Post on 15-Sep-2020
Integrated Genomic Analyses of Childhood Central Nervous System-Primitive Neuro-ectodermal Tumours (CNS-PNETs)
By
Daniel J. Picard
A thesis submitted in conformity with the requirements
for the degree of Master of Science
Graduate Department of Laboratory Medicine and Pathobiology University of Toronto
2014
© Copyright by Daniel J. Picard
ii
Integrated Genomic Analyses of Childhood Central Nervous System-Primitive Neuro-ectodermal Tumours (CNS-PNETs)
Daniel J. Picard
Master of Science
Graduate Department of Laboratory Medicine and Pathobiology University of Toronto
2014
Abstract
CNS-PNETs are rare, aggressive, paediatric embryonal brain tumours that are poorly
studied. We recently identified an aggressive sub-group of CNS-PNETs characterized by the
amplification of the C19MC microRNA cluster, however, little is known regarding the features
of other CNS-PNET tumours. This study was designed to define additional molecular sub-groups
of CNS-PNET by interrogating a large cohort of CNS-PNETs.
To elucidate the features of CNS-PNET, we examined transcriptional and copy number
profiles from primary hemispheric CNS-PNETs. We then validated and examined the clinical
significance of novel sub-group markers in 123 primary CNS-PNETs.
This thesis demonstrates that CNS-PNET can be categorized into three molecular sub-
groups that are distinguished by distinct primitive neural, oligo-neural and mesenchymal lineage
gene expression signatures and also correlated with distinct clinical features.
Data from my thesis has generated a substantial body of knowledge to fuel both
biological and clinical investigations of childhood CNS-PNETs.
iii
Acknowledgements
I am thankful to my supervisor, Dr. Annie Huang, and my graduate committee members, Dr.
Cynthia Hawkins and Dr. Herman Yeger, whose guidance and support from the initial to the
final stages of the project have provided a productive research environment and helped me
develop an understanding of the subject. Specifically, I would like to thank Dr. Huang for the
amazing opportunity to be a part of this study.
I wish to express my gratitude to past and present members of the Huang Lab, specifically David
Shih, Tiffany Chan, Patrick Sin-Chan and Jonathon Torchia, for their advice and support during
the course of my research. I am also grateful for helpful discussions with and advice from
members of other labs. I would also like to thank: Suzanne Miller and her supervisor Richard
Grundy for their help and contribution to the study; Pawel Buczkowicz, from Dr. Hawkin’s lab,
for his help with the Partek and SPSS packages and discussions regarding his experience with
bioinformatics; Anath Lionel, from Dr. Steve Scherer’s lab, for help with the copy number
analysis; and Sameer Agnihotri, from Abhijit Guha’s lab, for helpful and insightful scientific
discussions.
The progress of my project was also facilitated by the collaboration with and the generous
sharing of equipment and reagents from other labs; I am grateful to: Dr. Peter Dirks, Dr.
Meredith Irwin, Dr. Michael Taylor, and Dr. Jane McGlade.
I would also like to acknowledge members of the Pathology Department at Sickkids and the
Pathology Research Program at the University Health Network for their wonderful immuno-
histochemistry work and for being so accommodating.
This work was supported in part by a University of Toronto Fellowship.
iv
Table of Contents
Abstract ........................................................................................................................................... ii
Acknowledgments.......................................................................................................................... iii
List of Tables and Figures............................................................................................................. vii
List of Appendices ....................................................................................................................... viii
Abbreviations ................................................................................................................................. ix
Chapter 1 Introduction .................................................................................................................. 1
1.1. Paediatric brain tumours .......................................................................................................... 2
1.1.1. Classification of childhood brain tumours ..................................................................... 2
1.1.2. Clinical challenges in the treatment of childhood brain tumours .................................. 3
1.2. Childhood CNS-PNETs ........................................................................................................... 4
1.2.1. Clinical characteristics of CNS-PNET .......................................................................... 4
1.2.2. CNS-PNET models........................................................................................................ 5
1.3. Molecular and genetic characteristics of CNS-PNET ............................................................. 5
1.3.1. Review of published data .............................................................................................. 5
1.3.2. Limitations of current knowledge.................................................................................. 8
1.4. Objectives of Thesis ................................................................................................................. 9
1.4.1. Hypothesis ..................................................................................................................... 9
1.4.2. Objectives ...................................................................................................................... 9
Chapter 2 Methods and Materials ................................................................................................ 10
2.1 Clinical cohort and tumour materials ...................................................................................... 11
v
2.2 Gene expression and DNA copy number profiles .................................................................. 11
2.3 PCR, Fluorescence in-situ hybridization and immuno-histochemical validation studies ....... 12
2.4 Informatics and Statistical Analyses ....................................................................................... 13
Chapter 3 CNS-PNETs comprise 3 molecular sub-groups with distinct gene expression and copy
number features ............................................................................................................................. 15
3.0 Overview ................................................................................................................................. 16
Objective 1 ............................................................................................................................. 16
Hypothesis ............................................................................................................................. 16
3.1 Central review of a multi-centre CNS-PNET cohort .............................................................. 16
3.2 CNS-PNETs arising in the cerebral hemispheres comprise 3 molecular sub-groups ............. 20
3.3 Cell lineage markers, LIN28 and OLIG2, distinguish molecular sub-groups of CNS-PNETs
....................................................................................................................................................... 28
3.4 Molecular sub-groups of CNS-PNET exhibit distinct DNA copy number patterns ............... 29
3.5 Discussion ............................................................................................................................... 36
Chapter 4 LIN28 and OLIG2 correlate with survival and metastatic phenotypes in CNS-PNETs .
....................................................................................................................................................... 38
4.0 Overview ................................................................................................................................. 39
Objective 2 ............................................................................................................................. 39
Hypothesis ............................................................................................................................. 39
4.1 Defining clinical phenotypes of CNS-PNET molecular sub-groups ...................................... 39
4.1.1. Gender ......................................................................................................................... 42
4.1.2. Age............................................................................................................................... 42
vi
4.1.3. Metastasis .................................................................................................................... 45
4.2 Survival features of CNS-PNET molecular sub-groups ......................................................... 45
4.3 Discussion ............................................................................................................................... 46
Chapter 5 Discussion and future directions ................................................................................. 50
5.0 Discussion ............................................................................................................................... 51
5.1 Clinical relevance of findings ................................................................................................. 52
5.2 Future directions ..................................................................................................................... 53
References ..................................................................................................................................... 55
Appendices .................................................................................................................................... 60
vii
List of Tables and Figures
Tables
Table 1.1: Molecular genetics of supratentorial PNETs and pineoblastomas ................................ 6
Table 1.2: Molecular and cytogenetic abnormalities of CNS embryonal tumours ......................... 7
Table 2.1: Quantitative qRT-PCR primers and probes ................................................................. 13
Table 4.1: Clinical and molecular features of CNS-PNET ........................................................... 40
Table 4.2: Clinical and molecular features of CNS-PNET in relation to age ............................... 41
Figures
Figure 3.1 Flow chart of sample analysis ..................................................................................... 34
Figure 3.2 Gene expression profiles define 3 molecular sub-groups of CNS-PNET ................... 18
Figure 3.3 Molecular sub-groups of CNS-PNET exhibit distinct cell lineage ............................. 21
Figure 3.4 Molecular sub-groups of CNS-PNET exhibit distinct signalling signatures ............... 24
Figure 3.5 Validation of sub-group specific gene expression signatures ..................................... 26
Figure 3.6 qRT-PCR and immuno-histochemical analyses of IGF2 ............................................ 30
Figure 3.7 Cell lineage markers, LIN28 and OLIG2, correlate with molecular sub-groups of CNS-PNET ................................................................................................................................... 32
Figure 3.8 Molecular sub-groups of CNS-PNET have distinct DNA copy number alterations ... 34
Figure 4.1 Molecular sub-groups of CNS-PNET exhibit distinct clinical phenotypes ................. 43
Figure 4.2 Molecular sub-groups of CNS-PNET exhibit distinct survival features ..................... 47
viii
List of Appendices
Appendix 1 .................................................................................................................................... 61
Table A-1.1 Clinicopathologic characteristics of tumour samples ............................................... 62
Table A-1.2 Analyses performed on tumour samples .................................................................. 65
Appendix 2 .................................................................................................................................... 69
Manuscript: Markers of survival and metastatic potential in childhood CNS primitive neuro-ectodermal brain tumours: an integrative genomic analysis ......................................................... 70
ix
Abbreviations ATRT Atypical Teratoid Rhabdoid Tumour BAF47 Synonym of SWI/SNF related, matrix associated, actin dependent regulator of
chromatin, subfamily b, member 1 (SMARCB1) CDKN2A Cyclin-dependent Kinase Inhibitor 2A CDKN2B Cyclin-dependent Kinase Inhibitor 2B CNA Copy Number Aberration CNS Central Nervous System EGFR Epidermal Growth Factor Receptor ETANTR Embryonal Tumour with Abundant Neuropil and True Rosettes FDR False Discovery Rate FFPE Formalin Fixed Paraffin Embedded GBM GlioblastomaMultiforme GFAP Glial Fibrillary Acidic Protein H3.3 Synonym for H3 Histone, Family 3A (H3F3A) HCL Hierarchal Clustering IHC Immunohistochemistry INK4D Synonym of cyclin-dependent kinase inhibitor 2D (CDKN2D, p19) IPA Ingenuity Pathway Analysis MLPA Multiplex Ligation-Dependent Probe Amplification MYC v-myc Avian Myelocytomatosis Viral Oncogene Homolog MYCN v-myc Avian Myelocytomatosis Viral Oncogene Neuroblastoma Derived
Homolog NMF Non-negative Matrix Factorization PARP1 poly (ADP-ribose) polymerase 1 PCA Principal Component Analysis PDGFB Platelet-Derived Growth Factor Beta Polypeptide PDGFRA Platelet-Derived Growth Factor Receptor, Alpha Polypeptide PNET Primitive Neuro-Ectodermal Tumour PNET-NOS Primitive Neuro-Ectodermal Tumour-Not Otherwise Specified RB Retinoblastoma 1 RCAS Replication-Competent ASLV long terminal repeat (LTR) with a Splice acceptor RHOG Ras Homolog Family Member G SMARCB1 SWI/SNF related, matrix associated, actin dependent regulator of chromatin,
subfamily b, member 1 SNP Single Nucleotide Polymorphism SV40 Simian Vacuolating Virus 40 TMA Tissue Micro Array TP53 Tumour Protein p53 tv-a CD320 molecule, avian viral receptor WNT16 Wingless-Type MMTV Integration Site Family, Member 16
1
Chapter 1
Introduction
2
1.1.Paediatricbraintumours
1.1.1.Classificationofchildhoodbraintumours
Brain tumours are the most common childhood solid tumour1 and a leading cause of
childhood cancer related morbidity and mortality2. Studies have shown that paediatric and adult
brain tumours are molecularly distinct as exemplified by glioblastoma multiforme, where
amplification of EGFR is detected in >40% of adult glioblastoma multiforme but is rarely
observed in paediatric glioblastoma multiforme3. Of note, many adult cancers are the result of
carcinogenic exposures, whereas aberrations in developmental signalling pathways have been
demonstrated to be an important factor in paediatric brain tumour pathogenesis4. Compared to
adult brain tumours, paediatric brain tumours are rare and until recently, there have been few
studies characterizing the molecular landscape and pathogenesis of paediatric brain tumours.
Historically, embryonal tumours, which comprise a large fraction of paediatric brain
tumours, were grouped under the umbrella term primitive neuroectodermal tumour (PNET)5;
regardless of location in the central nervous system (CNS). There has been controversy over
classification, treatment, and cell of origin of these tumours since their defining feature is a
relatively homogeneous histological appearance consisting of poorly cohesive, undifferentiated
neuroepithelial cells, often with a high mitotic rate6. Due to these common features, embryonal
tumours were thought to arise from a common precursor cell of the subependymal matrix in the
CNS. The tendency for these neoplasms to infiltrate nearby normal tissues and disseminate
through cerebrospinal fluid (CSF) pathways was believed to contribute to their poor outcome.
However, evidence suggests that rather than being comprised of a large uniform group of
tumours, PNETs are a heterogeneous group of WHO grade IV neoplasms which include
medulloblastoma, atypical teratoid/rhabdoid tumour (ATRT), pineoblastoma, and CNS-PNET6.
Evidence of substantial tumour biological diversity across embryonal/PNET tumours is reflected
in patient survival; despite histologic similarities CNS-PNETs have distinctly poorer 5 year
overall survival of 20-50%7,8 as compared to medulloblastoma in which 5 year overall survival
for localized disease of 75-85% is achievable9. In fact, recent global gene expression profiling of
embryonal/PNET neoplasm favour the older concept of a cell type specific origin for distinct
embryonal tumours 10,11, which may be reflected by a combination of tumour location and
histopathologic patterns of differentiation.
3
1.1.2.Clinicalchallengesinthetreatmentofchildhoodbraintumours
In contrast to intrinsic brain tumours of adults, paediatric brain tumours span a much
wider histologic and biologic spectrum of diseases of which embryonal brain tumours/PNETs
comprise a major category. All embryonal/PNET brain tumours, which include medulloblastoma
and CNS-PNET, currently receive similar types of intensive post-surgical, chemo-
radiotherapeutic treatment regimens. Whole brain and spine irradiation is often applied, as
embryonal/PNET brain tumours, unlike malignant glial tumours, have a propensity to
metastasize within the cranio-spinal axes.
Substantial progress has been made with recent global genomic studies regarding
molecular pathogenesis of medulloblastoma. Amongst the embryonal tumours, medulloblastoma
patients have the best clinical outcomes12-19. However, despite histologic resemblance to
medulloblastoma, patients with CNS-PNETs fare poorly even with intensified high-dose
radiation and chemotherapy regimens designed for patients with metastatic medulloblastoma8,20-
23. Due to the rare incidence of CNS-PNETs, which represent 1/10th of medulloblastoma, effects
of specific intervention on CNS-PNET outcomes has been difficult to evaluate, as most clinical
data has been often been reported together with medulloblastoma. Furthermore, clinical studies
of CNS-PNETs have often been based on anatomical categorization to include all tumours
arising above the tentorium, including hemispheric CNS-PNETs and pineoblastomas as one
therapeutic entity. However, pineoblastomas exhibit distinct histo-pathologic features from CNS-
PNETs arising in the cerebral hemispheres24,25 and appear to respond differently to therapy as
compared to hemispheric CNS-PNETs6, and thus should probably be considered separately in
therapeutic trials.
The molecular and cellular make-up of CNS-PNET remain largely unknown24, hence
current tumour treatments are empiric and largely ineffective26. Tragically, even with the advent
of new intensive treatments, the prognosis for these patients diagnosed with CNS-PNETs is
dismal and often leads to debilitating and severe late effects, such as decreased intelligence in the
small proportion of survivors. In order to advance CNS-PNET therapeutics, it is important to
delineate the cellular and molecular pathogenesis of CNS-PNET to better inform diagnosis,
prognosis, and tumour specific treatment design.
4
1.2.ChildhoodCNS‐PNETs
1.2.1.ClinicalcharacteristicsofCNS‐PNET
CNS-PNETs are rare, predominantly hemispheric tumours comprising ~3-5% of all
paediatric brain tumours. Age at presentation is from birth to 20 years for paediatric CNS-
PNETs, however rare adult cases have been reported27,28. To date, 65% of CNS-PNETs are
reported in patients less than 5 years of age and no gender bias has been observed29. Patients
present with a variety of symptoms which are related to tumour location; however, increased
intracranial pressure is the most common symptom. CNS-PNETs can be visualized by computed
tomography (CT) and magnetic resonance imaging (MRI), but look similar regardless of location
or to other hemispheric lesions6. Although the 5 year disease-free survival is approximately 37%,
many patients suffer recurrences21.
