Data on the y axis are log-transformed. bMMRD from exome sequencing unless denoted with “(g)”. Cancers with >100 mutations/Mb are highlighted in orange. Cancer-type abbreviations and number of samples per group (representative cancers from ref. 2) are indicated in the Supplementary Note. For box plots, the thick horizontal line (green or black) indicate median, and upper and lower hinges correspond to the 25th and 75th percentiles. (b) Mutation frequencies, as calculated in 1-Mb bins, are plotted for each chromosome and reveal no evidence of localized hypermutation (kataegis22). The red dashed line indicates 100 mutations/Mb. (c) Total copy number changes in sporadic glioblastomas (n = 578, average = 55.48 changes/sample) and bMMRD glioblastomas (n = 4, average = 1.5 changes/sample). The Mann-Whitney non- parametric test was used to calculate P values. (d) Copy number profile of two bMMRD brain tumors. Chromosomal log R ratios and copy number plots are shown; in each plot, purple indicates total copy number and blue indicates copy number of the minor allele. (e) Tumor mutation frequency (log scale) as a function of age. bMMRD cancers are marked in orange. All other pediatric cancers are in green. The probability of observing ultra-hypermutation in a child with sporadic non-bMMRD was <10−13. Reprinted by permission from Nature Genetics (Shlien et al., 2015; 47:257-62). Copyright 2015.

These were the only cases of ultra-hypermutation in our analysis of childhood cancers; the probability of observing this staggering number of mutations in a child with sporadic non- bMMRD disease was <10−13 (Fig. 1e). Indeed, in a previous large profiling study of pediatric high-grade gliomas, three ultra-hypermutated tumors also had been found to harbor germline biallelic mismatch repair mutations4. To our knowledge, ours is the first report of a tumor genome profile that can be used to infer germline mutational status. DNA for non-neoplastic samples from patients with bMMRD (lymphocytes, n = 16) and controls had similar numbers of variants (Figure S.5.2). This contrasts with the high mutation load observed in non-neoplastic tissues of MMR-deficient mice5. To test whether this absence of excessive mutation was a result of residual mismatch repair activity, we evaluated MMR activity in non-neoplastic cells derived from patients with bMMRD using the G•T mismatch assay6,7. All cells lacked protein expression of the corresponding mutant MMR gene and were completely deficient in G•T mismatch repair (Figures S.5.3 and S.5.4). Therefore, it appears that secondary mutations are required to cause the ultra-hypermutation seen in bMMRD tumors. We examined each cancer for somatic mutations in the replication repair machinery. All ultra-hypermutated cancers harbored mutations in polymerase ε (Pol ε, POLE, 7/10 tumors) or polymerase δ (Pol δ , POLD1, 3/10 tumors). Nonmalignant tissue and non−ultra-hypermutated bMMRD cancers lacked mutations in polymerase genes (n = 17 and 7 tumors, respectively; Fig. 5.2a). These proofreading polymerases work cooperatively with MMR proteins. POLE was the most frequently mutated DNA repair gene in bMMRD (Figure S.5.5). Nonetheless, with a somatic mutation every ~5 kb, a large proportion of protein-coding genes would be expected to carry mutations and the presence of a high number of polymerase mutations could in theory be due to chance. We therefore undertook several analyses to address the potential role that polymerase mutations

103 might assume in bMMRD cancers. POLE mutations affected critical amino acid residues. Each bMMRD cancer with a mutation in POLE (bMMRD/POLE cancer; 7/7 tumors) harbored a mutation affecting the exonuclease domain or domains important to the intrinsic proofreading activity of Pol ε (Figures S.5.6.a). Residues S459 and S461 (substituted in one tumor and three tumors, respectively) are in the ExoIII exonuclease motif, adjacent to one of the exonuclease catalytic residues conserved in all polymerases (D462)8. F104 is in an F/YxPYFY motif conserved in both human Pol ε and δ (ref. 9). S297 and P436 closely flank the ExoI and ExoII motifs and are absolutely conserved in all POLE orthologs8. To assess how the proofreading capability of Pol ε was affected by these POLE mutations, we introduced mutations conferring the most frequent substitutions (S459F and S461P) into a construct encoding the Pol ε catalytic subunit10 and performed an in vitro assay measuring mutation accumulation. These mutations resulted in the loss of replication fidelity and a high mutation rate11 (Figure 5.2b). POLD1 mutations also affected conserved domains (Figure S.5.7). C319 and L606 were mutated in one and two tumors, respectively. C319 is immediately adjacent to one of the exonuclease catalytic sites in Pol δ (E318) within the ExoI motif. The recurrent L606 substitution (L606M) is in motif A of the polymerase domain12–14; the identical substitution in yeast Pol δ (L612M) has been shown to dramatically reduce replication fidelity15. POLE and POLD1 mutations are likely to have occurred as early events in each cancer’s life history (Figures 5.2c and S.5.8). All samples harbored vast numbers of genomic mutations at a low allelic fraction (<20%; subclonal variants), indicating a recent and explosive accumulation of mutations after POLE or POLD1 mutation. These data, coupled with the presence of mutator polymerases and high mutation loads, suggest that mutant Pol ε and Pol δ are drivers in bMMRD. To understand the extent to which polymerase defects affect the overall bMMRD genome, we explored their mutational profiles in greater depth. bMMRD/POLE cancers exhibited a mutational ‘signature’ that is specific to cancers with a mutant gene encoding DNA polymerase ε (ref. 2; Fig. 5.2b). Of the six classes of base substitution, the mutational landscape of these cancers was characterized by C>T and C>A changes (Figure 5.2a, mutation type). They also contained very few C>G mutations. We then analyzed the sequence context of each substitution based on the flanking 3’ and 5’ bases (Fig. 5.2a, mutation context). All bMMRD/POLE cancers had a common signature with highly distinctive features: most C>A and T>G transversions were followed by a 3’ thymine (>85% of C>A and >70% of T>G substitutions). Notably, these were frequently preceded by a thymine, that is, C>A at TCT and T>G at TTT (>30% of C>A and T>G substitutions). Thus, the genome

104 in tumors with mutant POLE incurred a signature mutation spectrum. bMMRD/POLD1 cancers displayed their own idiosyncratic mutational pattern, which differed markedly from that of the bMMRD/POLE tumors (Figure S.5.9). These cancers exhibited many C>A and C>T mutations, as well as an excess of T>A and T>C mutations, especially as compared to the bMMRD/POLE cancers (Fig. 5.2a, mutation context).

Figure 5.2: Consequences of polymerase mutations in bMMRD cancers (a) POLE and POLD1 driver mutations (blue circles) found in ultra-hypermutated malignant brain tumors (n=10) but not in low- grade tumors, other cancer types or benign polyps from patients with bMMRD (n=7). Mutation type indicates the simple

105 mutation spectrum of ultra-hypermutated cancers. Mutation context shows base substitution mutation spectra for POLE and POLD1 cancers. Each of the 96 mutated trinucleotides are represented in a heatmap. The base located 5’ to each mutated base is shown on the vertical axis and 3’ base is on the horizontal axis. C>G mutations are not included in this plot as there were too few of them. (b) Pol ε in vitro error rates for the tumor mutation hotspots. A reversion substrate similar to the CTàAT transversion error hotspot seen in human tumors was generated. This substrate only scores CTàAT transversions. Mutant frequencies were calculated for wild type (3 mutants out of 9,927 plaques scored), S461P, and S459F along with error rates for each (***, p ≤ 0.0001). Pvalues were calculated using chi-square tests. Error rates are the averages of two experiments, each conducted with independent DNA and enzyme preparations for each construct tested. (c) Timing of POLE and POLD1 mutation with respect to all other mutations in the genome, shown as a histogram. Clonality analysis of the ultra-hypermutated bMMRD tumors revealed that the driver polymerase mutations occurred in the earliest possible clone (arrows). The variant allele fractions of somatic mutations per tumor are plotted (i.e. the number of reads reporting a mutation). Samples with whole-genome sequencing data are indicated “(g)”. Reprinted by permission from Nature Genetics (Shlien et al., 2015; 47:257-62). Copyright 2015.

Although C>A changes feature prominently in both POLE and POLD1 cancers, these occur in a completely different sequence context: bMMRD/POLD1 cancers are characterized by C>A mutations at CCN, with a particular enrichment for C>A at CCT. To our knowledge, this is the first report of somatic POLD1 driver mutations in ultra-hypermutated cancers. This mutation spectrum was also recently found in engineered yeast with the same mutated residue (pol 3 L612M)16. This signature occurred early and matches that in previously described sporadic POLE-related cancers2. Next, we looked for the same signature in other common cancers. Substitutions in POLE had a similar effect on bMMRD cancers as tumors with known somatic MMR and POLE mutations (colorectal, and endometrial tumors17-18). Similar to data for our cohort, MMR/POLE cancers were hypermutated, contained few copy number changes and were microsatellite stable17 (Figure S.5.10). Lastly, unbiased hierarchical clustering of trinucleotide sequence revealed that all ultra-hypermutated bMMRD/POLE tumors grouped into a single cluster with sporadic MMR/POLE endometrial and colorectal cancers (Figure 5.3).

106

Figure 5.3: Mutation spectrum of inherited and sporadic cancers Shown is a cluster analysis based on mutation context of bMMRD cancers and sporadic colorectal and endometrial tumors. The 96 possible trinucleotides of all substitutions are on the y axis, and individual samples are on the x axis. All bMMRD/POLE cancers clustered together with ultra- hypermutated POLE endometrial and colorectal cancers of adulthood. Arrows indicate the mutation contexts enriched in POLE cancers. In the heatmap, colors represent the proportion of each trinucleotide (−log10 transformed) in that sample, such that the most common mutation types are in dark blue and the least common mutation types are in white. Reprinted by permission from Nature Genetics (Shlien et al., 2015; 47:257-62). Copyright 2015.

Our data suggest that bMMRD cancers bear the imprint of polymerase defects, in the form of a massive number of highly specific substitutions acquired in a short time. We wondered whether we could use these unique features of bMMRD/polymerase cancers to study the accumulation,

107 frequency and threshold (upper limit) of mutation accrual in cancer. We compared the mutation load of bMMRD and bMMRD/polymerase cancers with that found in other human cancers (Figure 5.4a). bMMRD tumors lacking a POLE mutation had an approximately 5- to 10-fold increase in mutation load relative to pediatric cancers from the same tissue type with intact MMR, whereas bMMRD/polymerase tumors displayed a 230-fold increase in exonic mutations relative to bMMRD alone.

Figure 5.4: Mutation threshold and rate in cancers with mismatch and polymerase mutations (a) Mutation burden in pediatric and adult cancers with and without mutations in MMR genes and/or polymerase defects (left). Box plots indicate the actual number of exonic mutations in human cancers, as determined by exome or genome sequencing. Box plots indicate 25th and 75th percentiles, and whiskers denote the upper and lower number of mutations per exome. Rate of mutation accumulation of serially collected bMMRD tumors (right). For three patients, the mutation frequency of tumor pairs was contrasted with the number of new exonic mutations shown (box) and the tumor type indicated (below the ovals). PXA,

108 pleomorphic xanthoastrocytoma; GBM, glioblastoma multiforme. (b) Surveillance MRI scans. Reprinted by permission from Nature Genetics (Shlien et al., 2015; 47:257-62). Copyright 2015. This mutation prevalence is similar to what has been previously reported in model organisms with engineered deficiencies in each pathway9. The genomes of inherited and sporadic MMR/polymerase cancers reached the same mutation level and did not exceed it (1−2 × 104) despite decades of difference in ages of onset (Figure 5.4). Finally, to study the rate of mutation accumulation and to establish the time required to develop bMMRD/polymerase cancer, we used specimens collected as part of our clinical surveillance protocol19. Sequential magnetic resonance imaging (MRI) and endoscopies enabled determination of tumor appearance and the collection of multiple specimens from carriers, which we used to measure the accumulation of somatic mutations over time. bMMRD gastrointestinal polyps (MMR10; Figure 5.4a) did not contain a higher mutational load than adult polyps20. Normal gastrointestinal mucosa and blood-derived DNA collected at different times contained few new mutations. It is reasonable to suggest that in the absence of secondary polymerase variants cancers mutate steadily, requiring many years to develop sufficient drivers. In contrast, serial analysis of recurrent brain cancers revealed rapid accumulation of mutations over a very short time. A bMMRD/POLE mutant glioblastoma that transformed from a low-grade glioma with wild-type POLE (MMR1) had similar mutations (including TP53 mutation: p53 substitution R273C) but exhibited 13,620 new exonic substitutions. Moreover, although we observed 72,354 new substitutions (by whole-genome sequencing) between a primary and relapsed bMMRD/POLE mutant glioblastoma (MMR94), the total amount of mutations remained the same within the threshold (Fig. 5.4a). To quantify mutation accumulation over time, we used repeated MRI data of four glioblastomas that developed from nonvisible masses (over 4−6 months). For example, a tumor of 381 mm3 corresponds to ~35 cell doublings and a mutation load of 21,284 mutations; it would therefore have 608 mutations per cell division (Fig. 5.4b). We appreciate that these are conservative estimates because they do not take into account the last few cycles of tumor growth (in which mutations would be below the detection of high- throughput sequencing) and potential loss owing to death of hypermutant cell clones. However, they are consistent between patients and similar to the numbers postulated from our in vitro polymerase assay. bMMRD/polymerase-mutant cancer that divides every 5−6 days will accumulate a staggering 250−600 mutations per cell cycle, thereby enabling bMMRD/polymerase cancers to acquire sufficient driver mutations in less than 6 months.