CNS-PNETs comprise a diverse histologic group which exhibit variable neuronal,
astrocytic, or ependymal differentiation and can be challenging to diagnose by routine histo-
pathology30. Presently, CNS–PNETs are sub-categorized based on location and specific
histological features. The most common sub-category is CNS-PNET not otherwise specified
(NOS), previously known as supratentorial-PNET (sPNET), a histologically poorly differentiated
tumour arising in the cerebrum. Rarer histologic sub-categories include cerebral neuroblastomas,
ganglioneuroblastoma, medulloepitheliomas, ependymoblastomas and ‘embryonal tumour with
abundant neuropil and true rosettes’ (ETANTR). Medulloepitheliomas are tumours characterized
by the presence of neural tube formation, whereas ependymoblastic rosettes are the defining
feature of ependymoblastomas31. ETANTRs are a recently described CNS-PNETs sub-group32,
with histological features similar to neuroblastoma and ependymoplastoma, characterized by true
and pseudo-rosettes on a background of abundant neuropil. However, CNS-PNET sub-
classification is under debate33 as emerging data suggest many of the histologic sub-classes may
represent closely related biological and molecular entities.
5
1.2.2.CNS‐PNETmodels
Studies of CNS-PNET in mouse models are limited since the cell of origin is unknown;
there are also few xenograft models of CNS-PNETs34. Specific genetic models of CNS-PNETs
remain to be described. However, hemispheric tumours resembling CNS-PNETs have been
described in mouse models of medulloblastoma with defects in p53, pRB and the CDKN2A
tumour suppressor pathways. Transgenic mice universally expressing the SV40 large T-antigen,
which inhibits both p53 and RB, develop CNS-PNETs with an incidence of 27%35. In addition,
mice with null mutations in PARP1 and p53, as well as INK4D and p53, develop CNS-PNET
with an incidence of 2% and 10%, respectively36,37. In 2008, Momota et al. applied the RCAS
somatic transgenic system to model CNS-PNET tumour formation38. These investigators
generated a ~30% incidence of CNS-PNET-like tumours in a p53 knockout mouse with over-
expression of c-Myc or c-Myc with -catenin using an RCAS somatic transgenic system.
Collectively, these murine studies indicate that multiple abnormalities are likely to be required
for tumour initiation and progression of CNS-PNETs.
1.3.MolecularandgeneticcharacteristicsofCNS‐PNET
1.3.1.Reviewofpublisheddata
Due to the absence of specific animal models and limited studies of primary tumours,
genetic features of CNS-PNET remain largely unknown. Studies of small number of CNS-
PNETs (total of 33 CNS-PNETs reported prior to 2009) indicate diverse copy number
aberrations (CNAs) in most CNS-PNET but only recurrent CNAs in <10-20% of tumours.
Importantly, these data indicate that CNS-PNETs have distinct CNAs from medulloblastoma
(Tables 1.1, 1.2)24,38. Specifically, they have rare iso-chromosome 17q (i17q), which is present
in 30- 40% of medulloblastoma24,39-41, and more frequent CNAs of chromosome 1p12-22.1, 9p41,
13q40, 14q42, 19p39,41 and 19q42. To date, MYCN, PDGFB, and PDGFRA amplification and
deletions of CDKN2A/2B39-41 have been detected in CNS-PNET. However, the identity of genes
underlying most CNAs, the relevance of observed CNAs and genetic alterations to the biology
and clinical attributes of CNS-PNETs remain largely unexplored.
6
Table1.1:MoleculargeneticsofsupratentorialPNETsandpineoblastomas
Reproduced from Li et al 24 with copyright permission from Neurosurgical Focus
7
Table1.2:MolecularandcytogeneticabnormalitiesofCNSembryonaltumours
Reproduced from Inda et al 39 with copyright permission from Histopathology
8
In 2009, Li et al 43 reported the first large cohort study of 40 CNS-PNET using global gene
expression and copy number profiling. They discovered amplification of a novel oncogenic
miRNA cluster, C19MC, in 25% of hemispheric CNS-PNETs by copy number analyses and
demonstrated that the C19MC marker identified a sub-group of CNS-PNETs with variant
histology that shared the same transcriptional signatures. Significantly, the C19MC amplicon
was not restricted to a specific histologic sub-class of CNS-PNETs, but was detected in
ETANTR, medulloepithelioma and ependymoblastoma samples as well as a proportion of CNS-
PNET-NOS. Korshunov et al44 confirmed the presence of C19MC amplicon in CNS-PNETs
variants with histologic features of rosette formation. Both studies showed that the C19MC
tumours had poor survival compared to other CNS-PNETs. To date, no specific histologic
features which correlate with non-C19MC amplified tumours have been identified. Interestingly,
a recent study suggested that a specific mutation of the H3.3 gene (H3.3G34R), which is seen in
about 25% of glioblastoma, is also found in subset of CNS-PNETs45. However, the H3.3
mutation K27M46 which is seen at high frequency in glioblastoma and diffuse intrinsic pontine
glioma, is not present in CNS-PNETs45. These findings highlight longstanding challenges with
histologic diagnosis of CNS-PNET and suggest that molecular features of some tumours
diagnosed as CNS-PNETs and glioblastoma may overlap.
1.3.2.Limitationsofcurrentknowledge
Although the study by Li et al43 led to the identification of one distinct sub-group of CNS-
PNETs, little is known regarding the molecular features of the other CNS-PNET tumours which
lack the C19MC marker. In order to define the full molecular spectrum of CNS-PNETs,
concerted molecular analyses of a substantial numbers of primary CNS-PNETs, which can only
be enabled by multi-institutional, international collaboration, is necessary. In my thesis, I
undertook a study designed to integrate gene expression, copy number, and immuno-
histochemical analyses in order to characterize 254 primary hemispheric CNS-PNETs collected
through a multi-institutional collaborative network. Integrated studies of such a large cohort
should enable the identification of molecular markers for additional CNS-PNET sub-groups and
help to define additional molecular correlates of clinical phenotypes in CNS-PNETs which will
be instrumental in informing prospective therapeutic studies.
9
1.4.ObjectivesofThesis
1.4.1.Hypothesis
CNS-PNETs comprise additional molecular sub-groups which may be defined by
integrated genomic analyses of a substantial CNS-PNET cohort.
1.4.2.Objectives
1. To comprehensively define molecular features of a substantial CNS-PNET cohort
using integrated global gene expression and copy number profile analyses.
2. To correlate molecular with clinico-pathologic features of CNS-PNETs in order to
define clinically relevant sub-groups of CNS-PNETs.
10
Chapter 2
Methods and Materials
These materials and methods were described in part in Picard D, Miller S, Hawkins CE, Bouffet E, Rogers HA, Chan TS, Kim SK, Ra YS, Fangusaro J, Korshunov A et al: Markers of survival and metastatic potential in childhood CNS primitive neuro-ectodermal brain tumours: an integrative genomic analysis. Lancet Oncol 2012, 13(8):838-848 and have been reproduced with copyright permission from Lancet Oncology.
11
2.1Clinicalcohortandtumourmaterials
Tumour materials and clinical information used in this study were collected through an
international collaborative network. Tumour samples and patient clinical information were
collected with consent as per protocols approved by the Hospital Research Ethics Board at
participating institutions including Children’s Cancer and Leukemia Group (CCLG) registered
centers in the UK, and from the Cooperative Human Tissue Network (CHTN, Columbus, Ohio
USA). CNS-PNET tissue microarrays used in this study were constructed at the Hospital for Sick
Children, University of Nottingham25, and Institute of Cancer Research in Sutton, UK. All
collected tumour samples were re-reviewed blindly by Dr. Cynthia Hawkins, and tested for loss
of SMARCB1/BAF47/INI1/SNF5 protein expression or SMARCB1 alterations by sequencing or
MLPA analyses to rule out misdiagnosed ATRT. Only hemispheric tumours diagnosed as CNS-
PNET according to the 2007 WHO CNS tumour classification criteria6 and without alterations of
SMARCB1 were included. Patient and tumour information are listed in Appendix 1 Table A-1.1.
2.2GeneexpressionandDNAcopynumberprofiles
DNA and RNA were extracted using standard methods from 77 and 51 primary CNS-
PNET samples respectively and analysed using Illumina Omni 2.5M SNP and Illumina HT-12
v4 gene expression arrays (http://www.illumina.com) to generate gene expression and DNA
copy number profiles. DNA and RNA hybridizations were performed at The Centre of Applied
Genomics (TCAG), Hospital for Sick Children, Toronto (http://www.tcag.ca/), according to the
manufacturer’s protocol. For clinical correlative analyses, only 59 of the 77 tumours with copy
number profiles that had complete clinical information were included. Details of molecular
analyses performed on individual tumour samples are shown in Appendix 1, Table A-1.2; all
data are deposited in the Wellcome Trust, European Genome-phenome Archive
(www.EBI.AC.UK; Accession number:EGAS00000000116).
12
2.3PCR,Fluorescencein‐situhybridizationandimmuno‐histochemicalvalidationstudies
For gene-specific qRT-PCR validation of array data, 10 ng cDNA synthesized from 1µg
of RNA (TaqMan® Reverse Transcription Kit, Applied Biosystems) was amplified using
specific TaqMan probes/primer sets (Table 2.1) and mRNA expression levels were determined
relative to actin using the ΔCt method. All qRT-PCR assays were performed in triplicate.
Immuno-histochemical analyses on tumour tissue microarray or FFPE tumour slides were
performed by the Pathology Research Program core laboratory (http://www.uhnres.utoronto.ca).
All tissue sections were treated with heat-induced epitope retrieval and blocked for endogenous
peroxidase and biotin. Antibodies used in this study included: anti-LIN28 (Cell Signalling
Technology, Boston, USA), anti-OLIG2, (Immuno-Biological Laboratories, Minneapolis, USA),
GFAP (DAKO, Burlington, CA), and anti-SYNAPTOPHYSIN (Millipore, Massachusetts,
USA). Antibody reactions were visualized using a Biogenix detection kit (BioGenix
Laboratories, San Ramon, USA). Immuno-reactivity for LIN28, GFAP and SYNAPTOPHYSIN
were scored manually based on intensity (1=low, 2=mod, 3=high) and distribution of stains
(1=≤10%, 2=10-50%, 3>50%). OLIG2 immuno-stains were quantified using the Aperio
Scanscope (Aperio, Vista CA, USA) system and the ImageScope software nuclear IHC
algorithm. For tumours on TMA, IHC values were determined based on average staining score of
at least 2 tissue cores, while tumours with FFPE slides were scored based on the extent of
staining in relation to the entire tumour section. Normal testicular tissue (human and murine) and
oligodendroglioma tumour tissue were used respectively as positive controls for LIN28 and
OLIG2 immuno-stains; samples processed in parallel without primary antibodies served as
negative controls. All IHC stains were scored blindly by myself and Tiffany Chan, and reviewed
by Dr. Annie Huang and Dr. Cynthia Hawkins. FISH was performed on FFPE TMA or
individual slides using established protocols. Test MYCN (2p24) and p16 (9p21) specific
PlatinumBright550 probe with corresponding LAF (2q11) and 9q21 PlatinumBright495 control
probes (Kreatech, Stretton Scientific, Stretton, UK) were used.
13
Table2.1:QuantitativeqRT‐PCRprimersandprobes
Gene Probe # ASCL1 Hs00269932_m1 COL1A1 Hs00164004_m1 COL1A2 Hs00164099_m1 COL5A1 Hs00609088_m1 CRABP1 Hs00171635_m1 ERBB3 Hs00176538_m1 FOXJ1 Hs00230964_m1 GFAP Hs00909233_m1 GLI2 Hs01119974_m1 GLI3 Hs00609233_m1 IGF2 Hs00171254_m1 LIN28A Hs00702808_s1 LIN28B Hs01013729_m1 MSX1 Hs00427183_m1 NCAM2 Hs00189850_m1 NES Hs00707120_s1 NEUROG2 Hs00702774_s1 NKX6-2 Hs00752986_s1 OLIG1 Hs00744293_s1 OLIG2 Hs00377820_m1 PDGFRA Hs00998018_m1 SMO Hs01090242_m1 TGFB3 Hs01086000_m1 TGFBR3 Hs01114253_m1 TUBB3 Hs00801390_s1 ZIC2 Hs00600845_m1
14
2.4InformaticsandStatisticalAnalyses
CNS-PNETs were classified into molecular sub-groups by unsupervised hierarchical
clustering (HCL), Non-negative Matrix Factorization (NMF)47 and principle component (PCA)
analyses of genes with the highest co-efficient of variation using the Partek Genomics Suite
(Partek, St Louis, MO). Genes enriched within tumour sub-groups were determined using a
supervised T-test adjusted for multiple hypotheses testing using the FDR method. Ingenuity
Pathway Analyses were performed on supervised gene sets to identify canonical signalling
pathways in each tumour sub-group. To determine regions of copy number gains and losses,
inferred copy number data was generated using the Illumina Genome Studio software and was
imported into Partek for CNA partitioning/segmentation analyses using a SNP window of 150.
Significance of CNAs in tumour sub-groups was then determined using Fisher’s exact test. Log
rank analysis using the Kaplan Meier method and Chi-square analyses were used respectively to
compare survival time and proportion of survivors across tumour sub-groups, while ANOVA
was used to determine significance of tumour sub-group in relation to age. To analyse the
significance of molecular sub-groups in relation to gender and metastatic status at diagnosis,
features in an individual molecular sub-group were compared to a pooled cohort of the other two
molecular sub-groups using Fisher’s exact test. Adjustment for multiple testing was not
performed as number of patients with complete information available for each clinical parameter
varied.
15
Chapter 3
CNS-PNETs comprise 3 molecular sub-groups with distinct gene expression and copy number features
Data presented in this chapter has been published in Picard D, Miller S, Hawkins CE, Bouffet E, Rogers HA, Chan TS, Kim SK, Ra YS, Fangusaro J, Korshunov A et al: Markers of survival and metastatic potential in childhood CNS primitive neuro-ectodermal brain tumours: an integrative genomic analysis. Lancet Oncol 2012, 13(8):838-848 and have been reproduced with copyright permission by Lancet Oncology.
Acknowledgements: Statistical analyses in this study were performed in consultation with Dr. Pingzhao Hu. Immunohistochemical analyses were performed by Daniel Picard and received from Drs Cynthia E Hawkins and Gino Somers. Fluorescence in-situ hybridization analyses were performed by Dr Suzanne Miller and Hazel A. Rogers.
16
3.0Overview
Biological studies of CNS-PNETs have to date been limited to low resolution genomic
studies on very small cohorts. In this chapter, I applied an integrated high resolution genomic
approach to analyse one of the largest CNS-PNET collection assembled to date. As diagnostic
criteria for CNS-PNETs remains in flux, I collaborated with Dr. Cynthia Hawkins to perform
detailed histologic and focused molecular characterization of all samples assembled through our
international study network to eliminate other diagnoses including rhabdoid brain tumours which
may share significant overlap in morphologic and clinical characteristics with paediatric CNS-
PNETs. In section 3.2 and 3.4, I describe high resolution gene expression analyses, which
identify molecular sub-groups of CNS-PNETs, and the association of characteristic CNAs with
distinct molecular sub-groups of CNS-PNETs.
Objective1
To comprehensively define molecular features of CNS-PNETs using global gene
expression and copy number profile analyses.
Hypothesis
CNS-PNETs comprise a genetically heterogeneous group of tumours that may be
classified into distinct molecular sub-groups.
3.1Centralreviewofamulti‐centreCNS‐PNETcohort
We received 254 malignant brain tumours with an institutional histopathologic diagnosis of
CNS-PNET from 20 international medical centres. In order to confirm that tumour samples
received exhibited histologic and molecular characteristic compatible with CNS-PNETs, we
performed a histologic review of all tumour samples using the WHO CNS tumour classification
for CNS-PNETs (Figure 3.1). In addition, immuno-staining was performed on all samples to
assess for SMARCB1 protein expression, an immuno-marker which is characteristic and
diagnostic for CNS rhabdoid tumours6. Using such an approach, only 56% of tumours (142/254)
in our cohort met the diagnostic criteria for primary CNS-PNETs thus underscoring the
significant diagnostic challenge posed by CNS-PNETs. In total we identified 108 samples with
17
Figure3.1Flowchartofsampleanalyses 254 Samples received with an institutional diagnosis of CNS-PNET were centrally reviewed.
112 samples (44%) which did not meet the WHO CNS tumour classification for primary CNS-
PNET diagnostic criteria were excluded. Material available for each case varied: 77 and 51,
respectively, samples had biomaterial avalilable for copy number and gene expression analysis,
and 95 FFPE tissue cases were available for immunohistochemical analysis.
18
Figure3.2Geneexpressionprofilesdefines3molecularsub‐groupsofCNS‐PNET
19
Figure 3.2 Gene expression profiles defines 3 molecular sub-groups of CNS-PNETMultiple unsupervised analyses were performed on human HT-12v4 expression array (Illumina)
data from 51 primary CNS-PNET samples. Cluster patterns were derived re-iteratively using an
initial set of 1000 genes identified by the highest co-efficient of variation to determine the most
stable tumour group clusters achievable with a minimal gene set. Three molecular sub-groups of
CNS-PNETs were independently indicated by A. Unsupervised Hierarchical Cluster (HCL)
analyses, B. Non-negative Matrix Factorization (NMF) and C. Principal Component Analysis
(PCA). Individual tumours corresponding to sub-groups 1, 2 and 3 are respectively indicated by
green, blue and purple coloured boxes or spheres.
20
snap frozen or paraffin embedded tumour material suitable for further molecular or immuno-
based studies which are described in sections 3.2, 3.3 and 4.1
3.2CNS‐PNETsarisinginthecerebralhemispherescomprise3molecularsub‐groups
In prior studies, our lab identified a sub-group of CNS-PNETs arising in the cerebral
hemispheres with highly primitive gene expression signature that were distinguished from other
CNS-PNETs by recurrent amplification of the C19MC miRNA cluster43. However, the molecular
spectrum and composition of the other sub-groups of CNS-PNETs could not be fully elucidated
due to limited tumour cohort size. In order to better define the molecular spectrum of CNS-
PNETs, I performed global gene expression analyses on an expanded cohort of 51 primary CNS-
PNET originating in the cerebral hemispheres. Gene expression data sets were analysed using
multiple unsupervised cluster algorithms including HCL (Partek), NMF (GenePattern) and PCA
(Partek) to define molecular sub-groups of CNS-PNETs. Clustering of gene expression data
performed iteratively using 200-1000 genes by HCL analyses revealed CNS-PNET comprises
three distinct molecular sub-groups. Consistent with HCL analyses, NMF analysis using the
same re-iterative approach revealed the highest co-phenetic co-efficient (0.9711) with a 200
genes data set that correlated with an optimal k value of 3 which also indicated the strongest
statistical support for 3 molecular classes of CNS-PNETs (Figure 3.2 A-B). Analyses of gene
expression data with PCA, a three dimensional orthogonal approach to clustering, showed sub-
group 1 tumours clearly segregated as a distinct molecular entity, while sub-group 2 and 3
tumours showed a closer spatial relationship and overlap in transcriptomic features not defined
(Figure 3.2 C) by HCL and NMF clustering, and highlight the importance of applying multiple
parallel analyses to define tumour sub-groups by transcriptome analyses.
To uncover enriched genes or pathways that characterize and correlate with each CNS-
PNET sub-group, I performed supervised analyses based on relative inter-group comparison to
identify the most highly differentially expressed gene sets between sub-groups which were then
subject to further analyses with gene and pathway enrichment tools. Using the Ingenuity
Pathway Analysis tool, I observed that the 3 molecular sub-groups of CNS-PNETs exhibited
21
Figure3.3Molecularsub‐groupsofCNS‐PNETexhibitdistinctcelllineage
22
Figure 3.3 Molecular sub-groups of CNS-PNET exhibit distinct cell lineage
Molecular sub-groups of CNS-PNETs independently indicated by all three unsupervised
analyses methods (Figure 3.2) were further subjected to supervised analyses. Three molecular
sub-groups of CNS-PNET with primitive neural, oligo-neural and mesenchymal gene expression
signatures identified using supervised analyses are shown in relation to a hierarchical cluster
map. Cell lineage genes enriched in each tumour sub-group were determined using a moderated
t-test adjusted for multiple testing (FDR≤0.05). Magnitude and significance of cell lineage genes
most significantly up-regulated in tumour sub-groups are denoted respectively by fold change
and p-values.
23
significant differences in expression patterns of neural lineage and differentiation genes (Figure
3.3). Expression profiles of sub-group 1 were most significantly enriched for genes associated
with embryonic or neural stem cells which include MEIS1/2, SOX3 and the HOX 2/B3/B4/B5/C4
genes. The RNA binding protein LIN28 and the retinoic acid–binding protein CRABP148, which
are implicated in stem cell pluripotency, were amongst the top over-expressed genes with nearly
20-30 fold greater expression in sub-group 1 tumours which have recurrent C19MC
amplification, as compared to sub-group 2 and 3 CNS-PNETs. In sub-group 2 tumours,
OLIG1/2, SOX8/10 and BCAN, which are markers of oligo-neural differentiation49, were the
most highly up-regulated genes. Additional genes with up-regulated expression in sub-group 2
CNS-PNETs included PDGFRA, ERBB3, NCAM2, members of the AKT/ERK/mTOR signalling
pathway, and the C1orf61 locus which encodes hsa-miR-9, a miRNA locus with functions in
neural fate determination. Sub-group 3 tumours exhibited limited expression of neural
differentiation genes, but were characterized by significant up-regulation of genes implicated in
epithelial and mesenchymal differentiation including COL1A, COL5A, FOXJ150, MSX151 and
IGF pathway genes, IGF2 and LHX2.
In addition to differential enrichment of cell lineage genes, CNS-PNET sub-groups also
exhibited significant differences in expression of canonical, developmental signalling pathway
components (Figure 3.4). IPA analyses revealed enrichment of WNT and SHH signalling genes
in sub-group 1, in which WNT and FZD family members, as well as PTCH1 and GLI2, and
SFRP1/2, an inhibitory WNT ligand, were up-regulated. Collectively these findings suggested
the SHH and/or non-canonical WNT signalling pathways were up-regulated in sub-group 1
tumours. In contrast gene and pathway enrichment analyses of sub-group 2 tumours revealed
down-regulation of the WNT, FZD and GLI gene families, as well as SMO and PTCH1 genes,
thus indicating down-modulation of the WNT and SHH signalling pathways in these tumours.
Our analyses did not identify specific canonical signalling pathways that were up-regulated in
sub-group 2 tumours. In sub-group 3 tumours, genes involved in the TGF pathway (TGFB3,
TGFBR2, BMP4 and SMAD6), as well as PTEN signalling pathway (BCL2 and FGFR1/L1),
were up-regulated. Of note, genes implicated in the epithelial to mesenchymal transition, which
24
Figure3.4Molecularsub‐groupsofCNS‐PNETexhibitdistinctsignallingsignatures
25
Figure 3.4 Molecular sub-groups of CNS-PNET exhibit distinct signalling signatures
Signalling pathways enriched in each tumour sub-group were determined by Ingenuity Pathways
Analyses (www.ingenuity.com) of sub-group-specific gene sets derived from supervised
analyses. Most significantly altered canonical pathways determined from analyses of 343, 276
and 325 genes respectively in sub-groups 1, 2 and 3 are represented in relation to tumour sub-
group. Proportion of up- or down-regulated genes within each category are respectively indicated
in red and green.
26
Figure3.5Validationofsub‐groupspecificgeneexpressionsignatures
27
Figure3.5Validationofsub‐groupspecificgeneexpressionsignaturesQuantitative RT-PCR analysis was performed to confirm gene expression patterns observed from
supervised analyses of microarray data; gene expression levels (∆Ct) determined relative to actin
are represented. A. Expression levels of select qRT-PCR validated genes significantly enriched
within each tumour sub-group (p≤0.05) is shown in a global skyline plot; mean value of n=3 is
plotted/sample. B. Mean expression levels of individual specific lineage/signalling genes (n=3
replicas) with most robust and significant over-expression and correlation with each tumour sub-
group are represented with SEM for specific mRNA (horizontal bars). Sub-group 1, 2 and 3
specific genes are indicated by green spheres, blue squares and purple triangles respectively.
28
included JAG2, SNAI2, TWIST1 and genes that overlap with TGF and PTEN, were prominently
enriched in this sub-group.
In summary, our collective analyses of CNS-PNETs gene expression characteristics
reveal CNS-PNETs comprise three molecular sub-groups with distinct transcriptomic features
differentially enriched for cell lineage genes and signalling pathways. In order to evaluate the
clinical relevance and significance of these molecular sub-groups, it will be important to
establish robust markers that can be used to study larger archived patient cohorts.
3.3CelllineagemarkersLIN28andOLIG2distinguishmolecularsub‐groupsofCNS‐PNETs
CNS-PNETs frequently express a spectrum of neural (MAP2; synatophysin) or glial
(Nestin, GFAP) differentiation markers, which are conventionally used in diagnostic histo-
pathology to distinguish CNS-PNETs from other malignant tumours arising in the cerebrum.
However, we observed that gene expression patterns of these conventional markers did not
correlate with molecular sub-grouping of CNS-PNET. Based on the differential enrichment of
cell lineage genes in transcriptional signatures of sub-group 1 and 2 CNS-PNETs, I investigated
whether other lineage specific markers identified by gene expression analyses could serve to
distinguish CNS-PNET molecular sub-groups.
I performed qRT-PCR analyses to first validate gene enrichment patterns seen in the sub-group
specific expression profiles and to identify genes that were most consistently expressed across
individual tumours within sub-groups (Figure 3.5). I selected 8 (sub-group 1), 9 (sub-group 2),
and 15 (sub-group 3) sub-group enriched genes that exhibited at least a 2 fold change in gene
expression with a value <0.05 q for qRT-PCR validation. These analyses revealed the LIN28,
OLIG2 and IGF2 loci to be most highly differentially and consistently expressed respectively
across tumours within CNS-PNET sub-groups 1, 2 and 3 tumours. To validate LIN28, OLIG2
and IGF2 expression at the protein level, I optimized and performed IHC analyses of LIN28,
OLIG2 and IGF2 on control normal tissues and an initial test cohort of 22 tumour samples also
analysed by qRT-PCR. IGF2 protein expression could not be reliably scored on normal or
tumour tissues (Figure 3.6), however IHC expression patterns for LIN28 and OLIG2 were
robust and correlated quantitatively with gene expression levels determined by gene expression
29
arrays and qRT-PCR analyses (Figure 3.7 A). Although a subset of sub-group 1 tumours also
exhibited variable levels of OLIG2 expression, LIN28 and OLIG2 expression correlated most
robustly with gene expression data in the three different molecular sub-groups, and a
combination of LIN28 and OLIG2 protein expression status reliably distinguished the three
molecular classes of CNS-PNETs. Cytoplasmic LIN28 and/or nuclear OLIG2 positivity defined
sub-group 1, nuclear OLIG2 immuno-positivity alone identified sub-group 2, whereas LIN28
and OLIG2 double negative stained tumours were considered sub-group 3 (Figure 3.7 B).
I further examined LIN28 and OLIG2 IHC in an additional 72 primary CNS-PNETs in
order to assess the distribution of CNS-PNET sub-groups. 15/72 samples with poor IHC quality
were excluded from further analyses. Based on sub-grouping analyses of 108 primary CNS-
PNETs by gene expression and/or IHC analyses for LIN28 or OLIG2 expression, we determined
that sub-group 1, 2 and 3 tumours respectively comprised 27%, 33% and 40% of all CNS-PNETs
analysed (Figure 3.7 C).
3.4Molecularsub‐groupsofCNS‐PNETexhibitdistinctDNAcopynumberpatterns
Copy number analyses can drive gene expression signatures and are known to be
important pathogenetic drivers in other paediatric brain tumours13,15. To investigate whether
specific CNA patterns were associated with molecular sub-groups of CNS-PNETs, we examined
genomic DNA from CNS-PNETs using ultra-high resolution Illumina Omni Quad genomic
arrays which interrogates 2.5 million SNPs. With the exception of the C19MC miRNA amplicon
previously identified in our lab43, there were few other recurrent high level copy number gains or
amplification observed in a cohort of 77 CNS-PNETs. Focal MYCN and CDK4 amplification
was detected in isolated tumours. Deletions centred on CDKN2A/2B (10 of 77 tumours) and
gains of Chr1q (11 of 77 tumours) were the most frequent CNAs observed (Figure 3.8). I
analyzed CNA patterns of 59 CNS-PNETs in relation to tumour sub-groups established by gene
expression and/or IHC analyses as well as based on genomic analyses of the C19MC locus by
FISH and/or qRT-PCR. In addition to C19MC amplification (15/19), gains of chr 2 (13/19) and 3
(5/19) were also significantly associated with sub-group 1 CNS-PNETs (p≤0.05). Gains of chr
8p (4/16), 13 (4/16) and 20 (4/16) were significantly more frequently in sub-group 3 tumours as
compared to sub-group 1 and 2 tumours (p≤0.05). Notably, frequent chr 9p loss centred on the
30
Figure3.6qRT‐PCRandimmuno‐histochemicalanalysesofIGF2
31
Figure 3.6 qRT-PCR and immuno-histochemical analyses of IGF2
A. Quantitative RT-PCR analysis was performed to confirm gene expression patterns observed
from supervised analyses of microarray data; gene expression levels (∆Ct) determined relative to
actin are represented. Mean expression levels of IGF2 (n=3 replicas) is represented with SEM
(horizontal bars). Sub-group 1, 2 and 3 specific genes are indicated by green spheres, blue
squares and purple triangles respectively. B-E. IGF2 immunohistochemistry was performed on
positive control placental tissue. Trophoblastic cells in the placenta are known to express high
levels of IGF2 and exhibit cytoplasmic and membranous immuno-staining for IGF2
(www.proteinatlas.org). IHC analyses for IGF2 were performed as described in methods using
polyclonal IGF2 antibodies (Santa Cruz, California). IGF2 staining patterns of placenta tissue
treated with primary antibody at 1:50 (B), 1:75 (C), 1:100 (D) dilutions and no primary antibody
as negative control (E) are shown.
32
Figure3.7Celllineagemarkers,LIN28andOLIG2,correlatewithmolecularsub‐groupsofCNS‐PNET
33
1 2 3Total
samples
GeneExpression
14 22 15 51
IHC 15 14 28 57
Total Samples /Group
29 36 43 108
% / Group 27% 33% 40% 100%
Tumour Sub-groupC
Figure 3.7 Cell lineage markers, LIN28 and OLIG2, correlate with molecular sub-groups of CNS-PNET
A. To validate OLIG2 and LIN28 as sub-group specific markers, immuno-histochemical
analyses were performed on 22 CNS-PNET samples with molecular sub-grouping information.