109 4. Discussion Our data directly reveal the consequences of complete ablation of replication error repair in human cancer. Once the proofreading ability of the DNA polymerases are lost in a mismatch repair−deficient cell, there is no defense against a rapid and catastrophic accumulation of point mutations (Figure S.5.11). Despite the extreme consequences of absent DNA replication repair, the resulting signatures are similar and consistent: mutations arise throughout the genome in a specific spectrum in the background of a near-diploid genome, and accumulate to a threshold without surpassing it. Ultra-hypermutated cells mutate continuously, potentially generating multiple independent subclones (Figures 5.2c), until confronting a threshold. The high mutation load and threshold may be this cancer’s Achilles’ heel, exploitable for therapeutic intervention.

5. Conclusion This is to our knowledge the first description of a massive simultaneous accumulation of point mutations associated with extremely rapid tumor initiation. The ultra-hypermutated phenotype occurs rapidly and is limited to substitutions, making it distinct from other tumors, which carry a variety of mutation types that typically accumulate in a slow and stepwise manner to provide sufficient clonal advantage21. bMMRD/polymerase-mutant cancers therefore suggest a new and unique mechanism for cancer initiation.

6. Methods 6.1. Patient and sample collection.

Patients were registered as a part of the International Biallelic Mismatch Repair Consortium, which includes multiple centers worldwide. Detailed information on each family and all patients can be found in our previous study23. Following Institutional Research Ethics Board approval, all data were centralized in the Division of Haematology/Oncology at The Hospital for Sick Children (SickKids) and the Familial Gastrointestinal Cancer Registry (FGICR) at the Zane Cohen Centre for Digestive Diseases at Mount Sinai Hospital, in Toronto, Canada. Consent forms were obtained from the parents or guardians, or from the patients, where applicable. Family history, demographic and clinical data were obtained from the responsible physician and/or genetic counselor at the corresponding centers.

110 Tumor and blood samples were collected from the Sickkids tumor bank. The diagnosis of bMMRD was made when a germ-line biallelic mutation in any of the four MMR genes (MLH1, MSH2, MSH6 and PMS2) was confirmed by sequencing in a clinically approved laboratory. The surveillance protocol developed by our group23 was used to gather clinical information, such as time to tumor development, and tumor samples from biopsies that were used for sequencing.

6.2. Copy number analysis.

DNA from bMMRD tumors was hybridized to Affymetrix SNP 6.0 arrays (n = 4 tumors). Copy number segmentation was performed using the Single Nucleotide Polymorphism-Fast Adaptive States Segmentation Technique (Biodiscovery Nexus Copy Number 7.5). This hidden Markov model−based approach was used with a significance threshold for segmentation set at 5.0 × 10−7 also requiring a minimum of three probes per segment and a maximum probe spacing of 1,000 kbp between adjacent probes before breaking a segment. The log ratio thresholds for copy gain and copy loss were set at 0.1 and −0.15, respectively. We compared bMMRD tumor copy number profiles to that of 578 glioblastoma samples previously hybridized to the same array platform. To account for possible differences in segmentation algorithms in the two data sets, copy number segments (either gains or losses) smaller than 5 Mb were excluded. The frequency of segments was compared using a Mann-Whitney nonparametric test.

7. References:

1. Wimmer, K. et al. Diagnostic criteria for constitutional mismatch repair deficiency syndrome: suggestions of the European consortium ‘care for CMMRD’ (C4CMMRD). J. Med. Genet. 51,

355–365 (2014).

2. Alexandrov, L.B. et al. Signatures of mutational processes in human cancer. Nature 500, 415–

421 (2013).

3. Liu, B. et al. Mismatch repair gene defects in sporadic colorectal cancers with microsatellite instability. Nat. Genet. 9, 48–55 (1995).

111 4. Wu, G. et al. The genomic landscape of diffuse intrinsic pontine glioma and pediatric non- brainstem high-grade glioma. Nat. Genet. 46, 444–450 (2014).

5. Narayanan, L., Fritzell, J.A., Baker, S.M., Liskay, R.M. & Glazer, P.M. Elevated levels of mutation in multiple tissues of mice deficient in the DNA mismatch repair gene Pms2. Proc.

Natl. Acad. Sci. USA 94, 3122–3127 (1997).

6. Thomas, D.C., Roberts, J.D. & Kunkel, T.A. Heteroduplex repair in extracts of human HeLa cells. J. Biol. Chem. 266, 3744–3751 (1991).

7. Panigrahi, G.B., Slean, M.M., Simard, J.P., Gileadi, O. & Pearson, C.E. Isolated short

CTG/CAG DNA slip-outs are repaired efficiently by hMutSbeta, but clustered slip-outs are poorly repaired. Proc. Natl. Acad. Sci. USA 107, 12593–12598 (2010).

8. Henninger, E.E. & Pursell, Z.F. DNA polymerase epsilon and its roles in genome stability.

IUBMB Life 66, 339–351 (2014).

9. Preston, B.D., Albertson, T.M. & Herr, A.J. DNA replication fidelity and cancer. Semin.

Cancer Biol. 20, 281–293 (2010).

10. Korona, D.A., Lecompte, K.G. & Pursell, Z.F. The high fidelity and unique error signature of human DNA polymerase epsilon. Nucleic Acids Res. 39, 1763–1773 (2011).

11. Bebenek, K. & Kunkel, T.A. Analyzing fidelity of DNA polymerases. Methods Enzymol.

262, 217–232 (1995).

12. Ghodgaonkar, M.M. et al. Phenotypic characterization of missense polymerase-delta mutations using an inducible protein-replacement system. Nat. Commun. 5, 4990 (2014).

13. Venkatesan, R.N. et al. Mutation at the polymerase active site of mouse DNA polymerase delta increases genomic instability and accelerates tumorigenesis. Mol. Cell. Biol. 27, 7669–

7682 (2007).

112 14. Schmitt, M.W. et al. Active site mutations in mammalian DNA polymerase delta alter accuracy and replication fork progression. J. Biol. Chem. 285, 32264–32272 (2010).

15. Nick McElhinny, S.A., Gordenin, D.A., Stith, C.M., Burgers, P.M. & Kunkel, T.A. Division of labor at the eukaryotic replication fork. Mol. Cell 30, 137–144 (2008).

16. Lujan, S.A. et al. Heterogeneous polymerase fidelity and mismatch repair bias genome variation and composition. Genome Res. 24, 1751–1764 (2014).

17. Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330–337 (2012).

18. Kandoth, C. et al. Integrated genomic characterization of endometrial carcinoma. Nature 497,

67–73 (2013).

19. Durno, C.A. et al. Oncologic surveillance for subjects with biallelic mismatch repair gene mutations: 10 year follow-up of a kindred. Pediatr. Blood Cancer 59, 652–656 (2012).

20. Nikolaev, S.I. et al. A single-nucleotide substitution mutator phenotype revealed by exome sequencing of human colon adenomas. Cancer Res. 72, 6279–6289 (2012).

21. Jones, S. et al. Comparative lesion sequencing provides insights into tumor evolution. Proc.

Natl. Acad. Sci. USA 105, 4283–4288 (2008).

22. Nik-Zainal, S. et al. Mutational processes molding the genomes of 21 breast cancers. Cell

149, 979-993 (2012).

Chapter 6: Appendix 2- Supplemental Data

1. Supplemental Experimental Procedures 1.1. Chapter 2: Defining the genomic landscape of choroid plexus tumors

Quantity and quality measures on nucleic acids DNA and RNA samples were quantified by spectrophotometry using the NanoDrop 1000 instrument (Thermo Scientific, Waltham, USA) and RNA integrity was measured by electrophoresis using the Experion instrument (BioRad, Hercules, USA).

Microarray processing Three microarrays platforms: Affymetrix Human Exon 1.0ST, Illumina Human Methylation450 BeadChip, and Affymetrix Genome-wide human SNP 6.0 were hybridized and processed according to the manufacturer’s instructions at the Toronto Centre for Applied Genomics (TCAG) in Toronto, Canada. Samples hybridized to Affymetrix OncoScan Express 2.0 microarrays were processed following the manufacturer’s instructions by the Affymetrix processing facility in Santa Clara, California, USA. Gene Expression parameters used in Partek GS6.5: Probes to import: Interrogating Probes and Control Probes. Probe filtering: Include Core Meta-Probeset: HuEx-1_0-st- v2.r2.dt1.hg18.core.mps. Pre-background Adjustment: Adjust for GC Content. Background correction: RMA Background correction. Normalization: One-Way Anova normalization, Log Probes using Base: 2, Probeset Summarization: Median Polish. ANOVA statistical test was used to identify gene expression differences between subgroups. Partek® Genomics Suite™ 6.5 (PGS) (Partek Inc. St. Louis, USA) was used for initial copy number analysis. Inferred copy number states (regions of deletion or duplication) were derived from Partek's genomic segmentation algorithm (Partek, St Louis,MO) such that segmented regions were significantly different from neighbouring regions (p<0.001) and contained 10 or more probes, with a signal to noise ratio of 0.3. Segmentation algorithm assumes ploidy for all non-gender chromosomes is diploid. Diploid samples were those with a copy number between 1.75-2.25. Annotation file: GenomeWideSNP_6.na33.annot.csv

114 Affymetrix Oncoscan FFPE Express 2.0 analysis: Undertaken using the SNP-FASST2 Segmentation Algorithm within Nexus Copy number software (Discovery Edition 7.0, BioDiscovery, Hawthorne, CA) and NCBI genome build 37. The SNP-FASST2 algorithm is an extension of the FASST2 Segmentation Algorithm (a Hidden Markov Model [HMM] based approach). With the SNP-FASST2 algorithm, B-allele frequency probes are assigned to a range of possible states and a combination of the BAF and Log-R states are used to make the final copy number and allelic event calls. The significance threshold for segmentation was set at 1 x 10-7 requiring a minimum of 5 probes per segment and a maximum probe spacing of 1000 Kbp between adjacent probes before breaking a segment. The log ratio thresholds for single copy gain and single copy loss were set at 0.35 and -0.35, respectively. The log ratio thresholds for two or more copy gain and homozygous loss were set at 1.1 and -1.1 respectively. The Homozygous Frequency Threshold was set to 0.95. The Homozygous Value Threshold was set to 0.8. The Heterozygous Imbalance Threshold was set to 0.4. The minimum LOH length was set at 500KB. Manual annotation of diploid chromosomes were performed to account for aneuploidy.

Unsupervised hierarchical clustering PVCLUST was performed in R using 1000 probesets with the largest median absolute deviation (MAD), while NMF was performed in Gene Pattern 2.0 using 5000 probesets with the largest MAD. PVRECT in R used a significance threshold of α=0.95. SigClust was used to test the significance of the clusters identified using NMF. The Rand index was used to assess the concordance between gene expression and methylation subgroup stratification by NMF. Significance was assessed by permutation of samples labels and computed over 10,000 iterations in order to generate a null distribution, as per Mack et al. 2014.

GSEA A defined list of curated pathway gene sets from various online databases was utilized as previously reported in Witt et al., 2011.

Methylation analysis Illumina GenomeStudio software was used for methylation analysis by normalizing data to internal controls and using background subtraction.

115

Statistical tests and methods Continuous variables were assessed for normality, using non-parametric methods: Mann- Whitney or Kruskal-Wallis when comparing two or more than two non-normal sets of data, respectively, and the parametric method: one-way analysis of variance (ANOVA) when comparing normal data. Categorical and binary variables were analyzed with Fisher’s Exact test or two-way ANOVA. Adjustment for multiple testing was performed using the false discovery rate (FDR) method. Statistical analyses of copy number and gene expression were performed in PGS 6.5, whereas methylation was analyzed in Genome Studio. Other statistical analyses were conducted in R (version 3.0.1). FISH Performed on 4-micron sections of formalin fixed paraffin embedded tumor samples using standard methods at The Hospital for Sick Children. BAC clones RP11-112P19 and RP11- 112C15 labeled with spectrum green for 1p34, and RP11-663F24 and RP11-1069P8 labeled with spectrum orange for 1q21.2 (control probes near the centromere) were used. We analyzed 200 nuclei for each specimen.

1.2. Chapter 3: Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes

Mouse tumourigenesis studies Mice harbouring conditional floxed alleles of Tp53, Rb, Pten and ROSAYFP (Chow et al. 2011) were housed and studied under St. Jude IACUC approved protocols. The choroid plexuses of E12.5 mouse embryos were electroporated in utero using pcDNA3.1 plasmids encoding Cre- Recombinase, GFP and/or candidate oncogenes exactly as we described previously (Gibson et al., 2010). Electroporation was performed using a CUY-21 electroporator (Nepa Gene Do., LTD, Tokyo, Japan) with five 50-ms pulses of 50 V with 950-ms intervals. Actively proliferating CPE and CPC cells were labelled in vivo by injection of 0.1mg/kg Bromodeoxyuridine (Brdu, Sigma- Aldrich, St. Louis, MO) two hours prior to euthanizing animals. Lentiviruses encoding the cDNA of Cre-Recombinase, Taf12, Rad54l or Nfyc, or shRNAs targeting these genes (three

116 shRNAs per gene designed using www.dharmacon.gelifesciences.com/design-center), were cloned into lentiviruses, packaged, titrated and injected in utero into the embryonic IV ventricle exactly as we described previously (Robinson et al. 2012). Mismatch control shRNAs were generated by mutating five bases within the 21mer sequence of each shRNA. Orthotopic implants of CPCs were generated by injecting the hindbrains of five to six week-old female CD1 nude mice (www.criver.com) with 5x106 CPC cells transduced with control or shRNA-lentivirus suspended in 10ml matrigel. CPC cells were injected into IV ventricle through cisterna magna using Hamilton syringe.

Immunohistochemistry and in situ hybridization Formalin fixed and paraffin embedded mouse CPCs (or for ATR inhibitor studies, cytospins of cultured CPC cells) were stained with standard immunohistochemical approaches using rabbit polyclonal antibodies to Ki67 (Vector Labs), Phospho-Histone H2A.XSer139 (Cell signalling, Danvers, MA), GFP (Invitrogen, Carlsbad, CA), copGFP (Evrogen, Moscow); or mouse monoclonal antibodies to BrdU (Sigma-Aldrich, St. Louis, MO), Taf12, Nfyc or Rad54l (Abcam, Cambridge, MA). For immunofluorescence staining, second antibodies were conjugated with either Alexa488 or Alexa 594 and sections mounted with Vectashield mounting medium containing DAPI (Vector laboratories, Inc, Burlingame, CA, USA). In situ hybridization (ISH) was performed using standard approaches and probes generated using the open reading frames of Taf12 (cDNA clone MGC:30438 IMAGE:3493047), Nfyc (cDNA clone MGC:28940 IMAGE:4009737), Rad54l (cDNA clone MGC:13963 IMAGE:3987790) and Ttr (generous gift of Dr. Edwin Monuki, IMAGE clone 1078224, Accession Number AA822938). Dual-color fluorescence in situ hybridization (FISH) to estimate DNA copy number was performed on 5 µm paraffin embedded tissue sections with BAC clones encoding Baat, Cdca8 and Stmn1 (BACPAC Resources, Oakland, CA). Probes were labelled with either AlexaFluor-488 or AlexaFluor-555 fluorochromes and nuclei counterstained with DAPI (Vector Labs).

Cell culture Fresh, primary, CPC tumour tissue was dissected from the IV ventricle of mice, minced and disaggregated using hyaluronidase (Atlanta Biologicals, Lawrenceville, GA) and collagenase type IV (Invitrogen, Carlsbad, CA). Disaggregated cells were filtered through a 40 mm cell strainer and transduced with control or shRNA-lentiviruses. The proliferation of single cell

117 suspensions cultured in neurobasal medium, in the presence or absence of the indicated concentration of ATR inhibitors was recorded using CellTiter-Glo as described (Gibson et al., 2010). Apoptosis was assessed in single cell suspensions using the CaspaseGlo assay (see Supplementary Experimental Procedures).

Quantitative Reverse Transcriptase PCR (qRT-PCR) Total cDNA was generated from cells and analysed using qRT-PCR and standard approaches. Primers pairs employed in assays included TAF12 (AGCCATCTCATCCTTGGATTTTT : ACGACCAAAATGCTCTTCCTACA), NFYC (AGGACAGATCCAGACACTTGCTAC : GATTGGCTGACTGAATAAACATGG), RAD54L (CCTGGTGAAGAACTGGTACAATG : ATGACCAGTCCAACATTTCCTTT) and GAPDH (AGGTCGGTGTGAACGGATTTG :TGTAGACCATGTAGTTGAGGTCA).

Whole genome sequencing analysis High quality DNA from 12 tumour-blood pairs was sent to the Hartwell Center for Bioinformatics and Biotechnology at St. Jude Children’s Research Hospital. After quality control measures, only 4 pairs were available for whole genome sequencing using a paired-end sequencing strategy as described in Zhang et al. 2012 (J. Zhang, Ding, et al. 2012). Detection of somatic single nucleotide variations (SNVs), indels, structural variations, and focal copy number changes from whole genome sequencing (WGS) was performed as previously described (Parker et al. 2014). Briefly, WGS mapping, coverage, quality assessment, single-nucleotide variants (SNVs), detection of small insertions or deletions (indels), tier annotation for sequence mutations, and prediction of adverse effects of missense mutations were processed using the same analysis pipeline (Robinson et al. 2012; J. Zhang, Ding, et al. 2012; J. Zhang, Benavente, et al. 2012). SVs were analysed by using the program CREST (Wang et al. 2011); CNAs were identified by comparing the read depth of matched tumour vs. normal tissue and were analyzed by using the CONSERTING algorithm (COpy Number SEgmentation by Regression Tree In Next-Gen sequencing). The reference human genome assembly GRCh37-lite was used to map all samples. We used the program cghMCR (http://master.bioconductor.org/packages/release/bioc/html/cghMCR.html) that calculates a modified version of the GISTIC score (Beroukhim et al. 2007) to identify recurrent copy number gain or loss. To minimize the effect caused by sample(s) with unusually

118 high/low seg.mean scores (e.g. very high-level amplification due to the formation of episomes), the seg.mean log2ratio values were set to have a ceiling of -3 for deletions and 4 for amplifications. The default parameters (method=”sum”, threshold=c(- 0.2, 0.2)) in the “cghMCR” package were used for all analysis. Tumour purity was estimated by accounting for loss of heterozygosity (LOH) and copy number change, and mutant allele fraction (MAF) of SNVs, as previously described (Chen et al. 2013). Chromothripsis was analyzed using the criteria proposed by Korbel and Campbell (Korbel & Campbell 2013). Oscillating pattern of copy-number states were manually inspected using the CIRCOS plots and statistical tests were applied to evaluate clustering of breakpoints and randomness of DNA fragment joins. For detecting clustering of breakpoints, we applied the Bartlett's goodness-of-fit test for exponential distribution to see if there was a strong departure from the null hypothesis, consistent with the chromothripsis hypothesis. For assessing randomness of DNA fragment joins, we applied the goodness-of-fit tests to evaluate if there was no significant departure from the multinomial distribution with equal probabilities, consistent with the chromothripsis hypothesis.

Validation of somatic lesions The genomic coordinates of putative somatic alterations identified by WGS, including SNVs, SVs, and indels, were used to generate a Nimbelgen Seqcap EZ bait set for enrichment of targeted regions (Roche). The baits were hybridized to Truseq sample libraries (Illumina) prepared from amplified genomic DNA (Roche). Pooled samples were sequenced on a HiSeq 2000 by using the paired-end multiplexed 100-cycle protocol. Resulting data were converted to FASTQ files by using CASAVA 1.8.2 (Illumina) and mapped with the Burrows-Wheeler Aligner (BWA) prior to pipeline analysis. Of the 409 somatic SNVs identified in the four cases, we were able to design a validation assay by custom capture for 392. Of these, 365 were validated as somatic mutations (overall validation rate, 93%).

1.3. Chapter 5: Appendix 1- Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers

Microsatellite instability testing. Microsatellite instability testing was performed in a clinically approved laboratory, as described previously23.

119 High-throughput sequencing, read mapping and identification of mutations. Tumors were sequenced using Agilent's exome enrichment kit (Sure Select V4; with >50% of baits above 25× coverage) or by whole genome sequencing to a depth ~40× (Supplementary Fig. 12). In all cases but one, the matched blood-derived DNA was also sequenced. Base calls and intensities from the Illumina HiSeq 2500 were processed into FASTQ files using CASAVA. The paired-end FASTQ files were aligned to the genome (to UCSC's hg19 GRCh37) with BWA24 (v0.5.9). Duplicate paired-end sequences were removed using Picard MarkDuplicates (v1.35) to reduce potential PCR bias. Aligned reads were realigned for known insertion/deletion events using SRMA25 (v0.1.155). Base quality scores were recalibrated using the Genome Analysis Toolkit26 (v1.1-28). Somatic substitutions were identified using MuTect27 (v1.1.4) or CaVEMaN22. Mutations were then filtered against common single-nucleotide polymorphisms (SNPs) found in dbSNP (v132), the 1000 Genomes Project (Feb 2012), a 69-sample Complete Genomics data set, and the Exome Sequencing Project (v6500). Mutation signature profiles were extracted using the single base substitution and the corresponding tri-nucleotide sequence context (i.e., reference base at mutation position and its 5′ and 3′ neighbors).

Comparison of bMMRD mutation frequency to sporadic cancers. Mutation frequencies (substitutions per Mb) for bMMRD tumors were calculated from genome or exome data as per previous publications2 and data on sporadic cancer, including age of onset, were obtained from ref. 2. Data shown in Figure 1 are from ref. 2 and from brain tumors sequenced at SickKids.

Validation of substitution mutations. Putative driver mutations in POLE and POLD1 were validated by Sanger sequencing (Supplementary Fig. 13).

Western blotting for MMR protein expression in non-neoplastic biallelic MMR mutant cells. Cell extracts were prepared as described7 and 40 µg of HeLa, wild-type lymphoblast, LoVo, MMR8 lymphoblast and MMR10 lymphoblast cell extracts were loaded in each well. Simultaneous western blotting for human MSH2, MSH3, MSH6 and actin was carried out as described7, 28. Another membrane was simultaneously probed for human PMS2 with 1/100 dilution of anti-PMS2 (BD Pharmingen 556415), human MLH1 with 1/500 dilution of anti- MLH1 (BD Pharmingen 554073) and actin. Both immunoblots were incubated in HRP- conjugated sheep anti-mouse secondary antibody, and chemiluminescence signals were

120 generated using Biorad Clarity Western ECL substrate. Images were captured on VWR CA11006-128 films with multiple exposures.

G•T mismatch repair reactions and repair efficiencies. Repair reactions were carried out as described previously6, 7. Briefly 20 fmol of circular substrate carrying a G•T mismatch and a nick 5′ to the mismatch was incubated with whole cell extracts (2−4 mg/ml of proteins), NTPs, dNTPs, creatine kinase and creatine phosphate for 1 h at 37 °C. Reactions were stopped in 2 mg/ml proteinase K, 2% SDS, 50 mM EDTA, pH 8.0, for 1 h followed by phenol-chloroform extraction. Mixtures were subjected to enzymatic purification kit of Qiagen and mini elute column. Products were eluted with 15 µl of elution buffer and digested with XmnI to linearize the substrate, and HindIII to assess whether correct repair had occurred. Products were resolved on 1% agarose gels, and probed by Southern blotting for quantitative analysis. Membranes were probed with radioactive probe and quantification was performed with a Typhoon FLA 9500 phosphorimager. Repair efficiency is the proportion of radiointensity of the repair products relative to all fragments.

Purification of polymerase ε. An expression vector encoding residues 1−1,189 of the catalytic subunit of human Pol ε was used in site-directed mutagenesis reactions to change Ser461 to proline. Human Pol ε was prepared as described10. Briefly, the human Pol ε was coexpressed in autoinduction medium with pRK603, which allows coexpression of TEV protease, at 25 °C until the culture was saturated. Peak fractions from the HisTrap column were pooled, dialyzed into 50 mM HEPES, pH 7.5, 1 mM DTT, 5% glycerol and bound to SP sepharose. Bound protein was eluted with a 0−1 M with NaCl gradient. Peak fractions were pooled, dialyzed into 50 mM Tris, pH 7.5, 1 mM DTT, 5% glycerol, 100 mM NaCl and bound to Q Sepharose. Bound protein was eluted with a 100 mM–M M NaCl gradient. Peak fractions were pooled, concentrated and passed through a pre-equilibrated Superdex200 size exclusion column. Fractions containing the purified 140 kDa protein were pooled, dialyzed into 50 mM Tris, pH 8.0, 1 mM DTT, 5% glycerol and aliquots were frozen and stored at −80 °C. The data for S459F have been described previously29.