LIN28 protein expression was restricted to sub-group 1 samples while OLIG2 protein expression
was predominantly restricted to sub-group 2 samples with the exception of PNET3, a C19MC
amplified tumour, which was immuno-positive for both LIN28 and OLIG2. Sub-group 3 tumours
did not exhibit immunostaining for LIN28 or OLIG2. B. Representative immuno-histochemical
staining pattern for sub-group 1, 2 and 3 molecular sub-groups of CNS-PNET (60X
magnification). Top panel shows a hematoxylin and eosin (H&E) stain, middle panel shows
characteristic, strong cytoplasmic reactivity for LIN28 in a sub-group 1 tumour but limited to no
reactivity in a sub-group 2 and 3 tumour. Lower panel shows characteristic, strong OLIG2
nuclear immuno-reactivity in a sub-group 2 tumour, but limited to no OLIG2 expression in sub-
group 1 and 3 tumours. Corresponding tissue microarray core is shown at low magnification in
inset. C. Summary of molecular sub-grouping of 108 CNS-PNETs determined by IHC and/or
gene expression analyses.
34
Figure3.8Molecularsub‐groupsofCNS‐PNEThavedistinctDNAcopynumberalterations
35
Figure 3.8 Molecular sub-groups of CNS-PNET have distinct DNA copy
number alterations A. DNA copy number profiles of primary CNS-PNET were generated
using the Human Omni 2.5 Quad SNP array and recurrent copy number patterns established
using CNA partition (Partek) analyses. Significant enrichment of specific copy number
abnormalities (CN lesion) in tumour sub-groups was determined based on 59 primary samples
with known molecular sub-grouping. Heat map shows select CN lesions which correlate
significantly, as determined using Chi-square analyses, with specific CNS-PNET sub-groups. B-
D Recurrent focal copy number abnormalities (CNAs) in CNS-PNETs are shown by heat map
and copy number profiles of specific CNAs generated using dChIP and plotted relative to
chromosome ideograms. B. Heat map and copy number plots of recurrent chr2 gains in sub-
group 1 CNS-PNET. C. Heat map, copy number plot and FISH validation of focal CDKN2A/B
loss on chr9p21 seen in 10 primary sub-group 2 and 3 CNS-PNET. D. Heat map and copy
number plots of recurrent focal C19MC locus amplification on chr19q13.41 in sub-group 1
tumours.
36
CDKN2A/2B locus were seen only in sub-groups 2 (6/22) and 3 (4/19) tumours. Collectively
our findings indicate that molecular sub-groups of CNS-PNET exhibit specific patterns of CNAs
and suggest that in addition to the C19MC amplicon other sub-group-specific CNAs may
underlie the distinct gene expression patterns seen in CNS-PNETs.
3.5Discussion
Our gene expression and copy number analyses were based on samples collected from
over 20 independent institutes. As described in section 3.1, there was substantial discordance
between institutional and repeat histo-pathology review performed by Dr. Cynthia Hawkins.
Most significantly, after appropriate immuno-histochemical or molecular testing for SMARCB1,
11% of our cohort was removed from the study as potential ATRTs. These findings underscore
the substantial challenge in current histo-pathologic classification of CNS-PNETs and the need
for unbiased molecular analyses to augment traditional histopathologic methods used for tumour
diagnoses.
A potential limitation of obtaining tissues from multiple different centres is the
heterogeneous methods of tissue handling that may confound molecular analyses. However,
given the substantial heterogeneity of tumour samples, an institutional bias was not observed
after application of conventional methods to correct for batch effects as described in methods. In
cluster analyses as described in section 3.2, I did not detect any significant difference in samples
based on their institutional origin (data not shown) suggesting that batch effects or institutional
origin was not a significant factor confounding our analyses.
Our analyses (Figure 3.4) revealed that CNS-PNETs differed in enrichment of cell
lineage and canonical signalling pathways thus suggesting different cells of origin and oncogenic
pathways may underlie the observed tumour sub-groups. A limitation of our cluster analyses is
that we performed inter-sub-group comparison with the goal of better defining tumour
heterogeneity amongst tumours diagnosed as CNS-PNETs at the molecular level. In contrast to
sub-group 1 and 3 tumours, we did not observe significant enrichment or activation of known
oncogenic pathways in sub-group 2 tumours. The nature of our analyses limits identification of a
candidate common pathway that is seen across all three sub-groups, such as the RAS pathway,
which might be the underlying driving pathway for sub-group 2. A more in depth comparison to
37
relevant normal fetal or childhood brain tissue may reveal up-regulated canonical or non-
canonical signalling pathways which specifically drive sub-group 2 tumourigenesis. However, as
CNS-PNET can arise in various sites within the hemisphere and at various ages, selecting the
appropriate control tissue may be difficult. It remains possible that pathway activation in sub-
group 2 samples might arise from post-translational mechanisms and thus will not be detected by
the genetic analyses performed.
Unlike sub-group 1 and 2 tumours, I could not identify a robust immuno-histochemical
marker for sub-group 3 tumours (Figure 3.6). Thus, further attempts at identifying other
potential markers for sub-group 3 tumours that can be tested using immuno-histochemical
methods remains to be undertaken. Candidate genes include ZIC2 or LHX2 which are highly
enriched in sub-group 3 tumours. Of note PCA analyses indicated some overlap in samples
between sub-group 2 and 3 tumours (Figure 3.2 C), thus suggesting that markers in addition to
OLIG2 may be needed to precisely distinguish sub-group 2 and 3 tumours.
Experiments performed in this chapter have provided some of the first insights into the
biology of this rare paediatric brain tumour. Specifically gene expression and copy number
analyses reveal 3 distinct sub-group of CNS-PNETs distinguished by different cell lineage gene
expression and signalling patterns which has provided an essential, initial molecular framework
for classification of CNS-PNET and will enable clinical correlative associations to be undertaken
relative to CNS-PNET molecular sub-grouping. The results of these studies are described in
Chapter 4.
38
Chapter 4
LIN28 and OLIG2 correlate with survival and metastatic phenotypes in CNS-PNETs
Data from this chapter was published in part in Picard D, Miller S, Hawkins CE, Bouffet E, Rogers HA, Chan TS, Kim SK, Ra YS, Fangusaro J, Korshunov A et al: Markers of survival and metastatic potential in childhood CNS primitive neuro-ectodermal brain tumours: an integrative genomic analysis. Lancet Oncol 2012, 13(8):838-848. and have been reproduced with copyright permission by Lancet Oncology.
Acknowledgments: Statistical analyses was performed with the help of Drs Pingzhao Hu and Derek Stephens.
39
4.0Overview
Until recently, biological studies of CNS-PNET have only been performed on small
cohorts (<10 samples per study) of tumours. Based on the first global analysis of a substantial
cohort of 41 CNS-PNETs, Li et al43 reported the identification of C19MC as a distinguishing
feature of one sub-group of CNS-PNETs. However, these studies, which were confirmed by
Korshunov et al44, were not powered to define other molecular classes of CNS-PNETs. In this
chapter, I applied the molecular sub-grouping established from studies described in Chapter 3 to
determine how molecular features of CNS-PNET sub-classes correlate with clinic-pathologic
features of an extended cohort of CNS-PNET.
Objective2
To correlate clinico-pathologic features with molecular sub-classes established for CNS-
PNETs.
Hypothesis
Molecular sub-groups of CNS-PNET will be associated with distinct clinico-pathologic
phenotypes.
4.1DefiningclinicalphenotypesofCNS‐PNETmolecularsub‐groups
We obtained 180 samples of CNS-PNETs with archived FFPE tissue for analyses. A
subset of 108 samples that had relevant clinical information and for which molecular sub-group
assignment was possible based on gene expression, IHC or CAN analyses were further examined
for clinco-pathologic and molecular correlation. I analyzed the clinical demographic data for
gender (n=107), age (n=100), and tumour stage (n=66) in section 4.1. (Table 4.1, 4.2, Appendix
1 Table A-1.1). In section 4.2, I analyzed survival time (n=58), as a whole, and in relation to
patient age.
40
Table4.1:ClinicalandmolecularfeaturesofCNS‐PNET
˜ Pearson Chi-Square ǂ Fisher's Exact Test * Log Rank (Mantel-Cox) test ¥ Feature in one individual group was compared to a pooled cohort of the other two groups Note: some patients were not included in analyses due to lack of specific clinical data - details of all patients in cohort are listed in Appendix 1 Table A-1.1
41
Table4.2:ClinicalandmolecularfeaturesofCNS‐PNETinrelationtoage
˜ Pearson Chi-Square ǂ Fisher's Exact Test * Log Rank (Mantel-Cox) test ¥ Feature in one individual group was compared to a pooled cohort of the other two groups Note: some patients were not included in analyses due to lack of specific clinical data - details of all patients in cohort are listed in Appendix 1 Table A-1.1
42
4.1.1. Gender
Gender bias has been observed in embryonal tumours such as medulloblastoma and
ATRTs, where there is a predominance of males, but no such bias has been reported in CNS-
PNET to date6. In this study, I analyzed information available from 107 sub-grouped patients (29
sub-group 1, 36 sub-group 2 and 42 sub-group 3) and I observed that sub-group 1 exhibit a
higher female to male ratio (18 females: 11 males) and a reverse trend (16 females: 26 males) in
sub-group 3 patients. However, no gender bias for sub-group 2 patients was observed (16
females: 20 males). As shown in Figure 4.1 A (Table 4.1), the female to male ratios between the
3 sub-groups, 1.64, 0.8 and 0.62 for sub-group 1, 2 and 3 respectively, are significantly different
(p=0.04).
4.1.2.Age
There is a well-known association of specific histologic categories of paediatric brain
tumours with patient age. Recently, an age-specific association for molecular sub-groups of
medulloblastoma has been reported52. Similar to that observed in medulloblastoma, my analyses
revealed an age specific distribution of CNS-PNET sub-groups. Sub-group 1 and 2 patients
exhibited a bi-modal age distribution with peaks age of diagnosis between 0-2 years for sub-
group 1 and 4-6 years for sub-group 2 CNS-PNETs, in contrast sub-group 3 patients had a single
age peak between 4-6 years. Overall sub-group 1 patients were significantly younger (median
age 2.9 years; 95% CI:2.4-5.2, p=0.005) as compared to sub-group 2 (median age 7.9; 95% CI:6-
9.7) and 3 (median age 5.9; 95% CI:4.9-7.8) patients (Table 4.1). Historical data suggest that
CNS-PNET is predominantly a disease of very young children (<3-4 years), and indeed we
observed a median age of 1.75 years for all patients in our cohort (n=100). However young
patients were significantly over-represented in sub-group 1 as compared to sub-group 2 and 3
(p=0.001) CNS-PNETs. Specifically, 77% (20/26) of sub-group 1 as compared to 28% (9/32) of
sub-group 2 and 43% (18/42) of 3 patients were ≤4 years of age at diagnosis (Figure 4.1 B,
Table 4.1). These observations suggest that CNS-PNETs presenting children ≤4 years are more
likely to have sub-group 1 tumours.
43
Figure4.1Molecularsub‐groupsofCNS‐PNETexhibitdistinctclinicalphenotypes
44
Figure 4.1 Molecular sub-groups of CNS-PNET exhibit distinct clinical phenotypes
Demographic and clinical information available on 108 primary CNS-PNET (Table 4.1, 4.2 and
Appendix 1 Table A-1.1) were examined to determine tumour sub-group correlation with: A-
Gender, B- Age at diagnosis, C- Metastatic status at diagnosis, D- Age and metastatic status at
diagnosis. Significance was determined using ANOVA (gender), Chi-square (age), and 2-sided
Fisher’s exact test (gender and metastatic status at diagnosis). Number of patients in each
category is indicated within bar graphs.
45
4.1.3.Metastasis
Metastasis is a clinical feature linked to poorer outcomes in medulloblastoma and other
embryonal brain tumours and represents an important factor in determining if a patient will
receive cranio-spinal radiation. Molecular sub-groups of CNS-PNET exhibited significant
differences in incidence of tumour metastases. Sub-group 3 tumours demonstrated the highest
incidence of tumour metastases at diagnosis: 53% (10/19) sub-group 3 patients were metastatic
(M+) at diagnosis while only 26% (5/19) sub-group 1 and 15% (3/20) sub-group 2 patients were M+
at diagnosis (p=0.037) (Figure 4.1 C). Proportion of localized (M0) to metastatic (M+) tumours
in sub-groups 1, 2 and 3 were respectively 2.8, 5.67 and 0.9 (p=0.03 (Table 4.1)).
Interestingly, although metastatic disease is reported to be more frequent in younger
children with embryonal brain tumours, analyses performed with the previous stratification for
age (≤ or > 4 years) showed no significant difference in frequency of metastases based on age.
This may be due to the small numbers found in the ≤4 category; sub-group 1 (n=15), sub-group 2
(n=6) and sub-group 3 (n=4). However, incidence of tumour metastases remained significantly
different amongst CNS-PNET sub-groups diagnosed in older children (Figure 4.1 D, Table 4.2).
A majority (9/15; 60%) of sub-group 3 patients > 4 years of age at diagnosis had metastatic
presentation as compared to only 1/4 (25%) of sub-group 1 and 2/14 (14%) sub-group 2 patients.
Proportion of M0 to M+ tumours, which were respectively 3, 6 and 0.67 for sub-group 1, 2 and 3
CNS-PNETs in children > 4 years, differed significantly in comparison of sub-group 3 to a
combined cohort of sub-group 1 and 2 patient (p=0.033). Since sub-group 1 consisted of 4
patients, I compared the incidence of metastases in patients >4 years between sub-group 2 and 3
alone and saw that incidence of metastases in sub-group 3 remained significant (p=0.014)
(Figure 4.1 D, Table 4.2). Notably, the incidence of metastases across sub-groups did not
correlate with patient outcomes, specifically although sub-group 3 tumours exhibited the highest
incidence of metastases; sub-group 1 patients exhibited the worst survival.
4.2SurvivalfeaturesofCNS‐PNETmolecularsub‐groups
Li et al43 and Korshunov et al44 have shown that C19MC amplified tumours are more
aggressive than other CNS-PNETs, however, specific molecular sub-grouping of the other non-
C19MC amplified CNS-PNETs had not yet been established. I compared survival features of all
46
3 molecular sub-groups of patients. Log rank analysis confirmed overall survival for group 1 was
significantly less than group 2 and 3 patients. Median survival for sub-group 1, 2 and 3 were
respectively 0.8 years (95% CI:0.47-1.2), 1.8 years (95% CI:1.4-2.3) and 4.3 years (95% CI:
0.82-7.8) p= 0.019 (Figure 4.2 A, Table 4.1). With the exception of two longer term survivors
which were sub-grouped by LIN28 IHC, all sub-group 1 patients were deceased within 4.2 years
of diagnosis.
As the majority of sub-group 1 tumours arise in younger children who are often treated
heterogeneously with radiation sparing therapeutic approaches due to concerns of neuro-
cognitive damage7,8, I examined whether the poor prognostic association of LIN28 expression in
sub-group 1 tumours held true for older children who are conventionally prescribed intensified
treatment regimens with higher dose cranio-spinal irradiation. We had very limited treatment
information for our patients, therefore, as most infant brain tumour protocols enroll patients up to
3-4 years of age7,8,23, we stratified patients by age ≤ or >4 years, to remove potential treatment
biases based on age. As shown in Figure 4.2 B and Table 4.2, while overall survival for all
young patients was similarly dismal, older children with LIN28 sub-group 1 tumours fared
significantly worse (median survival of 0.5 years, 95% CI: 0-1, p=0.004) than patients in sub-
group 2 (median survival of 1.8 years, 95% CI:1.5-2.2) and 3 (median survival of 4.8 years, 95%
CI:1.6-8). These findings indicate that sub-group 1 patients, identified by gene expression or
immuno-positivity for LIN28, describe a particularly high risk sub-group of CNS-PNETs across
all ages.