LacZ in vitro mutant frequency and error-rate calculations. The lacZ in vitro forward mutation assay was performed essentially as described previously11. Briefly, double-stranded M13mp2 DNA containing a 407-nt ssDNA gap was used as a substrate

121 in reactions containing 0.15 nM DNA, 50 mM Tris-Cl, pH 7.4, 8 mM MgCl2, 2 mM DTT, 100 µg/ml BSA, 10% glycerol, 250 µM dNTPs and 1.5 nM Pol ε-M630G at 37 °C. Completely filled product was transfected into Escherichia coli cells, which were used to determine the frequency of light blue and colorless plaques that occurred as a result of mutations arising during DNA synthesis. In this assay, accurate DNA synthesis yields dark blue plaques. One of the limitations of the forward assay is that sequence context-specific errors can be underestimated if that context is not well represented. To overcome this limitation, we generated a lacZ reversion substrate that only reports on C>A transversions in the CT>AT context. To generate the reversion substrate, we used site-directed mutagenesis to change A−11 to C−11 and prepared gapped substrate. The C−11−containing substrate gives rise to light blue plaques. C−11 A−11 transversion mutations are scored as dark blue plaques. A pilot sequencing study indicated that 100% of these revertant plaques contained the C−11 A−11 transversion mutation. LacZ mutant frequencies were calculated from combining at least two independent experiments. DNA from mutant plaques was purified and the lacZ gene was sequenced. Error rates were calculated according to the following equation: error rate (per nucleotide synthesized) = ((number of mutants of a particular class) × (mutant frequency)) / ((number of mutations sequenced) × (0.6) × (number of detectable sites)). The data for S459F have been described previously29.

Clustering of cancers by mutation spectra. Data from bMMRD were combined with somatic substitutions from sporadic endometrial cancers (n = 248) and colon cancers (n = 215), obtained from the TCGA (specifically, the Uterine Corpus Endometrial Carcinoma (UCEC) and the Colon Adenocarcinoma (COAD) studies). Only data sequenced on the Illumina platform were included. Only somatic substitutions were included. That is, insertions and deletions were discarded as were point mutations found in the 1000 Genomes Project (Feb 2012), a 69-sample Complete Genomics data set, and the Exome Sequencing Project (v6500). Data was then reannotated with ANNOVAR2 to remain consistent with the annotations used on the bMMRD samples. Substitutions were grouped on the basis of their 3′ and 5′ bases into 96 possible trinucleotide categories that were used for mutation spectrum analysis (Fig. 2) and clustering (Fig. 3). In the clustering analysis, the color of each grid represents the proportion of that trinucleotide in the sample (−log10 transformed). Pairwise comparisons were performed between samples, the

122 Euclidean distance of the trinucleotide proportions was determined, and clustering was performed using the Divisive Analysis clustering algorithm. Mutation frequencies were calculated (using 30 Mb as capture size). Samples with greater than 100 mutations/Mb were designated as ultra-hypermutated.

Calculation of mutation rate from repeated brain MRI scans of bMMRD patients. Calculation of mutation per cell cycle were performed based on established formulas. An example for one sample is provide (D132; Supplementary Table 3): Diameter of tumor from MRI = 45 mm Radius of tumor = 22.5 mm Estimated tumor volume = 4/3 π (0.0075)3 = 1.8 × 10−6 mm3 Diameter of average animal cell = 15 µm Radius of average animal cell = 0.0075 mm Volume of average animal cell = 4/3 π (0.0075)3 = 1.8 × 10−6 mm3 Estimated number of cells in tumor = 44,712/1.8 × 10−6 = 2.65 × 1010 cells Estimated number of cell divisions tumor has undergone = 2x = 2.65 × 1010

Where and x ≈ 35 cell divisions. Total mutations = 21,284 Number of cell cycles = 35 Mutations per cell cycle = 21,284/35 = 608 mutations per cell cycle

Accession codes European Genome-phenome Archive (EGA): EGAD00001000369 and EGAS00001001112.

2. Supplemental Figures 2.1. Chapter 3: Cross-species genomics identifies TAF12, NFYC and RAD54L as novel choroid plexus carcinoma oncogenes

123

Figure S3.1: A new mouse model of CPC Expression of the recombined ROSAYFP fluorescent lineage tracing allele in whole (A) and microscopic (B) preparations of postnatal day (P) 0 hindbrain choroid plexus following in utero electroporation with Cre-recombinase at embryonic day (E) 12.5 (scale bar=50mm). Recombined (C), but histologically normal (D) and Ttr+ (E) choroid plexus persisting in adult CPE (scale bar=15mm). In utero electroporation of hindbrain choroid plexus of Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP E12.5 embryos generated large YFP+ tumours (F,G) that recapitulate the histology (H); reduced Ttr (I) and increased Cytokeratin 8 (J) expression; high proliferation rate (K); and ultrastructural features (L,M) of human CPC (scale bar F to K=20mm; L to M=2mm). N, polymerase chain reactions of recombined (RC) alleles in mouse CPC (T) and intact (WT) alleles in normal tissue (N). O, Kaplan-Meier survival curves of E12.5 mice harbouring the indicate alleles that were Cre-electroporated in utero. P, unsupervised hierarchical clustering of gene expression profiles of mouse: CPC; E12.5 and adult CPE (eCP and aCP); WNT and Sonic Hedgehog (SHH) medulloblastomas; P7 dorsal brainstem (P7 DBS); P7 (P7 CB) and E12.5 cerebellum (eCB); and E12.5 lower rhombic lip (eLRL). Q, GSEA of ‘Lein choroid plexus markers’ in CPC reporting normalized enrichment score (NES).

124

Figure S3.2: Serial analysis of choroid plexus transformation in mice (A) In utero electroporation of the hindbrain choroid plexus of Tp53flx/flx ; Rbflx/flx; Ptenflx/flx ; ROSAYFP E12.5 embryos resulted in progressive expansion of YFP+ choroid plexus (top row, scale bar=50mm) that displayed: dysplasia (second row, hematoxylin and eosin [H&E]; dotted line and arrows indicate disruption of normal single layer epithelium; scale bar=10mm); loss of Ttr expression (third row, in situ hybridization; scale bar=20mm) and increasing proliferation (fourth row, nuclear BrdU incorporation marked with arrows [note dotted line encompasses same region in P0 H&E stain] scale bar=10mm); accumulation of DNA DSB (gH2ax stain marked with arrows; scale bar=10mm); and gain of 4qC6-qE2 (note two separate FISH probes used targeting Stmn1 and Cdca8 relative to control Baat at 4qB1). Graphs to the right report the quantification of BrdU incorporation (B), gH2ax stain (C) and gain of 4qC6-qE2 (D) in recombined (YFP+) cells. Mann Whitney, *= P<0.05, ***= P<0.0005.

125

Figure S3.3: Functional in utero assessment of 1p31.3-ter/4qC6-qE2 candidate CPC oncogenes (A) E12.5 choroid plexus was co-electroporated with plasmids encoding candidate and GFP to identify electroporated cells. Gels below show seven examples of 21 reverse-transcriptase (RT) PCR results of cells transfected with control plasmid or oncogene candidates with or without RT. (B) Sections from each animal were analysed both for morphological change (top, H&E) and proliferation of electroporated cells (bottom, BrdU+/GFP+; scale bar=10mm). Only Taf12, Nfyc and Rad54l demonstrated dysplasia and aberrant proliferation relative to the other 18 candidates, Pgd shown as an example. (C) Graph reporting the percentage of electroporated (GFP+) and proliferating choroid plexus epithelium cells. *P<0.05, **=P<0.005, Mann-Whitney.

126

Figure S3.4: Taf12, Nfyc and Rad54l are required to initiate and maintain CPC in mice (A) Reverse transcriptase PCR of CPC cells transduced with control or gene targeted shRNA lentivirus. (B) In vitro survival of CPC cells transduced with the indicated shRNA lentivirus. ***=P<0.0005, relative to controlshRNA cells. (C) Caspase 3/7 assays of control and shRNA-transduced CPC cells. *=P<0.05, relative to controlshRNA cells. (D) Kaplan-Meyer survival curves of mice implanted with CPC cells transduced with the indicated shRNA lentiviruses. *=P<0.05, **=P<0.005, ***=P<0.0005, Log-Rank relative to controlshRNA mice. (E) Kaplan-Meyer survival curves of Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP mice electroporated at E12.5 with Cre-recombinase and simultaneously injected in the hindbrain with the indicated shRNA (P<0.05, Log-Rank comparison of the three survival curves). (F) Morphology (H&E), lentiviral transduction (GFP), and Cre- recombination (RosaYFP) of hyperplastic choroid plexus mass in adult Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP mouse electroporated at E12.5 with Cre-recombinase and simultaneously injected in the hindbrain with Taf12shRNA (scale bar=20mm).

127

Figure S3.5: Taf12, Nfyc and Rad54l expression promote CPC in mice (A) Kaplan-Meyer survival curves of Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP mice injected in the hindbrain at E12.5 with Taf12, Nfyc and Rad54l expressing lentiviruses following sham (no Cre) or Cre-electroporation (Cre). Survival of Cre- electroporated mice not receiving lentivirus is also shown (Cre only). (B) Left: macroscopic, direct YFP fluorescence of the hindbrains of adult Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP mice receiving hindbrain injections of the indicated lentivirus at E12.5. Right panels show images of H&E stained or immunofluorescence stained sections of the corresponding hindbrain, left (scale bars=50mm). Kaplan-Meyer survival curves of Tp53flx/flx ; Rbflx/flx ; Ptenflx/flx ; ROSAYFP mice injected in the hindbrain at E12.5 (C) or P1 (D) with the indicated lentiviruses. Statistics report the Log-Rank comparison of the

128 survival curves. D, in situ hybridization (top and middle) and immunofluorescence (bottom) of Taf12, Nfyc and Rad54l expression in CPCs induced by lentiviral injection. All scale bars=50mm.

2.2. Chapter 5: Appendix 1- Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers

Figure S.5.1: Somatic rearrangements in bMMRD cancers Circos plots depicting whole-genome profile of two tumors. Each arc represents a single interchromosomal translocation (purple), large deletion (green), tandem duplication (black) or inversion (orange and blue). Rearrangements were determined by the Breakpoint via assembly algorithm (Brass1).

129

Figure S.5.2: Number of single-nucleotide variations (SNVs) in non-neoplastic bMMRD DNA SNVs were determined by standard target capture exome sequencing of blood-derived DNA from 16 bMMRD patients and 35 controls (including individuals unrelated (n = 11) to the patients as well as from parental DNA (n = 24)). Common SNVs found in public databases (dbSNP, 1000 Genomes Project or the NHLBI ESP Project from >6,500 samples) were removed. Shown are the numbers of rare SNVs in patients with bMMRD relative to control samples. Error bars represent s.e.m.

130

Figure S.5.3: MMR protein expression in non-neoplastic biallelic MMR mutant cells by protein blotting (A) For mutSα proteins MSH2 and MSH6. (B) For MutLa proteins PMS2 and MLH1. The lymphoblasts used were LoVo as a negative control, MMR10 and MMR8 lymphoblasts from patients with bMMRD, HeLa cells and normal lymphoblasts as a positive control. Each lane was separated by 8% Tris-glycine SDS-PAGE. Protein blotting was simultaneously carried out for mismatch repair proteins and actin.

131

Figure S.5.4: In vitro GT mismatch repair assay (A) A substrate containing a single GT mismatch was designed keeping a nick several hundred nucleotides away from the mismatch, as shown by the arrow. HindII restriction digestion, which cannot cleave the GT mismatch at its site but can efficiently cleave its repaired AT site, was used for measurement of mismatch repair ability. (B) Reactions were performed as described previously (Panigrahi et al., 2005). GT repaired products were digested by XmnI and HindIII. If repaired correctly, it would be sensitive to HindIII enzyme. The repaired products are marked by the brackets. The digested products were run on a 1% agarose gel and analyzed by Southern hybridization and quantification by Typhoon FLA 9500 Phosphorimager. The bar graphs represent three separate experiments. The last lane contains 50% of LoVo and 50% of MMR10 cell extracts. The protein concentration of the cell extracts was corrected for the purpose of quantification, and the graph was plotted accordingly.