4.3Discussion
Collectively my analyses demonstrate distinct clinico-pathologic features of molecular
sub-groups of CNS-PNETs. This information provides valuable diagnostic and prognostic tools
for in depth biological studies of disease mechanisms specific to CNS-PNET sub-groups and for
informing prospective treatment trials.
Our findings suggest that current treatment regimens for sub-group 1 CNS-PNETs are
largely ineffective and that such children should be considered for novel therapeutic approaches.
Older children with CNS-PNETs represented in the sub-group 2 and 3 patients routinely received
higher dose cranio-spinal irradiations. Interestingly, while overall survival for all young
47
Figure4.2Molecularsub‐groupsofCNS‐PNETexhibitdistinctsurvivalfeatures
48
Figure 4.2 Molecular sub-groups of CNS-PNET exhibit distinct survival features
Demographic and clinical information available on 66 primary CNS-PNET (Table 4.1, 4.2 and
Appendix Table A-1.1) were examined to determine tumour sub-group correlation with A-
Overall survival, and B- Overall survival and age. Significance was determined using log-rank
(survival) analyses.
49
patients was similarly dismal, sub-group 3 patients aged > 4 years of age trended (p=0.089)
towards better survival than sub-group 2 or sub-group 1, which had the worst overall survival.
This is particularly interesting as sub-group 3 tumours exhibited the highest incidence of
metastases which is generally considered a high risk feature of embryonal brain tumours. Thus
this observation potentially has significant implications for risk stratification for CNS-PNET
patients. Future studies with larger cohorts will clearly be important for validating this early
observation. We observed that sub-group 2 tumours, which are predominantly localized tumours,
also expressed high levels of OLIG2, a marker associated with glial lineage. These findings
reflect long standing challenges in distinguishing a sub-set of malignant gliomas arising in the
cerebral hemispheres with poorly differentiated features from CNS-PNETs45. The extent to
which sub-group 2 CNS-PNETs overlap with malignant glial entities remains to be established
with direct comparative studies of histologically verified glioblastoma.
In summary, our findings demonstrate for the first time that differential expression of cell
lineage markers LIN28 and OLIG2 distinguishes 3 molecular sub-groups of CNS-PNET and
identifies disease sub-groups with specific clinical features. Although our studies are limited by
the lack of complete clinical information for all patients, it represents a unique and highly
valuable study for this rare tumour entity. Our findings that LIN28 and OLIG2 respectively
identify CNS-PNETs with very different rates of treatment failures and distinct propensity for
metastasis has potential far reaching implications for future treatment of CNS-PNETs.
50
Chapter 5
Discussion and future directions
51
5.0Discussion
Advances in paediatric CNS-PNET research have been limited due to rare tumour
incidence6, lack of information regarding clinical and biological features of this disease and lack
of genetic markers for more precise histo-pathological diagnosis24,30. In this integrative study, I
examined the largest cohort of primary hemispheric CNS-PNET and have observed that CNS-
PNET comprise 3 distinct molecular and clinical sub-groups. By exploiting distinct molecular
expression signatures for each molecular sub-group, I was able to identify immuno-
histochemical markers that can be used to identify CNS-PNET sub-groups.
In my thesis, I have also further characterized the C19MC amplicon associated molecular
sub-group, which our group previously described2, and confirmed that these are distinctly
aggressive tumours and occur almost exclusively in children < 4 years of age. I showed that
these sub-group 1 tumours express high levels of the pluripotentcy gene LIN28, suggesting an
important oncogenic role for LIN28 in the pathogenesis of the sub-group 1 CNS-PNETs.
Interestingly, recent studies from our lab have demonstrated that C19MC amplified tumours with
or without high LIN28 expression span several histologic classes of CNS-PNETs including
ETANTR, medulloepithelioma and ependymoblastoma, as well as a proportion of “classic”
CNS-PNET53. Furthermore studies from our lab54 show that the LIN28/let-7 miRNA axis, which
is important in regulation of stemness, cellular metabolism and disease state55-58, is functional in
C19MC amplified/LIN28+ CNS-PNETs. Importantly, these studies have shown that LIN28
drives aberrant IGF/PI3K/mTOR signalling and that tumour cell lines with C19MC amplification
or high LIN28 expression are sensitive to Rapamycin, a pharmacologic mTOR antagonist which
represents a novel, candidate therapeutic for sub-group 1 CNS-PNETs.
LIN28 and OLIG2 are cell lineage markers expressed at various times during normal
development. LIN28 is an early precursor/primitive cell marker which has been reported to be
highly expressed in the developing germ cells and in germ cell associated tumours. To date, no
direct genetic alteration has been observed in our genomics data to account for high LIN28
expression. Ongoing epigenomic/methylation profiling studies in our lab indicate that LIN28
may be regulated at the epigenetic level in these tumours, thus suggesting that the association of
high LIN28 expression in sub-group 1 tumours may reflect an early lineage tumour cell of origin.
High expression of OLIG2, which has been reported in both neural and glial cell lineage
52
precursor cell types, was observed in sub-group 2 tumours which suggest an oligodendroglial
precursor cell of origin. Notably, OLIG2 is known to be co-expressed with ASCL1, a primitive
neural marker in oligodendroglial precursor cells (OPCs). Current data suggest that upon loss of
ASCL1 in the sub ventricular zone, OPCs differentiate into an astrocytic lineage59, expression of
the NKX2-2 and SOX10 gene are associated with a down-regulation of the myelination and
astrocytic commitment program by OPCs60,61. In our study, we observed variable amounts of
OLIG2 expression in sub-group 2 tumours, which was associated with high levels of SOX10.
Notably, as a subset of OLIG2 tumours also expressed high levels of ASCL1 and NKX2-2, these
findings suggest OLIG2+ve sub-group 2 CNS-PNETs exhibit a range of neural/glial features. As
OLIG2 is expressed in many tumours of glial lineage, it is possible that a subset of sub-group 2
CNS-PNET may have overlapping molecular features with malignant glial tumours of childhood.
However, H3.3 mutations, a defining feature of malignant childhood gliomas46, have not been
reported in CNS-PNETs. In contrast to sub-group 1 and 2 tumours, we observed that sub-group 3
tumours were enriched for genes associated with mesenchymal differentiation. Although the
potential cellular origin of sub-group 3 CNS-PNETs could not be determined based on our gene
expression data, we did observe enrichment ZIC262 and LHX263 which are expressed in early
neural precursors and have functions in brain patterning. Our study is limited by relatively small
numbers of tumour with substantial heterogeneity. Therefore, future in depth analyses of CNS-
PNET genomes and transcriptomes by global sequencing methods will better elucidate the
genetic spectrum of sub-group 2 and 3 CNS-PNETs.
5.1Clinicalrelevanceoffindings
Prior clinical studies of CNS-PNETs have often grouped tumours from various
anatomical locations including pineoblastoma as a supratentorial PNET and have also included
cohorts of medulloblastoma. Therefore patient features and survival trends have difficult to
interpret. My study represents the first study on a large cohort of primary CNS-PNET restricted
to only the cerebral hemisphere. My data indicates that hemispheric CNS-PNETs comprise three
molecular and clinically distinct sub-groups of CNS-PNET with characteristic age, gender,
metastasis and survival features.
Current treatment options for CNS-PNET are largely based on histologic similarities to
medulloblastoma, a more common malignant childhood brain tumour. However, current
53
approaches to CNS-PNETs treatment are often ineffective and are limited by lack of biological
studies to inform patient treatments. To date, only 2 biological studies2,44 of modest cohorts of
CNS-PNET have been undertaken, confirming an association of C19MC with more aggressive
features than other CNS-PNETs. In this study, we integrated molecular and genomic analyses
with immunohistochemistry of a large cohort of primary hemispheric CNS-PNETs and
demonstrated that cell lineage markers correlate with and reveal additional clinically relevant
sub-groups of CNS-PNETs. Data from my study suggest that LIN28 and OLIG2 represent
promising and readily applied markers that should be further explored in prospective patient
cohorts.
5.2Futuredirections
I have shown that chr 3q with a minimal region of overlap (MRO) centered on IGF2BP2,
is gained in group 1 tumours; interestingly, IGF2BP2 is significantly up-regulated in group 1
tumours. Indeed, although I have shown differences in CNAs between molecular sub-groups, I
observed many additional alterations that span sub-groups (e.g chr 1q, and MROs centered on
WNT16 (chr 7q) and RHOG (chr 11p) (Figure 3.8). The nature of genes altered by these CNAs,
and how they contribute to gene expression patterns and CNS-PNET tumourigenesis, remains to
be determined by ongoing and future sequencing studies.
Cell lines for CNS-PNETs are rare. Therefore, CNS-PNET cell lines which recapitulate
the histology and expression patterns of primary tumours are important tools. To further study
sub-group 2/3 tumours, it is imperative to generate such reagents. Recently a sub-group 3 cell
line and intra-cranial xenograft model was reported by Zhigang Liu et al64. Furthermore
preliminary reports of genetic models resembling sub-group 2 CNS-PNET has also been reported
in mice and zebra fish65. Cross-species analysis of these models with human CNS-PNETs will be
very powerful, not only for elucidating the cell of origin, but also for determining important
driver genes necessary for the formation of CNS-PNETs.
The biological relationship of CNS-PNETs arising in different anatomical locations
remains unclear. As my study comprised a substantial cohort of tumours restricted to the
hemispheres, data generated from my analyses will enable comparative studies of large tumour
cohorts across CNS-PNETs. Specifically these molecular and clinical data sets will be important
54
for determining the relationship of CNS-PNETs to pineoblastomas, as well as other hemispheric
tumours. Such studies will help refine and generate more powerful diagnostic markers that will
help distinguish malignant childhood brain tumours such as CNS-PNETs and other malignant
hemispheric tumours for which treatment approaches are substantially different.
In summary, data from my thesis has generated a substantial body of knowledge to fuel
both biological and clinical investigations of childhood CNS-PNETs and other related tumours.
Specifically, I have identified promising candidate disease markers that have the potential to be
highly informative diagnostic and prognostic markers which will help advance the therapeutic
challenges posed by patient with CNS-PNETs.
55
References
56
References 1. Kohler, B.A. et al. Annual report to the nation on the status of cancer, 1975‐2007, featuring
tumors of the brain and other nervous system. J Natl Cancer Inst 103, 714‐36 (2011). 2. Turner, C.D., Rey‐Casserly, C., Liptak, C.C. & Chordas, C. Late effects of therapy for pediatric
brain tumor survivors. J Child Neurol 24, 1455‐63 (2009). 3. Wong, K.‐K. et al. Genome‐Wide Allelic Imbalance Analysis of Pediatric Gliomas by Single
Nucleotide Polymorphic Allele Array. Cancer Research 66, 11172‐11178 (2006). 4. Grimmer, M.R. & Weiss, W.A. Childhood tumors of the nervous system as disorders of normal
development. Current Opinion in Pediatrics 18, 634‐638 (2006). 5. Rorke, L.B. The cerebellar medulloblastoma and its relationship to primitive neuroectodermal
tumors. J Neuropathol Exp Neurol 42, 1‐15 (1983). 6. Louis, D.N. & Wiestler, O.D. WHO Classification of Tumours of the Central Nervous System,
(IARC, Lyon, 2007). 7. Fangusaro, J. et al. Intensive chemotherapy followed by consolidative myeloablative
chemotherapy with autologous hematopoietic cell rescue (AuHCR) in young children with newly diagnosed supratentorial primitive neuroectodermal tumors (sPNETs): report of the Head Start I and II experience. Pediatr Blood Cancer 50, 312‐8 (2008).
8. Timmermann, B. et al. Role of radiotherapy in supratentorial primitive neuroectodermal tumor in young children: results of the German HIT‐SKK87 and HIT‐SKK92 trials. J Clin Oncol 24, 1554‐60 (2006).
9. Packer, R.J. et al. Phase III study of craniospinal radiation therapy followed by adjuvant chemotherapy for newly diagnosed average‐risk medulloblastoma. J Clin Oncol 24, 4202‐8 (2006).
10. Gilbertson, R. Paediatric embryonic brain tumours. biological and clinical relevance of molecular genetic abnormalities. Eur J Cancer 38, 675‐85 (2002).
11. Pomeroy, S.L. et al. Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 415, 436‐42 (2002).
12. Thompson, M.C. et al. Genomics identifies medulloblastoma subgroups that are enriched for specific genetic alterations. J Clin Oncol 24, 1924‐31 (2006).
13. Northcott, P.A. et al. Medulloblastoma comprises four distinct molecular variants. J Clin Oncol 29, 1408‐14 (2011).
14. Kool, M. et al. Integrated genomics identifies five medulloblastoma subtypes with distinct genetic profiles, pathway signatures and clinicopathological features. PLoS One 3, e3088 (2008).
15. Cho, Y.J. et al. Integrative genomic analysis of medulloblastoma identifies a molecular subgroup that drives poor clinical outcome. J Clin Oncol 29, 1424‐30 (2011).
16. Ellison, D.W. et al. Definition of disease‐risk stratification groups in childhood medulloblastoma using combined clinical, pathologic, and molecular variables. J Clin Oncol 29, 1400‐7 (2011).
17. Packer, R.J. et al. Long‐term neurologic and neurosensory sequelae in adult survivors of a childhood brain tumor: childhood cancer survivor study. J Clin Oncol 21, 3255‐61 (2003).
18. Gajjar, A. et al. Risk‐adapted craniospinal radiotherapy followed by high‐dose chemotherapy and stem‐cell rescue in children with newly diagnosed medulloblastoma (St Jude Medulloblastoma‐96): long‐term results from a prospective, multicentre trial. Lancet Oncol 7, 813‐20 (2006).
19. McNeil, D.E., Cote, T.R., Clegg, L. & Rorke, L.B. Incidence and trends in pediatric malignancies medulloblastoma/primitive neuroectodermal tumor: a SEER update. Surveillance Epidemiology and End Results. Med Pediatr Oncol 39, 190‐4 (2002).
57
20. Biswas, S. et al. Non‐pineal supratentorial primitive neuro‐ectodermal tumors (sPNET) in teenagers and young adults: Time to reconsider cisplatin based chemotherapy after cranio‐spinal irradiation? Pediatr Blood Cancer 52, 796‐803 (2009).
21. Reddy, A.T., Janss, A.J., Phillips, P.C., Weiss, H.L. & Packer, R.J. Outcome for children with supratentorial primitive neuroectodermal tumors treated with surgery, radiation, and chemotherapy. Cancer 88, 2189‐93 (2000).
22. Johnston, D.L. et al. Supratentorial primitive neuroectodermal tumors: a Canadian pediatric brain tumor consortium report. J Neurooncol 86, 101‐8 (2008).
23. Pizer, B.L. et al. Analysis of patients with supratentorial primitive neuro‐ectodermal tumours entered into the SIOP/UKCCSG PNET 3 study. Eur J Cancer 42, 1120‐8 (2006).
24. Li, M.H., Bouffet, E., Hawkins, C.E., Squire, J.A. & Huang, A. Molecular genetics of supratentorial primitive neuroectodermal tumors and pineoblastoma. Neurosurg Focus 19, E3 (2005).
25. Miller, S. et al. Genome‐wide molecular characterization of central nervous system primitive neuroectodermal tumor and pineoblastoma. Neuro Oncol 13, 866‐79 (2011).
26. Fangusaro, J., Massimino, M., Rutkowski, S. & Gururangan, S. Non‐cerebellar primitive neuroectodermal tumors (PNET): summary of the Milan consensus and state of the art workshop on marrow ablative chemotherapy with hematopoietic cell rescue for malignant brain tumors of childhood and adolescents. Pediatr Blood Cancer 54, 638‐40 (2010).
27. Kim, D.G. et al. Supratentorial primitive neuroectodermal tumors in adults. J Neurooncol 60, 43‐52 (2002).
28. Takeuchi, H. et al. Atypical neuronal‐glial tumors of cerebral hemispheres in adults with PNET‐like component: clinicopathological features of 5 cases. Clin Neuropathol 22, 47‐56 (2003).
29. Jakacki, R.I. Pineal and nonpineal supratentorial primitive neuroectodermal tumors. Childs Nerv Syst 15, 586‐91 (1999).
30. Burger, P.C. Supratentorial primitive neuroectodermal tumor (sPNET). Brain Pathol 16, 86 (2006).
31. Gessi, M. et al. Embryonal tumors with abundant neuropil and true rosettes: a distinctive CNS primitive neuroectodermal tumor. Am J Surg Pathol 33, 211‐7 (2009).
32. Eberhart, C.G., Brat, D.J., Cohen, K.J. & Burger, P.C. Pediatric neuroblastic brain tumors containing abundant neuropil and true rosettes. Pediatr Dev Pathol 3, 346‐52 (2000).
33. Louis, D.N. & International Agency for Research on Cancer. WHO classification of tumours of the central nervous system, 309 p. (International Agency for Research on Cancer, Lyon, 2007).
34. Fomchenko, E.I. & Holland, E.C. Mouse models of brain tumors and their applications in preclinical trials. Clin Cancer Res 12, 5288‐97 (2006).
35. Marcus, D.M. et al. Primitive neuroectodermal tumor of the midbrain in a murine model of retinoblastoma. Investigative Ophthalmology & Visual Science 32, 293‐301 (1991).
36. Tong, W.M. et al. Null mutation of DNA strand break‐binding molecule poly(ADP‐ribose) polymerase causes medulloblastomas in p53(‐/‐) mice. American Journal of Pathology 162, 343‐352 (2003).
37. Zindy, F. et al. Hemangiosarcomas, Medulloblastomas, and Other Tumors in Ink4c/p53‐null Mice. Cancer Research 63, 5420‐5427 (2003).
38. Momota, H. & Holland, E.C. Mouse models of CNS embryonal tumors. Brain Tumor Pathol 26, 43‐50 (2009).
39. Inda, M.M. et al. Genetic heterogeneity in supratentorial and infratentorial primitive neuroectodermal tumours of the central nervous system. Histopathology 47, 631‐7 (2005).
40. McCabe, M.G. et al. High‐resolution array‐based comparative genomic hybridization of medulloblastomas and supratentorial primitive neuroectodermal tumors. J Neuropathol Exp Neurol 65, 549‐61 (2006).
58
41. Pfister, S. et al. Supratentorial primitive neuroectodermal tumors of the central nervous system frequently harbor deletions of the CDKN2A locus and other genomic aberrations distinct from medulloblastomas. Genes Chromosomes Cancer 46, 839‐51 (2007).
42. Russo, C. et al. Comparative genomic hybridization in patients with supratentorial and infratentorial primitive neuroectodermal tumors. Cancer 86, 331‐9 (1999).
43. Li, M. et al. Frequent amplification of a chr19q13.41 microRNA polycistron in aggressive primitive neuroectodermal brain tumors. Cancer Cell 16, 533‐46 (2009).
44. Korshunov, A. et al. Focal genomic amplification at 19q13.42 comprises a powerful diagnostic marker for embryonal tumors with ependymoblastic rosettes. Acta Neuropathol 120, 253‐60 (2010).
45. Gessi, M. et al. H3.3 G34R mutations in pediatric primitive neuroectodermal tumors of central nervous system (CNS‐PNET) and pediatric glioblastomas: possible diagnostic and therapeutic implications? J Neurooncol 112, 67‐72 (2013).
46. Khuong‐Quang, D.A. et al. K27M mutation in histone H3.3 defines clinically and biologically distinct subgroups of pediatric diffuse intrinsic pontine gliomas. Acta Neuropathol 124, 439‐47 (2012).
47. Brunet, J.P., Tamayo, P., Golub, T.R. & Mesirov, J.P. Metagenes and molecular pattern discovery using matrix factorization. Proc Natl Acad Sci U S A 101, 4164‐9 (2004).
48. Skottman, H. et al. Gene expression signatures of seven individual human embryonic stem cell lines. Stem Cells 23, 1343‐56 (2005).
49. Ligon, K.L., Fancy, S.P., Franklin, R.J. & Rowitch, D.H. Olig gene function in CNS development and disease. Glia 54, 1‐10 (2006).
50. Jacquet, B.V. et al. Specification of a Foxj1‐dependent lineage in the forebrain is required for embryonic‐to‐postnatal transition of neurogenesis in the olfactory bulb. J Neurosci 31, 9368‐82 (2011).
51. Houzelstein, D. et al. The homeobox gene Msx1 is expressed in a subset of somites, and in muscle progenitor cells migrating into the forelimb. Development 126, 2689‐701 (1999).
52. Northcott, P.A. et al. Medulloblastomics: the end of the beginning. Nat Rev Cancer 12, 818‐34 (2012).
53. Korshunov, A. et al. Embryonal tumor with abundant neuropil and true rosettes (ETANTR), ependymoblastoma, and medulloepithelioma share molecular similarity and comprise a single clinicopathological entity. Acta Neuropathol Epub ahead of print(2013).
54. Spence, T. et al. A novel C19MC amplified cell line links Lin28/let‐7 to mTOR signaling in embryonal tumor with multilayered rosettes. Neuro Oncol 16, 62‐71 (2014).
55. West, J.A. et al. A role for Lin28 in primordial germ‐cell development and germ‐cell malignancy. Nature 460, 909‐13 (2009).
56. Viswanathan, S.R. et al. Lin28 promotes transformation and is associated with advanced human malignancies. Nature genetics 41, 843‐8 (2009).
57. Zhu, H. et al. The Lin28/let‐7 axis regulates glucose metabolism. Cell 147, 81‐94 (2011). 58. Frost, R.J. & Olson, E.N. Control of glucose homeostasis and insulin sensitivity by the Let‐7 family
of microRNAs. Proceedings of the National Academy of Sciences of the United States of America 108, 21075‐80 (2011).
59. Nakatani, H. et al. Ascl1/Mash1 promotes brain oligodendrogenesis during myelination and remyelination. J Neurosci 33, 9752‐68 (2013).
60. Qi, Y. et al. Control of oligodendrocyte differentiation by the Nkx2.2 homeodomain transcription factor. Development 128, 2723‐33 (2001).
61. Srinivasan, R. et al. Genome‐wide analysis of EGR2/SOX10 binding in myelinating peripheral nerve. Nucleic Acids Res 40, 6449‐60 (2012).
59
62. Elms, P., Siggers, P., Napper, D., Greenfield, A. & Arkell, R. Zic2 is required for neural crest formation and hindbrain patterning during mouse development. Dev Biol 264, 391‐406 (2003).
63. Hoch, R.V., Rubenstein, J.L. & Pleasure, S. Genes and signaling events that establish regional patterning of the mammalian forebrain. Semin Cell Dev Biol 20, 378‐86 (2009).
64. Liu, Z. et al. A Patient Tumor‐derived Orthotopic Xenograft Mouse Model Replicating the Group 3
Supratentorial Primitive Neuroectodermal Tumor in Children. Neuro‐Oncology (2013). 65. 2013 Pediatric Neuro‐Oncology Basic and Translational Research Conference. Neuro Oncol 15,
i1‐i51 (2013).
60
Appendix
Appendix 1-2 were published in Picard D, Miller S, Hawkins CE, Bouffet E, Rogers HA, Chan TS, Kim SK, Ra YS, Fangusaro J, Korshunov A et al: Markers of survival and metastatic potential in childhood CNS primitive neuro-ectodermal brain tumours: an integrative genomic analysis. Lancet Oncol 2012, 13(8):838-848. Reproduced with copyright permission from Lance Oncology
61
Appendix 1
62
TableA‐1.1Clinicopathologiccharacteristicsoftumoursamples
Annotation Sub-
group Gender Diagnosis
Age at Diagnosis
(years) Treatment Metastasis
Status Survival Status
Survival Time
(years)
PNET3 1 female PNET/EP 4.08 Chemo M0 Dead 0.5
PNET4 2 female PNET 10.75 Chemo/XRT M0 Dead 0.67
PNET5 1 female PNET/EP 1.67 Chemo M0 Dead 0.58
PNET6 1 female PNET/EP 2.42 Chemo M0 Dead 1
PNET7 2 male PNET n/a n/a n/a n/a n/a
PNET8 3 male PNET 6 n/a M+ Alive n/a
PNET9 3 male PNET 4.08 n/a M+ Alive 0.42
PNET10 n/a n/a PNET n/a n/a n/a n/a n/a
PNET11 n/a n/a PNET n/a n/a n/a n/a n/a
PNET15 1 male PNET/ME 0.42 None M0 Alive n/a
PNET17 3 male PNET 1.83 Chemo/XRT M0 Dead 6.33
PNET20 3 female PNET 11 n/a n/a Dead 1
PNET22 2 female PNET 9.67 n/a n/a Dead 2
PNET25 3 male PNET 2 n/a n/a n/a n/a
PNET28 3 male PNET 4 n/a n/a n/a n/a
PNET30 2 male PNET 12 n/a n/a n/a n/a
PNET31 2 female PNET n/a n/a n/a n/a n/a
PNET36 2 male PNET 10 Chemo/XRT n/a Alive 4
PNET37 3 female PNET 8 Chemo/XRT n/a Alive 3.92
PNET40 1 male PNET 2.5 None M0 Dead 0.01
PNET42 1 female PNET 3.17 Chemo/XRT M0 Dead 0.83
PNET43 1 male PNET 1.5 Chemo M+ Dead 0.83
PNET44 2 male PNET 11 Chemo M0 Dead 1.75
PNET47 2 female PNET 2.58 Chemo/XRT M+ Alive 11.17
PNET48 3 female PNET 1 Chemo M0 Dead 2.33
PNET49 n/a male PNET 17 n/a n/a n/a n/a
PNET51 n/a female PNET 3 n/a n/a n/a n/a
PNET52 n/a female PNET 8 n/a n/a n/a n/a
PNET53 n/a female PNET 3 n/a n/a n/a n/a
PNET54 1 male PNET/EP 3 n/a M0 n/a n/a
PNET55 n/a female PNET 2 n/a n/a n/a n/a
PNET56 1 female PNET/ETNATR 1.67 n/a M0 Dead 0.25
PNET59 3 n/a PNET n/a n/a n/a n/a n/a
PNET61 3 female PNET 0.67 n/a n/a Dead n/a
PNET62 3 female PNET 1 n/a n/a Dead n/a
PNET64 3 male PNET 16 n/a n/a Alive n/a
PNET65 3 male PNET 10 n/a n/a Alive n/a
PNET66 2 female PNET 6 n/a n/a Dead n/a
PNET67 1 male PNET/ETNATR 3 n/a n/a Dead n/a
63
PNET68 3 male PNET 2 n/a n/a Alive n/a
PNET71 2 male PNET 7 n/a n/a Alive n/a
PNET72 3 male PNET 0.06 n/a n/a Dead n/a
PNET75 3 male PNET 0.92 n/a n/a Dead n/a
PNET78 3 male PNET 4 n/a n/a Alive n/a
PNET79 3 male PNET/ME 4 n/a n/a Dead n/a
PNET81 1 female PNET/ETNATR 3 n/a n/a Alive n/a
PNET82 1 female PNET/ME 12 n/a n/a Alive n/a
PNET91 1 female PNET n/a n/a n/a n/a n/a
PNET93 1 female PNET n/a n/a n/a n/a n/a
PNET94 1 female PNET 15 n/a n/a n/a n/a
PNET95 3 female PNET 3.08 n/a n/a n/a n/a
PNET96 3 male PNET 8 n/a n/a n/a n/a
PNET97 2 male PNET 12 n/a n/a n/a n/a
PNET99 n/a male PNET 6 n/a M+ Dead 1.25
PNET100 n/a female PNET 11 n/a M+ Dead 8.7
PNET101 n/a male PNET 3 n/a M0 Alive 6.6
PNET105 n/a female PNET n/a n/a n/a n/a n/a
PNET106 n/a male PNET n/a n/a n/a n/a n/a
PNET109 1 female PNET 1.58 n/a n/a Dead n/a
PNET111 1 male PNET 2.33 n/a n/a Dead n/a
PNET112 2 male PNET 2 n/a n/a Dead n/a
PNET113 3 male PNET 7 n/a n/a Dead n/a
PNET114 2 male PNET n/a n/a n/a n/a n/a
PNET116 2 male PNET 14.2 Chemo/XRT M0 Dead 1.79
PNET118 1 male PNET 2.9 XRT M0 Dead 0.38
PNET119 2 male PNET 17.9 Chemo/XRT M0 Dead 0.58
PNET122 3 male PNET 16.7 Chemo/XRT M0 Dead 4.33
PNET123 2 male PNET 17.9 Chemo M+ Dead 0.88
PNET126 2 male PNET 3 n/a n/a n/a n/a
PNET129 2 male PNET 8 n/a n/a n/a n/a
PNET131 3 female PNET 0.83 n/a n/a n/a n/a
PNET132 3 male PNET 4 n/a n/a n/a n/a
PNET135 1 male PNET n/a n/a n/a Dead 0
PNET138 1 male PNET 2.83 n/a M0 Dead 1.58
PNET139 n/a n/a PNET n/a n/a n/a n/a n/a
PNET140 n/a n/a PNET n/a n/a n/a n/a n/a
PNET141 n/a n/a PNET n/a n/a n/a n/a n/a
PNET142 n/a n/a PNET n/a n/a n/a n/a n/a
PNET143 2 female PNET 1.92 n/a M0 Dead 0.75
PNET146 3 male PNET 10.58 n/a M+ Alive 3.58
PNET148 1 female PNET 2.2 n/a M0 Alive 3.6
PNET149 2 male PNET 6.42 n/a M0 Dead 1.25
64
PNET157 3 male PNET 2.25 n/a M0 Dead 8.17
PNET158 2 male PNET 0.03 n/a M0 Dead 6.08
PNET160 3 male PNET 5.75 XRT M+ Alive 1.75
PNET161 1 female PNET 1.75 Chemo/XRT M0 Dead 3.17
PNET163 2 female PNET 1.42 Chemo M0 Dead 0.08
PNET164 1 female PNET 8.92 n/a M+ Dead 1.33
PNET166 n/a female PNET 0.06 n/a M0 Dead 0.25
PNET167 3 female PNET 1.08 n/a n/a Dead 2.71
PNET169 2 female PNET 4.08 n/a n/a Dead 0.17
PNET170 2 female PNET n/a None n/a n/a n/a
PNET171 2 male PNET 11.17 Chemo/XRT M+ Dead 4.42
PNET172 3 female PNET 3.5 Chemo/XRT M0 Alive 2.17
PNET173 2 male PNET 13.17 Chemo/XRT M0 Dead 1.83
PNET174 2 male PNET 7.83 Chemo/XRT M0 Dead 2.33
PNET187 2 female PNET 15.83 Chemo/XRT M0 Alive 2.92
PNET188 2 male PNET 1.67 Chemo M0 Dead 3.42
PNET190 3 male PNET 11.75 XRT M+ Dead 2.5
PNET191 2 female PNET 4.42 Chemo M0 Alive 5.17
PNET196 2 female PNET 1.67 None M0 Dead 0
PNET197 3 male PNET 10.17 Chemo/XRT M0 Alive 7.5
PNET199 2 female PNET 9.67 Chemo/XRT M0 Dead 2.08
PNET200 3 female PNET 7 Chemo/XRT M+ Dead 1.67
PNET226 1 female PNET 3.4 n/a n/a Dead 0
PNET255 1 female PNET 1.92 Chemo/XRT M+ Alive 4.67
PNET256 2 female PNET 13 Chemo/XRT M0 Dead 1.17
PNET258 3 male PNET 7.5 Chemo/XRT M+ Alive 2.08
PNET259 1 male PNET 3.67 Chemo/XRT M+ Dead 0.67
PNET260 2 female PNET 8.5 XRT M0 Alive 7.75
PNET265 n/a female PNET 1.33 Chemo M+ Dead 0.5
PNET266 3 male PNET 8.92 XRT M+ Dead 5.92
PNET267 n/a female PNET 4.25 Chemo/XRT M0 Dead 2
PNET268 n/a male PNET 2.75 None M0 Dead 0
PNET269 n/a male PNET 1.75 None M0 Dead 0
PNET270 n/a female PNET 0.42 Chemo M+ Dead 1.17
PNET271 n/a female PNET 5.92 XRT M0 Dead 0.58
PNET272 1 female PNET 0.83 Chemo M+ Dead 3.42
PNET273 1 female PNET 8.25 Chemo/XRT M0 Dead 0.67
PNET274 3 male PNET 15.5 Chemo/XRT M0 Alive 21.33
PNET276 n/a male PNET 10.25 Chemo/XRT M0 Dead 1.33
PNET277 n/a female PNET 3.58 Chemo/XRT M0 Alive 22
PNET279 n/a female PNET 4.17 None M0 Dead 0.17
PNET281 3 female PNET 11.75 Chemo/XRT M+ Dead 4.83
PNET282 3 female PNET 7.08 Chemo/XRT M0 Alive 9.33
65
PNET283 3 male PNET 11.83 None M0 Dead 0
PNET284 3 male PNET 4.92 Chemo/XRT M+ Dead 0.75
PNET285 n/a female PNET 2 Chemo M0 Dead 0.25
PNET286 1 male PNET 5.17 Chemo M0 Dead 3.17