132

Figure S.5.5: Frequency of somatic mutations in DNA replication and repair genes Shown is the proportion of ultra-hypermutated cancers that contain somatic non-synonymous mutations in genes known to be involved in DNA replication or DNA repair. POLE, the most frequently mutated gene, is highlighted.

133

134

Figure S.5.6: Putative polymerase driver mutations found in ultra-hypermutated bMMRD tumors A multispecies alignment for POLE is shown, with each putative driver polymerase mutation indicated (red box). The alignment was performed using BLAST and NCBI’s CDD database. Amino acids deemed critical for catalysis are highlighted2–4 (yellow). (a) Mutated polymerase ε residues are shown, and the exonuclease domain is highlighted (horizontal blue line). The Exo motifs are indicated in orange. Asterisks denote recurrent mutations also found in colorectal or endometrial cancers: mutations at residues F104, P436 and S459 were previously found in colorectal cancers5,6, and S297 was previously found in endometrial cancer7. Polymerase active site mutations have also been found in the germ line of individuals predisposed to colorectal adenomas and carcinomas8. Note that only positions 83–127 and 268–472 of POLE are shown. (b) Examples of somatic POLE passenger mutations, which are not conserved.

Figure S.5.7: Putative polymerase driver mutations found in ultra-hypermutated bMMRD tumors. A multispecies alignment for POLD1 is shown, with each putative driver polymerase mutation indicated (red box). The alignment was performed using BLAST and NCBI’s CDD database. Amino acids deemed critical for catalysis are highlighted (yellow). Mutated polymerase D residues are shown, and the PolII and ExoI motifs are highlighted (horizontal green and orange lines, respectively). Note that only positions 581-624 and 303-348 of POLD1 are shown.

135

Figure S.5.8: Timing of the polumerase E mutational signature in bMMRD/POLE cancers The top panel shows the variant allele fraction of all somatic point mutations in four bMMRD/POLE cancer genomes (similar to Fig. 2d). The percentages of POLE signature mutations are plotted in the lower panel. That is, each point represents the proportion of NpCpT > NpApT (red) or NpTpT (teal) at a given allele fraction. Note that the POLE signature is established early in each cancer and that it continues and is relatively unchanged across the ilfe history of the cancer.

136

Figure S.5.9: Correlation of mutation signatures Shown is the similarity of mutation type for all ultra-hypermutated brain cancers (n=10). Mutations were classified according to their 5’ and 3’ flanking bases (forming a trinucleotide). The proportions of each trinucleotide within all possible mutation classes were compared (that is, the proportion of each of 16 trinucleotides within the 6 classes of substitution: C>A, C>G, C>T, T>A, T>C and T>G). Colors represent the Pearson’s correlation between samples.

137

Figure S.5.10: Mutation burden in sporadic colorectal and endometrial cancers (A) The total number of somatic mutations in sporadic colorectal and endometrial cancers with either MMR mutations, POLE mutations or both. (B) The relationship between mutation burden and copy number. Colon cancers are plotted on the top row, and endometrial cancers are plotted on the bottom row. In all panels, each dot represents one tumor specimen. The red dots represent the genotype of interest. That is, on the left, red dots are samples with mutation in an MMR gene but not in POLE; in the middle, red dots represent POLE-mutated samples, without MMR mutations; and, on the right, red dots represent samples with both MMR and POLE mutations, which contain the highest number of point mutations and fewest copy number alterations. Data are from http://www.cbioportal.org/..

138

Figure S.5.11: Suggested model for the mutation signature of bMMRD malignant cancers A multilayer defense for protecting genome stability prevents replication errors in most individuals. An inherited mismatch repair defect leads to the gradual accumulation of mutations and, thus, to increased cancer risk during adulthood. However, the addition of impaired POLE or POLD1 DNA polymerases results in an extremely rapid accumulation of mutations and onset of cancer in young children.

139

Figure S.5.12: Next generation sequencing coverage of bMMRD malignant cancers (A) Coverage of exome sequencing data. Shown is a box plot with each tumor's coverage, for all bases in the exome target. The red dotted line shows the minimum coverage of 25× required for downstream analysis. (B) Coverage of genome sequence data. Shown is the average coverage for all tumor genomes sequenced. The red dotted line shows the minimum desired coverage of 30×.

140

141

Figure S.5.13: Validation of POLE and POLD1 mutations Putative driver mutations in (A) POLE and (B) POLD1 were validated by Sanger sequencing using forward and reverse sequencing.

142

References

Achatz, M.I.W. et al., 2007. The TP53 mutation, R337H, is associated with Li-Fraumeni and Li- Fraumeni-like syndromes in Brazilian families. Cancer letters, 245(1-2), pp.96–102.

Achatz, M.I.W., Hainaut, P. & Ashton-Prolla, P., 2009. Highly prevalent TP53 mutation predisposing to many cancers in the Brazilian population: a case for newborn screening? The Lancet Oncology, 10(9), pp.920–925.

Addo, N.K. et al., 2011. Preoperative vincristine for an inoperable choroid plexus papilloma: a case discussion and review of the literature. Case report, 8(2), pp.149–153.

Agamanolis, D.P. & Malone, J.M., 1995. Chromosomal Abnormalities in 47 Pediatric Brain Tumors. Cancer genetics and cytogenetics, 81(2), pp.125–134.

Aguzzi, A., Bradner, S. & Paulus, W., 2000. Choroid plexus tumors. In P. Kleihues & W. Cavenee, eds. World Health Organization Classification of Tumors. Pathology and Genetics of Tumours of the Nervous System. Lyon: IARC Press, pp. 84–86.

Aicardi, J., 2005. Aicardi syndrome. Brain & development, 27(3), pp.164–71.

Alexandrov, L.B. et al., 2013. Signatures of mutational processes in human cancer. Nature, 500(7463), pp.415–421.

Anderson, M.D. et al., 2013. Bevacizumab use in disseminated choroid plexus papilloma. Journal of Neuro-Oncology, 114, pp.251–253.

Awatramani, R. et al., 2003. Cryptic boundaries in roof plate and choroid plexus identified by intersectional gene activation. Nat Genet, 35(1), pp.70–75.

Aylon, Y. & Oren, M., 2011. p53: Guardian of ploidy. Molecular oncology, 5(4), pp.315–323.

143 Baker, D.J. et al., 2009. Whole chromosome instability caused by Bub1 insufficiency drives tumorigenesis through tumor suppressor gene loss of heterozygosity. Cancer cell, 16(6), pp.475–86.

Baker, S., Fearon, E. & Nigro, J., 1989. Chromosome 17 deletions and p53 gene mutations in colorectal carcinomas. Science, 244(4901), pp.217–221.

Baker, S., Preisinger, A. & Jessup, J., 1990. p53 gene mutations occur in combination with 17p allelic deletions as late events in colorectal tumorigenesis. Cancer research, 50, pp.7717– 7722.

Baselga, J. et al., 1996. Phase II study of weekly intravenous recombinant humanized anti- p185HER2 monoclonal antibody in patients with HER2/neu-overexpressing metastatic breast cancer. Journal of Clinical Oncology, 14(3), pp.737–744.

Benatti, P. et al., 2011. Specific inhibition of NF-Y subunits triggers different cell proliferation defects. Nucleic Acids Research, 39(13), pp.5356–5368.

Berger, C. et al., 1998. Choroid Plexus Carcinomas in Childhood: Clinical Features and Prognostic Factors. Neurosurgery, 42(3), pp.470–475.

Beroukhim, R. et al., 2007. Assessing the significance of chromosomal aberrations in cancer: Methodology and application to glioma. Proceedings of the National Academy of Sciences , 104 (50 ), pp.20007–20012.

Bettegowda, C. et al., 2012. Treatment of choroid plexus tumors: a 20-year single institutional experience. Journal of neurosurgery. Pediatrics, 10(5), pp.398–405.

Bhattacharjee, M.B. et al., 1997. Cytogenetic analysis of 120 primary pediatric brain tumors and literature review. Cancer genetics and cytogenetics, 97(1), pp.39–53.

Biegel, J.A., 1999. Cytogenetics and molecular genetics of childhood brain tumors. Neuro- oncology, 1(2), pp.139–151.

144 Bieniossek, C. et al., 2013. The architecture of human general transcription factor TFIID core complex. Nature, 493(7434), pp.699–702.

Blackburn, E.H., 2001. Switching and signaling at the telomere. Cell, 106, pp.661–673.

Blackburn, E.H., 2005. Telomerase and Cancer: Kirk A. Landon--AACR prize for basic cancer research lecture. Molecular Cancer Research, 3(9), pp.477–482.

Blamires, T.L. & Maher, E.R., 1992. Choroid plexus papilloma. A new presentation of von Hippel-Lindau (VHL) disease.,

Bouchalova, P. et al., 2014. Mutant p53 accumulation in human breast cancer is not an intrinsic property or dependent on structural or functional disruption but is regulated by exogenous stress and receptor status. Journal of Pathology, 233(3), pp.238–246.

Brassesco, M.S., Valera, E.T., Becker, A.P., et al., 2009. Grade II atypical choroid plexus papilloma with normal karyotype. Child’s Nervous System, 25(12), pp.1623–1626.

Brassesco, M.S., Valera, E.T., Neder, L., et al., 2009. Polyploidy in atypical grade II choroid plexus papilloma of the posterior fossa. Neuropathology : official journal of the Japanese Society of Neuropathology, 29(3), pp.293–8.

Brinster, R.L. et al., 2015. Transgenic mice harboring SV40 t-antigen genes develop characteristic brain tumors. Cell, 37(2), pp.367–379.

Brown, P. et al., 2004. Molecular mechanisms of cerebrospinal fluid production. Neuroscience, 129(4), pp.955–968.

Bryan, T.M. et al., 1995. Telomere elongation in immortal human cells without detectable telomerase activity. The EMBO journal, 14(17), pp.4240–8.

Bunz, F. et al., 2002. Targeted Inactivation of p53 in Human Cells Does Not Result in Aneuploidy. Cancer Research , 62 (4 ), pp.1129–1133.

145 Cannon, D.M. et al., 2015. Choroid plexus tumor epidemiology and outcomes : implications for surgical and radiotherapeutic management. Journal of neuro-oncology, 121, pp.151–157.

Castelo-Branco, P. et al., 2013. Methylation of the TERT promoter and risk stratification of childhood brain tumours: an integrative genomic and molecular study. The lancet oncology, 14(6), pp.534–42.

Chango, A. et al., 2006. Quantitative methylation-sensitive arbitrarily primed PCR method to determine differential genomic DNA methylation in down syndrome. Biochemical and Biophysical Research Communications, 349(2), pp.492–496.

Chen, E.Y. et al., 2013. Cross-Species Array Comparative Genomic Hybridization Identifies Novel Oncogenic Events in Zebrafish and Human Embryonal Rhabdomyosarcoma. PLoS Genet, 9(8), p.e1003727.

Chen, X. et al., 2014. Recurrent somatic structural variations contribute to tumorigenesis in pediatric osteosarcoma. Cell Reports, 7(1), pp.104–112.

Cheng, J. & Haas, M., 1990. Frequent mutations in the p53 tumor suppressor gene in human leukemia T-cell lines. Molecular and cellular biology, 10(10), pp.5502–5509.

Chow, L.M.L. et al., 2011. Cooperativity within and among Pten, p53, and Rb Pathways Induces High-Grade Astrocytoma in Adult Brain. Cancer Cell, 19(3), pp.305–316.

Cin, H. et al., 2011. Oncogenic FAM131B-BRAF fusion resulting from 7q34 deletion comprises an alternative mechanism of MAPK pathway activation in pilocytic astrocytoma. Acta Neuropathologica, 121(6), pp.763–774.

Currle, D.S. et al., 2005. Direct and indirect roles of CNS dorsal midline cells in choroid plexus epithelia formation. Development (Cambridge, England), 132(15), pp.3549–3559.

Cushing, H., 1914. Studies on the Cerebro-Spinal Fluid : I. Introduction. The Journal of Medical Research, 31(1), pp.1–19.

146 Custodio, G. et al., 2011. Increased Incidence of Choroid Plexus Carcinoma Due to the Germline TP53 R337H Mutation in Southern Brazil A. Yeudall, ed. PLoS ONE, 6(3), p.e18015.

Dang, L. et al., 2005. Notch3 signaling initiates choroid plexus tumor formation. Oncogene, 25(3), pp.487–91.

Davidsson, J., Veerla, S. & Johansson, B., 2013. Constitutional trisomy 8 mosaicism as a model for epigenetic studies of aneuploidy. Epigenetics & chromatin, 6(1), p.18.

Dhillon, R.S. et al., 2013. Progression of choroid plexus papilloma. Journal of clinical neuroscience : official journal of the Neurosurgical Society of Australasia, 20(12), pp.1775–8.

Dickens, D. et al., 2005. Successful Treatment of an Unresectable Choroid Plexus Carcinoma in a Patient With Li-Fraumeni Syndrome. Journal of Pediatric Hematology Oncology, 27(1), pp.46–49.

Diller, L., Kassel, J. & Nelson, C., 1990. p53 functions as a cell cycle control protein in osteosarcomas. Molecular and Cellular Biology, 10(11), pp.5772–81.

Dittmer, D. et al., 1993. Gain of function mutations in p53. Nature Genetics, 4(1), pp.42–46.

Dohrmann, G., 1970. The choroid plexus: a historical review. Brain research, 18, pp.197–218.

Donehower, L., Harvey, M. & Slagle, B., 1992. Mice deficient for p53 are developmentally normal but susceptible to spontaneous tumours. Nature, 356(6366), pp.215–221.

Donovan, M.J. et al., 1994. Chromosome aberrations in choroid plexus papillomas. Genes, Chromosomes and Cancer, 11(4), pp.267–270.

Downing, J.R. et al., 2012. The Pediatric Cancer Genome Project. Nat Genet, 44(6), pp.619–622.

Druker, B.J. et al., 2001. Efficacy and Safety of a Specific Inhibitor of the BCR-ABL Tyrosine Kinase in Chronic Myeloid Leukemia. New England Journal of Medicine, 344(14), pp.1031–1037.

147 Dudley, R.W.R. et al., 2015. Pediatric choroid plexus tumors : epidemiology , treatments , and outcome analysis on 202 children from the SEER database. Journal of neuro-oncology, 121, pp.201–207.

Duensing, A. & Duensing, S., 2005. Guilt by association? p53 and the development of aneuploidy in cancer. Biochemical and biophysical research communications, 331(3), pp.694–700.

Dürrbaum, M. et al., 2014. Unique features of the transcriptional response to model aneuploidy in human cells. BMC genomics, 15(1), p.139.

Eden, C.J. et al., 2014. Orthotopic models of pediatric brain tumors in zebrafish. Oncogene, 34(13), pp.1736–1742.

Elmore, L.W. & Holt, S.E., 2001. Telomerase, Aging and Disease. In M. Mattson, ed. Elsevier, pp. 89–102.

Emuss, V. et al., 2005. Mutations of C-RAF are rare in human cancer because C-RAF has a low basal kinase activity compared with B-RAF. Cancer Research, 65(21), pp.9719–9726.

Fang, X. & Zhang, P., 2011. Aneuploidy and tumorigenesis. Seminars in cell & developmental biology, 22(6), pp.595–601.

Fearon, E.R. & Vogelstein, B., 1990. A genetic model for colorectal tumorigenesis. Cell, 61(5), pp.759–767.

Flaherty, K.T. et al., 2010. Inhibition of Mutated, Activated BRAF in Metastatic Melanoma. New England Journal of Medicine, 363(9), pp.809–819.

Flynn, R.L. et al., 2015. Alternative lengthening of telomeres renders cancer cells hypersensitive to ATR inhibitors. Science, 347(6219), pp.273–277.

Forshew, T. et al., 2009. Activation of the ERK/MAPK pathway: a signature genetic defect in posterior fossa pilocytic astrocytomas. The Journal of Pathology, 218(2), pp.172–181.

148 Frappart, P.-O. et al., 2009. Recurrent genomic alterations characterize medulloblastoma arising from DNA double-strand break repair deficiency. Proceedings of the National Academy of Sciences , 106 (6 ), pp.1880–1885.

Frum, R.A. & Grossman, S.R., 2014. Mutant p53 and MDM2 in Cancer. Subcellular Biochemistry, 85, pp.187–197.

Galipeau, P.C. et al., 1996. 17p (p53) allelic losses, 4N (G2/tetraploid) populations, and progression to aneuploidy in Barrett’s esophagus. Proceedings of the National Academy of Sciences of the United States of America, 93(14), pp.7081–7084.

Garber, J.E. et al., 1990. Choroid plexus tumors in the breast cancer-sarcoma syndrome. Cancer, 66(12), pp.2658–2660.

Garritano, S. et al., 2010. Detailed haplotype analysis at the TP53 locus in p.R337H mutation carriers in the population of Southern Brazil: evidence for a founder effect. Human Mutation, 31(2), pp.143–150.

Gemoll, T. et al., 2013. Chromosomal aneuploidy affects the global proteome equilibrium of colorectal cancer cells. Analytical Cellular Pathology, 36(5-6), pp.149–161.

Gessi, M., Giangaspero, F. & Pietsch, T., 2003. Atypical teratoid/rhabdoid tumors and choroid plexus tumors: when genetics “surprise” pathology. Brain pathology (Zurich, Switzerland), 13(3), pp.409–414.

Gibson, P. et al., 2010. Subtypes of medulloblastoma have distinct developmental origins. Nature, 468(7327), pp.1095–1099.

Gocha, A.R.S., Harris, J. & Groden, J., 2013. Alternative mechanisms of telomere lengthening: Permissive mutations, DNA repair proteins and tumorigenic progression. Mutation Research - Fundamental and Molecular Mechanisms of Mutagenesis, 743-744, pp.142–150.

Goldstein, I. et al., 2010. Understanding wild-type and mutant p53 activities in human cancer: new landmarks on the way to targeted therapies. Cancer Gene Therapy, 18(1), pp.2–11.

149 Gonzalez, K.D. et al., 2009. Beyond Li Fraumeni Syndrome: Clinical Characteristics of Families With p53 Germline Mutations. Journal of Clinical Oncology, 27(8), pp.1250–1256.

Gozali, A.E. et al., 2012. Choroid plexus tumors; management, outcome, and association with the Li-Fraumeni syndrome: The Children’s Hospital Los Angeles (CHLA) experience, 1991-2010. Pediatric Blood & Cancer, 58(6), pp.905–909.

Greenberg, M.L., 1999. Chemotherapy of choroid plexus carcinoma. Child’s nervous system : ChNS : official journal of the International Society for Pediatric Neurosurgery, 15(10), pp.571–7.

Griffin, C.A. et al., 1992. Chromosome abnormalities in low-grade central nervous system tumors. Cancer Genetics and Cytogenetics, 60(1), pp.67–73.

Grill, J. et al., 2002. Comparative genomic hybridization detects specific cytogenetic abnormalities in pediatric ependymomas and choroid plexus papillomas. Cancer genetics and cytogenetics, 136(2), pp.121–5.

Haddad, N.M.N., Cavallerano, J.D. & Silva, P.S., 2013. Von Hippel-Lindau Disease: A Genetic and Clinical Review 1. Seminars in Ophthalmology, 28, pp.377–386.

Hanahan, D. & Weinberg, R.A., 2011. Hallmarks of cancer: the next generation. Cell, 144(5), pp.646–674.

Hanel, W. et al., 2013. Two hot spot mutant p53 mouse models display differential gain of function in tumorigenesis. Cell death and differentiation, pp.898–909.

Hanel, W. & Moll, U.M., 2012. Links between mutant p53 and genomic instability. Journal of cellular biochemistry, 113(2), pp.433–9.

Harvey, M., McArthur, M. & Montgomery, C., 1993. Spontaneous and carcinogen-induced tumorigenesis in p53-deficient mice. Nature Genetics, 5, pp.225–229.

Hasselblatt, M. et al., 2006. Identification of novel diagnostic markers for choroid plexus tumors: a microarray-based approach. The American journal of surgical pathology, 30(1), pp.66–74.

150 Heaphy, C.M. et al., 2011. Prevalence of the Alternative Lengthening of Telomeres Telomere Maintenance Mechanism in Human Cancer Subtypes. The American journal of pathology, 179(4), pp.1608–1615.

Hébert, J.M., Mishina, Y. & McConnell, S.K., 2002. BMP Signaling Is Required Locally to Pattern the Dorsal Telencephalic Midline. Neuron, 35(6), pp.1029–1041.

Huang, X. et al., 2009. Sonic hedgehog signaling regulates a novel epithelial progenitor domain of the hindbrain choroid plexus. Development , 136 (15 ), pp.2535–2543.

Huether, R. et al., 2014. The landscape of somatic mutations in epigenetic regulators across 1,000 paediatric cancer genomes. Nature Communications, 5(3630).

Hunter, N.L. & Dymecki, S.M., 2007. Molecularly and temporally separable lineages form the hindbrain roof plate and contribute differentially to the choroid plexus. Development (Cambridge, England), 134(19), pp.3449–3460.

Ikota, H. et al., 2011. Clinicopathological and immunohistochemical study of 20 choroid plexus tumors: their histological diversity and the expression of markers useful for differentiation from metastatic cancer. Brain tumor pathology, 28(3), pp.215–21.

Janssen, S.F. et al., 2013. Gene expression and functional annotation of the human and mouse choroid plexus epithelium. PloS one, 8(12), p.e83345.

Japp, A.S. et al., 2015. High-Resolution Genomic Analysis Does Not Qualify Atypical Plexus Papilloma as a Separate Entity Among Choroid Plexus Tumors. Journal Neuropathology Experimental Neurology, 74(2), pp.110–120.

Jeibmann, A. et al., 2007. Malignant progression in choroid plexus papillomas. Journal of neurosurgery, 107(3 Suppl), pp.199–202.

Jeibmann, A. et al., 2006. Prognostic implications of atypical histologic features in choroid plexus papilloma. Journal of neuropathology and experimental neurology, 65(11), pp.1069–1073.

151 Johanson, C. et al., 2011. The distributional nexus of choroid plexus to cerebrospinal fluid, ependyma and brain: toxicologic/pathologic phenomena, periventricular destabilization, and lesion spread. Toxicologic pathology, 39(1), pp.186–212.

Johnson, R.A. et al., 2010. Cross-species genomics matches driver mutations and cell compartments to model ependymoma. Nature, 466(7306), pp.632–636.

Jones, D.T.W. et al., 2009. Oncogenic RAF1 rearrangement and a novel BRAF mutation as alternatives to KIAA1549:BRAF fusion in activating the MAPK pathway in pilocytic astrocytoma. Oncogene, 28(20), pp.2119–2123.

Judkins, A.R. et al., 2005. INI1 protein expression distinguishes atypical teratoid/rhabdoid tumor from choroid plexus carcinoma. Journal of neuropathology and experimental neurology, 64(5), pp.391–7.

Junttila, M.R. & Evan, G.I., 2009. p53--a Jack of all trades but master of none. Nature reviews. Cancer, 9(11), pp.821–9.

Kamaly-Asl, I.D., Shams, N. & Taylor, M.D., 2006. Genetics of choroid plexus tumors. Neurosurgical focus, 20(1), p.E10.

Karim, A. et al., 2006. Concomitant choroid plexus papillomas involving the third and fourth ventricles: A case report and review of the literature. Clinical Neurology and Neurosurgery, 108(6), pp.586–589.

Kemp, C. et al., 1993. Reduction of p53 gene dosage does not increase initiation or promotion but enhances malignant progression of chemically induced skin tumors. Cell, 74(5), pp.813–822.

Kim, T. et al., 2013. Functional genomic analysis of chromosomal aberrations in a compendium of 8000 cancer genomes Functional genomic analysis of chromosomal aberrations in a compendium of 8000 cancer genomes. Genome Research, 23(2), pp.217–227.

Kloosterman, W.P. et al., 2011. Chromothripsis as a mechanism driving complex de novo structural rearrangements in the germline. Human Molecular Genetics.

152 Kloosterman, W.P., Koster, J. & Molenaar, J.J., 2014. Prevalence and clinical implications of chromothripsis in cancer genomes. Current opinion in oncology, 26(1), pp.64–72.

Koh, E.J. et al., 2014. Clinical outcome of pediatric choroid plexus tumors: retrospective analysis from a single institute. Child’s nervous system : ChNS : official journal of the International Society for Pediatric Neurosurgery, 30(2), pp.217–25.

Korbel, J.O. & Campbell, P.J., 2013. Criteria for inference of chromothripsis in cancer genomes. Cell, 152(6), pp.1226–36.

Krutilkova, V. et al., 2005. Identification of five new families strengthens the link between childhood choroid plexus carcinoma and germline TP53 mutations. European Journal of Cancer, 41(11), pp.1597–1603.

Lafay-Cousin, L. et al., 2011. Choroid plexus tumors in children less than 36 months: the Canadian Pediatric Brain Tumor Consortium (CPBTC) experience. Child’s Nervous System, 27(2), pp.259–64.