PNET287 n/a male PNET 0.5 Chemo M0 n/a 0
PNET288 1 male PNET 2.25 None M0 Dead 3.17
PNET289 n/a male PNET 7 Chemo/XRT M+ Dead 1.67
PNET290 3 male PNET 0.17 n/a n/a Dead 0
PNET291 1 female PNET 3.75 Chemo/XRT n/a Dead 1
PNET293 2 female PNET 1.75 n/a n/a n/a n/a
PNET294 3 female PNET 10 n/a n/a n/a n/a
PNET295 2 male PNET 5.33 n/a n/a n/a n/a
PNET296 n/a female PNET 9.17 n/a n/a n/a n/a
PNET297 3 female PNET 15.33 Chemo/XRT n/a Dead 2.08
PNET300 3 female PNET 5.67 n/a n/a n/a n/a
PNET301 1 female PNET 3.25 Chemo M0 Alive 13.75
PNET302 n/a female PNET 8.75 Chemo/XRT M0 Dead 1.75 n/a = not available
Chemo = Chemotherapy
XRT = Radation therapy
PNET/ME: PNET with features of medulloepithelioma
PNET/EP: PNET with features of ependymal or ependymoblastic differentiation
PNET/ETNATR: PNET variant with excess neuropil and true rosettes
65
TableA‐1.2Analysesperformedontumoursamples
Annotation Illumina HT-12 v4 OmniQuad 2.5 SNP Immunohistochemistry
PNET3 √ √ √
PNET4 √ √ √
PNET5 √ √ √
PNET6 √ √ √
PNET7 √ √ no material
PNET8 no material √ √
PNET9 √ √ √
PNET10 no material √ no material
PNET11 no material √ no material
PNET15 √ √ no material
PNET17 √ √ no material
PNET20 √ √ √
PNET22 √ √ √
PNET25 √ √ √
PNET28 no material √ √
PNET30 √ √ √
PNET31 √ √ no material
PNET36 √ √ no material
PNET37 √ √ no material
PNET40 √ √ no material
PNET42 √ √ no material
PNET43 √ √ no material
PNET44 √ √ no material
PNET47 √ √ no material
PNET48 √ √ no material
PNET49 no material √ no material
PNET51 no material √ no material
PNET52 no material √ no material
PNET53 no material √ no material
PNET54 no material √ no material
PNET55 no material √ no material
PNET56 no material no material √
PNET59 no material no material √
PNET61 no material no material √
PNET62 no material no material √
PNET64 no material no material √
PNET65 no material no material √
PNET66 no material no material √
PNET67 no material no material √
PNET68 no material no material √
66
PNET71 no material no material √
PNET72 no material no material √
PNET75 no material no material √
PNET78 no material no material √
PNET79 no material no material √
PNET81 no material no material √
PNET82 no material no material √
PNET91 √ √ no material
PNET93 √ √ no material
PNET94 √ √ no material
PNET95 √ √ no material
PNET96 √ √ no material
PNET97 √ √ no material
PNET99 no material √ no material
PNET100 no material √ no material
PNET101 no material √ no material
PNET105 no material √ no material
PNET106 no material √ no material
PNET109 √ √ √
PNET111 √ √ √
PNET112 no material no material √
PNET113 no material no material √
PNET114 √ √ no material
PNET116 no material no material √
PNET118 no material no material √
PNET119 no material no material √
PNET122 no material no material √
PNET123 no material no material √
PNET126 no material no material √
PNET129 no material no material √
PNET131 no material no material √
PNET132 √ no material √
PNET135 √ √ no material
PNET138 √ √ √
PNET139 no material √ no material
PNET140 no material √ no material
PNET141 no material √ no material
PNET142 no material √ no material
PNET143 √ √ no material
PNET146 √ √ no material
PNET148 no material √ no material
PNET149 √ √ no material
PNET157 √ √ no material
67
PNET158 √ √ √
PNET160 no material no material √
PNET161 no material no material √
PNET163 no material no material √
PNET164 no material no material √
PNET166 no material no material √
PNET167 no material no material √
PNET169 no material no material √
PNET170 √ √ no material
PNET171 √ √ √
PNET172 √ √ √
PNET173 √ √ √
PNET174 √ √ √
PNET187 √ √ √
PNET188 √ √ no material
PNET190 √ √ √
PNET191 √ √ √
PNET196 √ √ √
PNET197 √ √ √
PNET199 √ √ no material
PNET200 √ √ no material
PNET226 no material √ no material
PNET255 no material no material √
PNET256 no material no material √
PNET258 no material no material √
PNET259 no material no material √
PNET260 no material no material √
PNET265 no material no material √
PNET266 no material √ √
PNET267 no material no material √
PNET268 no material no material √
PNET269 no material no material √
PNET270 no material no material √
PNET271 no material √ √
PNET272 no material √ √
PNET273 no material no material √
PNET274 no material no material √
PNET276 no material no material √
PNET277 no material no material √
PNET279 no material no material √
PNET281 no material no material √
PNET282 no material no material √
PNET283 no material √ √
68
PNET284 no material √ √
PNET285 no material √ √
PNET286 no material no material √
PNET287 no material no material √
PNET288 no material no material √
PNET289 no material no material √
PNET290 no material no material √
PNET291 no material no material √
PNET293 no material no material √
PNET294 no material no material √
PNET295 no material no material √
PNET296 no material no material √
PNET297 no material no material √
PNET300 no material no material √
PNET301 no material no material √
PNET302 no material no material √
n/a = not available
69
Appendix 2
838 www.thelancet.com/oncology Vol 13 August 2012
Articles
IntroductionBrain tumours are the most common paediatric solid neoplasms1 and a leading cause of cancer-related morbid-ity and mortality in children.2 Embryonal tumours are the largest group of malignant paediatric brain tumours and include medulloblastoma, atypical rhabdoid teratoid tumour, and CNS primitive neuro-ectodermal brain tumours (PNETs). Despite histological resemblance to medullo blastoma, patients with CNS PNETs fare poorly even with intensifi ed therapy designed for patients with metastatic medulloblastoma.3,4 By contrast with this disease, in which substantial progress has been made in molecular understanding5,6 and clinical outcomes,7 the molecular and cellular make-up of CNS PNET is largely unknown8 and tumour treatments are often ineff ective. To improve outcomes from CNS PNET, delineation of the cellular and molecular pathogenesis of CNS PNET will be important to inform diagnosis, prognosis, and design of tumour-specifi c treatments.
CNS PNETs are predominantly hemispheric tumours and make up about 3–5% of all paediatric brain tumours. Such cancers are histologically heterogeneous with variable neuronal, ependymal, or glial diff erentiation9 and can be challenging to diagnose by routine histopath ology.10 Although diagnostic techniques and molecular-based tumour classifi cations have improved for atypical rhabdoid teratoid tumours11 and medulloblastoma, the working classifi cation for CNS PNET is not settled and thus therapeutic and molecular studies can be chal-lenging. In recent studies, our research groups identifi ed a distinctly aggressive subgroup of CNS PNETs that showed fre quent amplifi cation of an oncogenic miRNA cluster (C19MC).12,13 However, the molecular composition of most CNS PNETs is unknown. Although genomic studies suggest sub stantial heterogeneity in DNA copy number profi les,8,12,14 the signifi cance of these fi ndings in relation to clinical phenotypes is unclear. Similarly, gene-expression studies of small cohorts12,15 have yielded few
Markers of survival and metastatic potential in childhood CNS primitive neuro-ectodermal brain tumours: an integrative genomic analysisDaniel Picard*, Suzanne Miller*, Cynthia E Hawkins, Eric Bouff et, Hazel A Rogers, Tiff any S Y Chan, Seung-Ki Kim, Young-Shin Ra, Jason Fangusaro, Andrey Korshunov, Helen Toledano, Hideo Nakamura, James T Hayden, Jennifer Chan, Lucie Lafay-Cousin, Pingzhao Hu, Xing Fan, Karin M Muraszko, Scott L Pomeroy, Ching C Lau, Ho-Keung Ng, Chris Jones, Timothy Van Meter, Steven C Cliff ord, Charles Eberhart, Amar Gajjar, Stefan M Pfi ster, Richard G Grundy†, Annie Huang†
SummaryBackground Childhood CNS primitive neuro-ectodermal brain tumours (PNETs) are very aggressive brain tumours for which the molecular features and best treatment approaches are unknown. We assessed a large cohort of these rare tumours to identify molecular markers to enhance clinical management of this disease.
Methods We obtained 142 primary hemispheric CNS PNET samples from 20 institutions in nine countries and examined transcriptional profi les for a subset of 51 samples and copy number profi les for a subset of 77 samples. We used clustering, gene, and pathway enrichment analyses to identify tumour subgroups and group-specifi c molecular markers, and applied immuno histochemical and gene-expression analyses to validate and assess the clinical signifi cance of the subgroup markers.
Findings We identifi ed three molecular subgroups of CNS PNETs that were distinguished by primitive neural (group 1), oligoneural (group 2), and mesenchymal lineage (group 3) gene-expression signatures with diff erential expression of cell-lineage markers LIN28 and OLIG2. Patients with group 1 tumours were most often female (male:female ratio 0·61 for group 1 vs 1·25 for group 2 and 1·63 for group 3; p=0·043 [group 1 vs groups 2 and 3]), youngest (median age at diagnosis 2·9 years [95% CI 2·4–5·2] for group 1 vs 7·9 years [6·0–9·7] for group 2 and 5·9 years [4·9–7·8] for group 3; p=0·005), and had poorest survival (median survival 0·8 years [95% CI 0·5–1·2] in group 1, 1·8 years [1·4–2·3] in group 2 and 4·3 years [0·8–7·8] in group 3; p=0·019). Patients with group 3 tumours had the highest incidence of metastases at diagnosis (no distant metastasis:metastasis ratio 0·90 for group 3 vs 2·80 for group 1 and 5·67 for group 2; p=0·037).
Interpretation LIN28 and OLIG2 are promising diagnostic and prognostic molecular markers for CNS PNET that warrant further assessment in prospective clinical trials.
Funding Canadian Institute of Health Research, Brainchild/SickKids Foundation, and the Samantha Dickson Brain Tumour Trust.
Lancet Oncol 2012; 13: 838–48
Published OnlineJune 11, 2012
http://dx.doi.org/10.1016/S1470-2045(12)70257-7
See Comment page 753
*Authors contributed equally
†Joint lead authors
Division of Hematology-Oncology, Arthur and Sonia
Labatt Brain Tumour Research Centre, Department of
Pediatrics (D Picard BSc, Prof E Bouff et MD,
T S Y Chan BSc, A Huang MD), Department of Pathology (C E Hawkins MD), and The
Centre for Applied Genomics (P Hu PhD), Hospital for Sick
Children, University of Toronto, Toronto, ON, Canada;
Children’s Brain Tumour Research Centre, Queen’s
Medical Centre, University of Nottingham, Nottingham, UK
(S Miller PhD, H A Rogers PhD, Prof R G Grundy MD);
Department of Neurosurgery, Seoul National University Children’s Hospital, Seoul, South Korea (S-K Kim MD);
Department of Neurosurgery, Asan Medical Center, Seoul,
South Korea (Prof Y-S Ra MD); Division of Pediatric
Hematology/Oncology and Stem Cell Transplantation,
Children’s Memorial Hospital, Chicago, IL, USA
(J Fangusaro MD); Clinical Cooperation Unit
Neuropathology, German Cancer Research Center,
Heidelberg, Germany (A Korshunov MD); Oncology
Department, Schneider Hospital, Petach Tikva, Israel
(H Toledano MD); Department of Neurosurgery, Kumamoto
University, Kumamoto, Japan (H Nakamura MD); Northern
Institute for Cancer Research, Newcastle University,
Newcastle Upon Tyne, UK
Articles
www.thelancet.com/oncology Vol 13 August 2012 839
(J T Hayden MD, Prof S C Cliff ord PhD); Department of Pathology and Laboratory Medicine, University of Calgary, Calgary, AB, Canada (J Chan MD); Department of Pediatric Oncology, Alberta Children’s Hospital, Calgary, AB, Canada (L Lafay-Cousin MD); Department of Neurosurgery, University of Michigan Medical School, Ann Arbor, MI, USA (X Fan MD, Prof K M Muraszko MD); Department of Neurology, Children’s Hospital Boston, Boston, MA, USA (Prof S L Pomeroy MD); Texas Children’s Cancer Center, Baylor College of Medicine, Houston, TX, USA (C C Lau MD); Department of Anatomical and Cellular Pathology, Chinese University of Hong Kong, Hong Kong, China (Prof H-K Ng MD); Department of Paediatric Molecular Pathology, Institute of Cancer Research, Sutton, UK (C Jones PhD); Department of Pediatrics, Virginia Commonwealth University, Richmond, VA, USA (T Van Meter PhD); Division of Pathology, Johns Hopkins University School of Medicine, Baltimore, MD, USA (Prof C Eberhart MD); Neuro-oncology Division, St Jude Children’s Research Hospital, Memphis, TN, USA (A Gajjar MD); and German Cancer Research Centre, and Paediatric, Haematology and Oncology, Heidelberg University Hospital, Heidelberg, Germany (S M Pfi ster MD)
Correspondence to:Dr Annie Huang, Division of Hematology-Oncology, Arthur and Sonia Labatt Brain Tumour Research Centre, Department of Pediatrics, Hospital for Sick Children, University of Toronto, Toronto, ON, Canada M5G 1X8annie.huang@sickkids.ca
See Online for appendix
insights into the clinical diversity of CNS PNETs. In this study, we undertook a multicentre, international col-laboration with the aim of providing a concerted molecular analysis of a substantial number of primary CNS PNETs. To assess clinical signifi cance of potential CNS PNET molecular subgroups, we examined whether subgroups diff ered in patient characteristics and outcome.