Lafay-Cousin, L. et al., 2010. Use of ifosfamide, carboplatin, and etoposide chemotherapy in choroid plexus carcinoma. Journal of Neurosurgery: Pediatrics, 5(6), pp.615–621.

Lang, G.A. et al., 2004. Gain of function of a p53 hot spot mutation in a mouse model of Li- Fraumeni syndrome. Cell, 119(6), pp.861–872.

Lawrence, M.S., 2011. The Spectra of Somatic Mutations Across Many Tumor Types. In 1st annual TCGA symposium.

Lee, R.S. et al., 2012. A remarkably simple genome underlies highly malignant pediatric rhabdoid cancers. Journal of Clinical Investigation, 122(8), pp.2983–2988.

Lehtinen, M.K. et al., 2013. The Choroid Plexus and Cerebrospinal Fluid: Emerging Roles in Development, Disease, and Therapy. Journal of Neuroscience, 33(45), pp.17553–17559.

Lena, G. et al., 1990. Choroid plexus tumours in children. Review of 24 cases. Acta Neurochirurgica, 106(1-2), pp.68–72.

153 Li, M. et al., 2010. The ATM – p53 pathway suppresses aneuploidy- induced tumorigenesis. PNAS, 107(32), pp.14188–14193.

Liau, J.-Y. et al., 2015. Leiomyosarcoma With Alternative Lengthening of Telomeres Is Associated With Aggressive Histologic Features, Loss of ATRX Expression, and Poor Clinical Outcome. American Journal of Surgical Pathology, 39(2), pp.236–244.

Liu, G. et al., 2000. High metastatic potential in mice inheriting a targeted p53 missense mutation. Proceedings of the National Academy of Sciences of the United States of America, 97(8), pp.4174–9.

Van Loo, P. et al., 2010. Allele-specific copy number analysis of tumors. Proceedings of the National Academy of Sciences of the United States of America, 107(39), pp.16910–16915.

Louis, D.N. et al., 2007. The 2007 WHO classification of tumours of the central nervous system. Acta neuropathologica, 114(2), pp.97–109.

Lovejoy, C. a et al., 2012. Loss of ATRX, Genome Instability, and an Altered DNA Damage Response Are Hallmarks of the Alternative Lengthening of Telomeres Pathway. PLoS genetics, 8(7), p.e1002772.

Lu, X. et al., 2001. Selective Inactivation of p53 Facilitates Mouse Epithelial Tumor Progression without Chromosomal Instability Selective Inactivation of p53 Facilitates Mouse Epithelial Tumor Progression without Chromosomal Instability.

Luo, J., Solimini, N.L. & Elledge, S.J., 2009. Principles of cancer therapy: oncogene and non- oncogene addiction. Cell, 136(5), pp.823–37.

Magrangeas, F. et al., 2011. Chromothripsis identifies a rare and aggressive entity among newly diagnosed multiple myeloma patients. Blood, 118(3), pp.675–8.

Mairet-Coello, G., Tury, A. & DiCicco-Bloom, E., 2009. Insulin-Like Growth Factor-1 Promotes G1/S Cell Cycle Progression through Bidirectional Regulation of Cyclins and Cyclin- Dependent Kinase Inhibitors via the Phosphatidylinositol 3-Kinase/Akt Pathway in Developing Rat Cerebral Cortex. The Journal of Neuroscience , 29 (3 ), pp.775–788.

154 Malkin, D. et al., 1990. Germ line p53 mutations in a familial syndrome of breast cancer, sarcomas, and other neoplasms. Science, 250(4985), pp.1233–1238.

Malkin, D. et al., 2001. Tissue-specific expression of SV40 in tumors associated with the Li- Fraumeni syndrome. Oncogene, 20(33), pp.4441–4449.

Mangerel, J. et al., 2014. Alternative lengthening of telomeres is enriched in, and impacts survival of TP53 mutant pediatric malignant brain tumors. Acta neuropathologica, 128(6), pp.853–62.

Marinoni, I. et al., 2014. Loss of DAXX and ATRX are associated with chromosome instability and reduced survival of patients with pancreatic neuroendocrine tumors. Gastroenterology, 146(2), pp.453–460.e5.

Matallanas, D. et al., 2011. Raf family kinases: old dogs have learned new tricks. Genes & cancer, 2(3), pp.232–260.

Merino, D. & Malkin, D., 2014. p53 and Hereditary Cancer. In S. P. Deb & S. Deb, eds. Mutant p53 and MDM2 in Cancer. Springer, pp. 1–16.

Merino, D.M. et al., 2015. Molecular characterization of choroid plexus tumors reveals novel clinically relevant subgroups. Clinical cancer research : an official journal of the American Association for Cancer Research, 21(1), pp.184–193.

Mertens, F. et al., 1995. Recurrent chromosomal imbalances in choroid plexus tumors. Cancer Genetics and Cytogenetics, 80(1), pp.83–84.

Mills, K.D. et al., 2004. Rad54 and DNA Ligase IV cooperate to maintain mammalian chromatid stability. Genes and Development, 18(11), pp.1283–1292.

Misaki, K. et al., 2011. MGMT promoter methylation and temozolomide response in choroid plexus carcinoma. Brain Tumor Pathology.

Morigaki, R. et al., 2012. Choroid plexus papilloma in a girl with hypomelanosis of Ito. Journal of Neurosurgery: Pediatrics, 10(3), pp.182–185.

155 Mosleh, O. et al., 2013. Successful Treatment of a Recurrent Choroid Plexus Carcinoma with Surgery Followed by High-Dose Chemotherapy and Stem Cell Rescue. Pediatric hematology and oncology, (October 2012), pp.1–6.

Muller, P.A.J. & Vousden, K.H., 2014. Mutant p53 in Cancer: New Functions and Therapeutic Opportunities. Cancer Cell, 25(3), pp.304–317.

Muller, P.A.J. & Vousden, K.H., 2013. P53 Mutations in Cancer. Nature cell biology, 15(1), pp.2–8.

Murphy, K.L., Dennis, A.P. & Rosen, J.M., 2000. A gain of function p53 mutant promotes both genomic instability and cell survival in a novel p53-null mammary epithelial cell model. The FASEB Journal, 14(14), pp.2291–2302.

Nabetani, A. & Ishikawa, F., 2011. Alternative lengthening of telomeres pathway: recombination-mediated telomere maintenance mechanism in human cells. Journal of biochemistry, 149(1), pp.5–14.

Nardini, M. et al., 2013. Sequence-Specific Transcription Factor NF-Y Displays Histone-like DNA Binding and H2B-like Ubiquitination. Cell, 152(1), pp.132–143.

Neumann, E. et al., 1993. Cytogenetic analysis of 109 pediatric central nervous system tumors. Cancer genetics and cytogenetics, 71(1), pp.40–9.

Nicholson, J.M. & Cimini, D., 2015. Link between Aneuploidy and Chromosome Instability. In K. W. Jeon, ed. International Review of Cell and Molecular Biology. Elsevier Ltd, pp. 299– 317.

Nielsen, C.M. & Dymecki, S.M., 2010. Sonic hedgehog is required for vascular outgrowth in the hindbrain choroid plexus. Developmental Biology, 340(2), pp.430–437.

Nigro, J., Baker, S. & Preisinger, A., 1989. Mutations in the p53 gene occur in diverse human tumor types. Nature, 342, pp.705–708.

156 Ogiwara, H. et al., 2012. Choroid plexus tumors in pediatric patients. British Journal of Neurosurgery, 26(1), pp.32–7.

Olive, K.P. et al., 2004. Mutant p53 gain of function in two mouse models of Li-Fraumeni syndrome. Cell, 119(6), pp.847–60.

Oren, M. & Rotter, V., 2010. Mutant p53 gain-of-function in cancer. Cold Spring Harbor perspectives in biology, 2(2), p.a001107.

Oromendia, A.B. & Amon, A., 2014. Aneuploidy : implications for protein homeostasis and disease Protein complex. Disease models & mechanisms, 7, pp.15–20.

Ozery-Flato, M. et al., 2011. Large-scale analysis of chromosomal aberrations in cancer karyotypes reveals two distinct paths to aneuploidy. Genome Biology, 12(6), p.R61.

Palmer, C. et al., 2010. Recurrence of fourth ventricular choroid plexus papilloma with late occurrence of a spinal drop metastasis. British journal of neurosurgery, 24(4), pp.482–4.

Pandit, B. et al., 2007. Gain-of-function RAF1 mutations cause Noonan and LEOPARD syndromes with hypertrophic cardiomyopathy. Nature Genetics, 39(8), pp.1007–1012.

Parker, M. et al., 2014. C11orf95–RELA fusions drive oncogenic NF-κB signalling in ependymoma. Nature.

Paulus, W. et al., 1990. Clinicopathological correlations in epithelial choroid plexus neoplasms : a study of 52 cases. Acta Neuropathologica, 80, pp.635– 641.

Paulus, W., Brandner, S. & others, 1999. Synaptophysin in choroid plexus epithelial cells- no useful aid in differential diagnosis. Journal of neuropathology and experimental neurology, 58(10), p.1111.

Petitjean, A. et al., 2007. TP53 mutations in human cancers: functional selection and impact on cancer prognosis and outcomes. Oncogene, 26(15), pp.2157–65.

157 Peyre, M., Bah, A. & Kalamarides, M., 2012. Multifocal choroid plexus papillomas: case report. Acta Neurochirurgica, 154(2), pp.295–299.

Pianetti Filho, G., Fonseca, L.F. & da Silva, M.C., 2002. Choroid plexus papilloma and Aicardi syndrome: case report. Arquivos de neuro-psiquiatria, 60(4), pp.1008–10.

Punnett, H.H. et al., 1994. Cytogenetic Analysis of a Choroid Plexus Papilloma. Genes, Chromosomes and Cancer, 10, pp.282–285.

Rausch, T. et al., 2012. Genome Sequencing of Pediatric Medulloblastoma Links Catastrophic DNA Rearrangements with TP53 Mutations. Cell, 148(1-2), pp.59–71.

De Reuck, J. & Vanderdonckt, P., 1986. Choroid plexus and ependymal cells in CSF cytology. Clinical Neurology and Neurosurgery, 88(3), pp.177–179.

Ribeiro, R. & Sandrini, F., 2001. An inherited p53 mutation that contributes in a tissue-specific manner to pediatric adrenal cortical carcinoma. Proceedings of the National Academy of Sciences, 98(16), pp.9330–5.

Rickert, C.H., 2004. Prognosis-related molecular markers in pediatric central nervous system tumors. Journal of neuropathology and experimental neurology, 63(12), pp.1211–1224.

Rickert, C.H. & Paulus, W., 2001. Tumors of the choroid plexus. Microscopy research and technique, 52(1), pp.104–11.

Rickert, C.H., Wiestler, O.D. & Paulus, W., 2002. Chromosomal imbalances in choroid plexus tumors. The American journal of pathology, 160(3), pp.1105–1113.

Rieber, J. et al., 2009. Novel oncogene amplifications in tumors from a family with Li-Fraumeni syndrome. Genes, Chromosomes and Cancer, 48(7), pp.558–568.

Robinson, G. et al., 2012. Novel mutations target distinct subgroups of medulloblastoma. Nature, 488(7409), pp.43–48.

158 Roland, B. & Pinto, A., 1996. Hyperdiploid karyotype in a choroid plexus papilloma. Cancer Genetics and Cytogenetics, 90(2), pp.130–131.

Ruland, V. et al., 2014. Choroid plexus carcinomas are characterized by complex chromosomal alterations related to patient age and prognosis. Genes Chromosomes and Cancer, 53(5), pp.373–380.

Sáenz Robles, M.T. et al., 1994. Induction versus progression of brain tumor development: differential functions for the pRB- and p53-targeting domains of simian virus 40 T antigen. Molecular and Cellular Biology, 14(4), pp.2686–2698.

Safaee, M. et al., 2012. Choroid plexus papillomas: advances in molecular biology and understanding of tumorigenesis. Neuro-oncology, 94117.

Sakai, W. et al., 2008. Secondary mutations as a mechanism of cisplatin resistance in BRCA2- mutated cancers. Nature, 451(7182), pp.1116–1120.

Sameshima, T. et al., 2010. Choroid plexus papilloma originating in the sella turcica--case report. Neurologia medico-chirurgica, 50(2), pp.144–146.

Sarnat, H.B., 1998. Histochemistry and immunocytochemistry of the developing ependyma and choroid plexus. Microscopy research and technique, 41(1), pp.14–28.

Savage, N.M., Crosby, J.H. & Reid-Nicholson, M.D., 2012. The cytologic findings in choroid plexus carcinoma: Report of a case with differential diagnosis. Diagnostic Cytopathology, 40(1), pp.1–6.

Schmitz, K.-M. et al., 2009. TAF12 Recruits Gadd45a and the Nucleotide Excision Repair Complex to the Promoter of rRNA Genes Leading to Active DNA Demethylation. Molecular Cell, 33(3), pp.344–353.

Scuoppo, C. et al., 2012. A tumour suppressor network relying on the polyamine-hypusine axis. Nature, 487(7406), pp.244–248.

159 Segal, M.B. & Pollay, M., 1977. The secretion of cerebrospinal fluid. Experimental eye research, 25 Suppl, pp.127–148.

Seidinger, A.L. et al., 2011. Association of the highly prevalent TP53 R337H mutation with pediatric choroid plexus carcinoma and osteosarcoma in Southeast Brazil. Cancer, 15(117), pp.2228–2235.

Serrano, M. et al., 1997. Oncogenic ras Provokes Premature Cell Senescence Associated with Accumulation of p53 and p16 INK4a. Cell, 88, pp.593–602.

Sévenet, N. et al., 1999. Constitutional mutations of the hSNF5/INI1 gene predispose to a variety of cancers. American journal of human genetics, 65(5), pp.1342–8.

Shaulian, E. et al., 1992. Identification of a minimal transforming domain of p53: negative dominance through abrogation of sequence-specific DNA binding. Molecular and cellular biology, 12(12), pp.5581–5592.

Shaw, P. et al., 1992. Induction of apoptosis by wild-type p53 in a human colon tumor-derived cell line. Proceedings of the National Academy of Sciences, 89(10), pp.4495–4499.

Sheltzer, J.M. et al., 2012. Transcriptional consequences of aneuploidy. Proceedings of the National Academy of Sciences , 109 (31 ), pp.12644–12649.

Shin, K.-H. et al., 2003. Hypermethylation of the hTERT promoter inhibits the expression of telomerase activity in normal oral fibroblasts and senescent normal oral keratinocytes. British journal of cancer, 89(8), pp.1473–1478.

Shintaku, M. et al., 2008. Ossifying choroid plexus papilloma recurring with features of atypical papilloma. Neuropathology : official journal of the Japanese Society of Neuropathology, 28(2), pp.160–4.

Shlien, A. et al., 2008. Excessive genomic DNA copy number variation in the Li-Fraumeni cancer predisposition syndrome. Proceedings of the National Academy of Sciences of the United States of America, 105(32), pp.11264–11269.

160 Siegel, J.J. & Amon, A., 2012. New Insights into the Troubles of Aneuploidy. Annual review of cell and developmental biology, (June), pp.1–26.

Sigal, A. & Rotter, V., 2000. Oncogenic mutations of the p53 tumor suppressor: The demons of the guardian of the genome. Cancer Research, 60(24), pp.6788–6793.

Solomon, D. a et al., 2011. Mutational Inactivation of STAG2 Causes Aneuploidy in Human Cancer. Science, 333, pp.1039–1043.

Somyajit, K. et al., 2013. ATM- and ATR-Mediated Phosphorylation of XRCC3 Regulates DNA Double-Strand Break-Induced Checkpoint Activation and Repair. Molecular and Cellular Biology , 33 (9 ), pp.1830–1844.

Stephens, P.J. et al., 2011. Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell, 144(1), pp.27–40.

Stingele, S. et al., 2012. Global analysis of genome, transcriptome and proteome reveals the response to aneuploidy in human cells. Molecular Systems Biology, 8(1).

Storm, S.M. & Rapp, U.R., 1993. Oncogene activation: c-raf-1 gene mutations in experimental and naturally occurring tumors. Toxicology letters, 67(1-3), pp.201–210.

Stratton, M.R., Campbell, P.J. & Futreal, P.A., 2009. The cancer genome. Nature, 458(7239), pp.719–724.

Strazielle, N. & Ghersi-Egea, J.F., 2000. Choroid plexus in the central nervous system: biology and physiopathology. Journal of neuropathology and experimental neurology, 59(7), pp.561–574.

Subramanian, A. et al., 2005. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America, 102(43), pp.15545–50.

Suh, Y.A. et al., 2011. Multiple stress signals activate mutant p53 in vivo. Cancer Research, 71(23), pp.7168–7175.

161 Sun, M.Z. et al., 2014. Gross total resection improves overall survival in children with choroid plexus carcinoma. Journal of Neuro-Oncology, 116, pp.179–185.

Tabori, U. et al., 2010. TP53 alterations determine clinical subgroups and survival of patients with choroid plexus tumors. Journal of Clinical Oncology, 28(12), pp.1995–2001.

Taggard, D.A. & Menezes, A.H., 2000. Three choroid plexus papillomas in a patient with Aicardi syndrome. A case report. Pediatric neurosurgery, 33(4), pp.219–223.

Takahashi, M. et al., 2009. Efficacy of multi-staged surgery and adjuvant chemotherapy for successful treatment of atypical choroid plexus papilloma in an infant: case report. Neurologia medico-chirurgica, 49(10), pp.484–7.

Tanaka, K. et al., 2009. Rapid regrowth of an atypical choroid plexus papilloma located in the cerebellopontine angle. Journal of Clinical Neuroscience, 16, pp.121–124.

Tatevossian, R.G. et al., 2010. MYB upregulation and genetic aberrations in a subset of pediatric low-grade gliomas. Acta Neuropathologica, 120(6), pp.731–743.

Tong, Y. et al., 2015. Cross-Species Genomics Identifies TAF12, NFYC, and RAD54L as Choroid Plexus Carcinoma Oncogenes. Cancer Cell, 27(5), pp.712–727.

Torres, E.M. et al., 2010. Identification of aneuploidy-tolerating mutations. Cell, 143(1), pp.71– 83.

Tuna, M., Knuutila, S. & Mills, G.B., 2009. Uniparental disomy in cancer. Trends in molecular medicine, 15(3), pp.120–8.

Turkoglu, E. et al., 2014. Clinical outcome of adult choroid plexus tumors: retrospective analysis of a single institute. Acta neurochirurgica, 156, pp.1461–1468.

Uchiyama, C.M. et al., 1997. Choroid Plexus Papilloma and Cysts in the Aicardi Syndrome: Case Reports. Pediatric Neurosurgery, 27(2), pp.100–104.

162 Upender, M.B. et al., 2004. Chromosome Transfer Induced Aneuploidy Results in Complex Dysregulation of the Cellular Transcriptome in Immortalized and Cancer Cells. Cancer Research , 64 (19 ), pp.6941–6949.

Uziel, T. et al., 2005. The tumor suppressors Ink4c and p53 collaborate independently with Patched to suppress medulloblastoma formation. Genes and Development, 19(22), pp.2656– 2667.

Valera, E.T. et al., 2009. Multiple dicentric chromosomes behind polyploidy in grade II atypical choroid plexus papilloma: a complementary cytogenetic evaluation. Neuropathology : official journal of the Japanese Society of Neuropathology, 29(2), pp.200–2.

Venkatachalam, S. et al., 1998. Retention of wild‐type p53 in tumors from p53 heterozygous mice: reduction of p53 dosage can promote cancer formation. The EMBO Journal, 17(16), pp.4657–4667.

Voulgari, A. et al., 2008. TATA Box-Binding Protein–Associated Factor 12 Is Important for RAS-Induced Transformation Properties of Colorectal Cancer Cells. Molecular Cancer Research , 6 (6 ), pp.1071–1083.

Vousden, K.H. & Ryan, K.M., 2009. P53 and Metabolism. Nature reviews. Cancer, 9(10), pp.691–700.

Wang, J. et al., 2011. CREST maps somatic structural variation in cancer genomes with base- pair resolution. Nat Meth, 8(8), pp.652–654.

Wang, P.-Y. et al., 2013. Increased Oxidative Metabolism in the Li–Fraumeni Syndrome. New England Journal of Medicine, 368(11), pp.1027–1032.

Wang, Y. et al., 1998. Induced p53 expression in lung cancer cell line promotes cell senescence and differentially modifies the cytotoxicity of anti-cancer drugs. Oncogene, 1299(May), pp.1923–1930.

163 Watanabe, M. et al., 2012. BMP4 sufficiency to induce choroid plexus epithelial fate from embryonic stem cell-derived neuroepithelial progenitors. The Journal of neuroscience : the official journal of the Society for Neuroscience, 32(45), pp.15934–45.

Weaver, B. a a et al., 2007. Aneuploidy Acts Both Oncogenically and as a Tumor Suppressor. Cancer Cell, 11(1), pp.25–36.

Weaver, B. a. & Cleveland, D.W., 2006. Does aneuploidy cause cancer? Current Opinion in Cell Biology, 18(6), pp.658–667.

White, R., Rose, K. & Zon, L., 2013. Zebrafish cancer: the state of the art and the path forward. Nat Rev Cancer, 13(9), pp.624–636.

Wiestler, B. et al., 2013. ATRX loss refines the classification of anaplastic gliomas and identifies a subgroup of IDH mutant astrocytic tumors with better prognosis. Acta Neuropathologica, 126(3), pp.443–451.

Witt, H. et al., 2011. Delineation of Two Clinically and Molecularly Distinct Subgroups of Posterior Fossa Ependymoma. Cancer cell, 20(2), pp.143–157.

Wolburg, H. & Paulus, W., 2009. Choroid plexus: biology and pathology. Acta neuropathologica, 119(1), pp.75–88.

Wolf, D. & Rotter, V., 1985. Major deletions in the gene encoding the p53 tumor antigen cause lack of p53 expression in HL-60 cells. Proceedings of the National Academy of Sciences, 82, pp.790–794.

Wolff, J.E.A. et al., 2002. Choroid plexus tumours. British Journal of Cancer, 87(10), pp.1086– 1091.

Wrede, B. et al., 2009. Atypical choroid plexus papilloma: clinical experience in the CPT-SIOP- 2000 study. Journal of neuro-oncology, 95(3), pp.383–392.

164 Wright, W.D. & Heyer, W.-D., 2014. Rad54 Functions as a Heteroduplex DNA Pump Modulated by Its DNA Substrates and Rad51 during D Loop Formation. Molecular Cell, 53(3), pp.420–432.

Wu, G. et al., 2012. Somatic histone H3 alterations in pediatric diffuse intrinsic pontine gliomas and non-brainstem glioblastomas. Nature Genetics, 44(3), pp.251–253.

Xue, W. et al., 2012. A cluster of cooperating tumor-suppressor gene candidates in chromosomal deletions. Proceedings of the National Academy of Sciences , 109 (21 ), pp.8212–8217.

Yuasa, H., Tokito, S. & Tokumaga, M., 1992. Primary Carcinoma of the choroid plexus in Li- Fraumeni syndrome. Neurosurgery, 32(1), pp.131–134.

Zakrzewska, M. et al., 2005. Mutational analysis of hSNF5/INI1 and TP53 genes in choroid plexus carcinomas. Cancer Genetics and Cytogenetics, 156(2), pp.179–182.

Zebisch, A. et al., 2009. Loss of RAF kinase inhibitor protein is a somatic event in the pathogenesis of therapy-related acute myeloid leukemias with C-RAF germline mutations. Leukemia : official journal of the Leukemia Society of America, Leukemia Research Fund, U.K, 23(6), pp.1049–1053.

Zebisch, A. et al., 2006. Two transforming C-RAF germ-line mutations identified in patients with therapy-related acute myeloid leukemia. Cancer Research, 66(7), pp.3401–3408.

Zender, L. et al., 2015. An Oncogenomics-Based In Vivo RNAi Screen Identifies Tumor Suppressors in Liver Cancer. Cell, 135(5), pp.852–864.

Zhang, G., Yang, J. & Cui, Z., 2015. New Insight into Cancer Aneuploidy in Zebrafish. In International Review of Cell and Molecular Biology. pp. 149–170.

Zhang, J., Benavente, C. a., et al., 2012. A novel retinoblastoma therapy from genomic and epigenetic analyses. Nature, 481(7381), pp.329–334.

Zhang, J., Ding, L., et al., 2012. The genetic basis of early T-cell precursor acute lymphoblastic leukaemia. Nature, 481(7380), pp.157–163.

165 Zhang, J. et al., 2013. Whole-genome sequencing identifies genetic alterations in pediatric low- grade gliomas. Nature genetics, 45(6), pp.602–12.

Zhang, Y. et al., 2012. E2F1: A potential negative regulator of hTERT transcription in normal cells upon activation of oncogenic c-Myc. Medical Science Monitor : International Medical Journal of Experimental and Clinical Research, 18(1), pp.RA12–RA15.

Zheng, W. & Zhao, Q., 2002. Establishment and characterization of an immortalized Z310 choroidal epithelial cell line from murine choroid plexus. Brain Research, 958, pp.371–380.