MethodsParticipants and study designWe obtained 254 samples with an institutional diagnosis of CNS PNET from participating institutions including six registered Children’s Cancer and Leukaemia Group centres in the UK and the Cooperative Human Tissue Network in Columbus, OH, USA (centres listed in the appendix).
CNS PNET tissue microarrays used in this study were constructed at the Hospital for Sick Children (Toronto, ON, Canada),12 University of Nottingham (Nottingham, UK),14 and the Institute of Cancer Research (Sutton, UK). All collected samples were initially reviewed for age of patient, location, and primary tumour occurrence
(fi gure 1) and then subject to histopathologic review by CEH, who was masked to fi ndings of previous assessments at the other centres. Samples were tested for loss of INI1 immunoreactivity or changes in INI1 by sequencing or multiplex ligation-dependent probe amplifi cation ana lyses to rule out misdiagnosed atypical rhabdoid teratoid tumours. We included only hemispheric tumours diag nosed as CNS PNET according to the 2007 WHO CNS tumour classifi cation criteria9 without mutations in INI1. For correlative analyses with clinical characteristics, we included only tumours with complete clinical information (full details of the patients and tumour information is listed in the appendix). We obtained tumour samples and clinical information with consent as per protocols approved by the hospital research ethics boards at participating institutions.
ProceduresAll tumour samples confi rmed to be CNS PNET and with snap frozen tumour material were processed for gene expression or DNA array analyses to initially establish tumour molecular subgroups. Tumour
77 included in copy-number analyses 95 included in immunohistochemical analyses
142 samples confirmed as paediatric primary hemispheric CNS PNET as per WHO CNS classification
18 no gene expression, immunohistochemistry, or clinical data
15 incomplete immunohistochemistry
59 samples with copy number and molecular subgroup information
80 included in immunohistochemical analyses
51 included in gene-expression analyses
23 duplicate samples
108 independent primary tumours with molecular grouping by gene expression and/or immunohistochemistry and complete clinical data
254 samples received
112 excluded 5 located in the posterior fossa 5 patients aged >18 years 11 duplicate samples 25 recurrent CNS-PNET 29 atypical teratoid rhabdoid tumours 15 ependymoma 2 glioblastoma 11 pineoblastoma 9 non-CNS tumours
Figure 1: Sample analysisPNET=primitive neuro-ectodermal brain tumour.
Articles
840 www.thelancet.com/oncology Vol 13 August 2012
0 5 10 15 20
Axonal guidance
WNT
SHH
0 5 10 15 20
Axonal guidance
0 5 10 15
Axonal guidance
PTEN signaling
TGFβ signaling
Group 1 Group 2 Group 3
*
*
**
*
*
*
**
**
**
**
**
**
**
**
Pluripotency
Pluripotency Pluripotency
SHH
WNT
WNT
Gene p value
NKX6-2 7 <0·0001
HOXA2 5 0·0008
HOXB3 3 <0·0001
HOXB4 4 <0·0001
HOXB5 6 <0·0001
HOXC4 5 <0·0001
SALL4 13 <0·0001
SOX3 13 <0·0001
PROM1 3 <0·0001
<0·0001
OMG 7 <0·0001
BCAN 7 <0·0001
NCAM2 5 <0·0001
COL1A2 4 0·0034
COL5A1 6 <0·0001
COL21A1 2 0·0008
LHX2 10 <0·0001
ZIC2 11 <0·0001
MSX1 7 <0·0001
SNAI2 3 0·0012
TWIST1 3 0·202
SHISA2 4 0·0040
MEIS1/2 5 <0·0001
OLIG1/2 15
SOX8/10 6 0·0130
LIN28/B 18 <0·0001
Fold change
Semaphorin signaling
Prim
itive
-neu
ral
Olig
oneu
ral
Mes
ench
ymal
Number of genes Number of genes
Number of genes
CRABP1 31 <0·0001
FOXJ1 2 0·023
Group 1 Group 2 Group 3
A
B Pathway analysis
Upregulated Downregulated
*p<0·05 **p<0·001
Figure 2: Molecular subgroups of CNS primitive
neuro-ectodermal brain tumours
(A) Heat map of highly expressed cell lineage genes in
each subgroup identifi ed using a supervised t test adjusted for
multiple testing (false discovery rate ≤0·05) relative
to a hierarchical cluster map of all tumours; magnitude (fold
change) and signifi cance (p value) of cell lineage genes
upregulated in each tumour subgroup is shown.
(B) Signifi cant changes in canonical pathways are
presented from analyses of 343 genes in group 1, 276 genes in group 2,
and 325 genes in group 3.
Articles
www.thelancet.com/oncology Vol 13 August 2012 841
grouping for samples with only formalin-fi xed paraffi n-embedded (FFPE) materials available were determined with immunohistochemical analyses. All tumour samples with established molecular grouping in for-mation and clinical data for patients’ demographics, metastatic status, and survival were then examined to determine whether molecular subgroups correlated with specifi c CNS PNET phenotypes.
To assess gene expression and DNA copy number profi les, we extracted RNA from 51 primary CNS PNET samples and DNA from 77 samples with standard methods. We used Illumina Omni 2.5M single-nucleo tide polymorphism (SNP) for ultra-high resolution copy number analyses (interrogating 2·5 million SNPs) and Illumina HT-12.v4 gene-expression arrays (San Diego, CA, USA) to generate DNA copy number and gene expression profi les. We did DNA and RNA hybridisations at the Centre of Applied Genomics Facility at the Hospital for Sick Children, according to the manufacturer’s protocol.
For the gene expression profi les, we did multiple unsupervised analyses to identify molecular subgroups of CNS PNETs. To defi ne genes or pathways that charac-terise each CNS PNET subgroup, we then did supervised analyses of each subgroup relative to the others, and examined the most highly diff erentially expressed gene sets between subgroups for gene and pathway enrich-ment. To assess the clinical signifi cance of identifi ed molecular subgroups, we sought markers of each sub-group that could be examined by immunohistochem istry on a larger cohort of clinically well characterised tumours. We did quantitative RT-PCR analyses to validate group-specifi c gene clusters identifi ed by supervised analyses and examined expression levels of individual genes across groups to identify the most robust, upregulated loci that can distinguish tumour subgroups.
To determine the relationship of copy number changes to molecular subgroups, we included 59 (77%) of 77 tumours with copy number profi les that had established molecular grouping.
The appendix shows details of molecular analyses done on individual tumour samples. All data are deposited in the Wellcome Trust, European Genome-Phenome Archive (accession number EGAS00000000116).
For gene-specifi c quantitative RT-PCR validation of array data, we amplifi ed 10 ng cDNA synthesised from 1 μg of RNA (TaqMan Reverse Transcription Kit, Applied Biosystems, Burlington, ON, Canada) by use of specifi c TaqMan probes-primer sets (see appendix) and deter-mined mRNA expression levels relative to actin with the ΔCt method. We did all RT-PCR assays in triplicate. Immunohistochemical analyses of tumour tissue micro-array or FFPE tumour slides were done by the Pathology Research Program laboratory at the University Health Network (Toronto, ON, Canada). We treated all tissue sections with heat-induced epitope retrieval and blocked for endogenous peroxidase and biotin. We assessed expression of markers for primitive neural, glial (nestin,
glial fi brillary acidic protein [GFAP]) or neuronal (synaptophysin) diff erentiation—which are measure-ments conventionally used in histopathological diagnosis of CNS PNET—for all tumours. The anti bodies used in this study were anti-LIN28 (Cell Signalling Technology, Boston, MA, USA), OLIG2 (Immuno-Biological Laboratories, Minneapolis, MN, USA), GFAP (DAKO, Burlington, CA, USA), and synaptophysin (Millipore, MA, USA). Antibody reactions were visual ised with a Biogenix detection kit (BioGenex Laboratories, San Ramon, CA, USA). Immunoreactivity for LIN28, GFAP, and synaptophysin were scored manually on the basis of intensity (1 was low, 2 was moderate, and 3 was high) and distribution of stains (1 was ≤10%, 2 was 10–50%, and 3 was >50%). OLIG2 immunostains were quantifi ed with the Aperio Scanscope (Aperio, Vista, CA, USA) system and the ImageScope software nuclear immunohistochemistry
Hae
mat
oxyl
in a
nd e
osin
LIN
28O
LIG2
Group 3LIN28-negative/OLIG2-negative
Group 2OLIG2-positive
Group 1LIN28-positive
100 μm200 μm
100 μm200 μm
100 μm200 μm 100 μm200 μm 100 μm200 μm
100 μm200 μm100 μm200 μm
100 μm200 μm 100 μm200 μm
LIN28 OLIG2p=0·044 p=0·0002
0
5
10
15
0
20
20
30
40
50
mRN
A re
lativ
e to
actin
(10–2
) IGF2p=0·0021
0
1
2
3
4
5
mRN
A re
lativ
e to
actin
mRN
A re
lativ
e to
actin
(10–2
)
B
A
Group 2Group 1 Group 3Group 2Group 1 Group 3Group 2Group 1 Group 3
Figure 3: Cell lineage markers of molecular subgroups of CNS PNETs(A) Quantitative RT-PCR analyses of 51 primary CNS PNETs profi led by gene expression arrays (appendix) showing enriched mean expression levels of LIN28 (group 1), OLIG2 (group 2), and IGF2 (group 3; three replicas) are shown with standard errors of mean (bars) and transcript levels shown as circles, squares, and triangles. (B) Characteristic immunohistochemical analyses from the validation of 72 samples of CNS PNET; LIN28 and OLIG2 immunostains (20× magnifi cation) are shown in relation to a haematoxylin and eosin stain; insets (1× magnifi cation) show corresponding tissue microarray cores. PNET=primitive neuro-ectodermal brain tumour.
For the European Genome-Phenome Archive see https://www.ebi.ac.uk/ega/
Articles
842 www.thelancet.com/oncology Vol 13 August 2012
algorithm. For tumours on tissue microarray, we established immunohistochemistry values on the basis of average staining score of at least two tissue cores, while tumours with FFPE slides were scored on the basis of the extent of staining in relation to the entire tumour section. Normal testicular tissue (human and mouse) was used as a positive control for LIN28 and oligodendroglioma tumour tissue was used as a positive control for OLIG2 immunostains; samples processed in parallel without primary antibodies were used as negative controls. DP and TC scored all immuno histochemistry stains while masked to cancer status, which were reviewed by AH and CEH. FISH was done on FFPE tissue microarrays or individual slides with established protocols. To confi rm robustness of LIN28 and OLIG2 immunohistochemistry for subgrouping, we tested an initial cohort of 22 samples with subgroups established by gene expression studies for LIN28 and OLIG2 expression by immunohistochemistry (appendix) We used MYCN (2p24) and p16 (9p21) specifi c PlatinumBright550 probe with corresponding LAF (2q11) and 9q21 PlatinumBright495 control probes (Kreatech, Stretton Scientifi c, Stretton, UK).
Statistical analysisWe classifi ed CNS PNETs into molecular subgroups by unsupervised hierarchical clustering, non-negative matrix factorisation,16 and principal component analyses of genes with the highest coeffi cient of variation with the Partek Genomics Suite version 6.5 (Partek, St Louis, MO, USA). We assessed genes enriched within tumour sub groups with a supervised t test adjusted for multiple hypotheses testing with the false-discovery-rate method. Ingenuity pathway analyses were done on supervised gene sets to identify canonical signalling pathways in each tumour subgroup. To establish regions of copy number gains and losses, inferred copy number data were generated with the Illumina Genome studio software and were imported into Partek for copy number variation partitioning-segmentation analyses with a SNP window of 150. We then determined signifi cance of copy number alterations in tumour subgroups with Fisher’s exact test. We used the log-rank analysis with the Kaplan-Meier method to compare survival times and χ² analyses to compare the proportion of survivors across tumour subgroups, whereas ANOVA was used to assess signifi cance of tumour subgroups in relation to age. To analyse the signifi cance of molecular subgroups in relation to sex and metastatic status at diagnosis, we compared features in an individual molecular subgroup to a pooled cohort of the other two molecular subgroups with Fisher’s exact test. Adjustment for multiple testing was not done because patients with complete infor-mation available for every clinical parameter varied. A p value of less than 0·05 was regarded as signifi cant for all analyses. All statistical analyses were done with SPSS version 19·0.
Role of the funding sourceThe sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. DP, SM, RGG, and AH had access to the raw data. The corresponding author had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication.
ResultsUnsupervised hierarchical and non-nega tive matrix factorisation clustering with 200–1000 genes consistently identifi ed three distinct molecular sub groups of CNS PNET, with non-negative matrix factorisation analyses suggesting a strongest cophenetic coeffi cient at k=3 (fi gure 2, appendix). Principle com ponent analyses suggested that group 1 tumours, which have frequent C19MC locus amplifi cation, segregated distinctly, whereas group 2 and 3 tumours showed greater proximity and some overlap (appendix).
Supervised analyses revealed that the three subgroups showed signifi cant diff erences in neural lineage and diff erentiation genes (fi gure 2). Expression profi les of group 1 were most signifi cantly enriched for genes
Subgroup 1 (n=29)
Subgroup 2 (n=36)
Subgroup 3 (n=43)
p value Comparison
Sex
n 29 36 42
Male 11 20 26
Female 18 16 16
Ratio 0·61 1·25 1·63 0·043* Group 1 vs groups 2 and 3
Age at diagnosis
n 26 32 42
Median, years 2·9 7·9 5·9 0·005† Groups 1 vs 2 vs 3
95% CI 2·4–5·2 6·0–9·7 4·9–7·8
≤4 years 20 9 18
>4 years 6 23 24
Ratio 3·33 0·39 0·75 0·001‡ Groups 1 vs 2 vs 3
Metastasis status
n 19 20 19
M0 14 17 9
M+ 5 3 10
Ratio 2·80 5·67 0·90 0·037* Group 3 vs groups 1 and 2
Status
n 26 26 34
Dead 21 20 20
Alive 5 6 14
Ratio 4·20 3·33 1·43 0·13‡ Groups 1 vs 2 vs 3
Survival time
n 20 23 23
Median, years 0·8 1·8 4·3 0·019§ Groups 1 vs 2 vs 3
95% CI 0·5–1·2 1·4–2·3 0·8–7·8
Some patients were not included in analyses because of a lack of specifi c clinical data; details of all patients are shown in the appendix. *Fisher’s exact test. †ANOVA. ‡Pearson’s χ². §Log-rank (Mantel-Cox) test.
Table 1: Clinical and molecular characteristics of children with CNS primitive neuro-ectodermal brain tumours, by molecular subgroup
Articles
www.thelancet.com/oncology Vol 13 August 2012 843
associated with embryonic or neural stem cells. Notably, LIN28 and CRABP1,17 which are implicated in stem-cell pluripotency, were among the most overexpressed genes with nearly 20–30 fold greater expression in group 1 as compared with group 2 and group 3. In group 2 tumours, OLIG1/2, SOX8/10, and BCAN (which are markers of oligoneural diff erentiation18) were the most upregulated genes, whereas group 3 tumours showed reduced ex pression of neural diff erentiation genes but upregulation of epithelial and mesenchymal diff erentiation genes including COL1A2, COL5A, FOXJ1,19 and MSX1.20
Pathway enrichment analyses also suggested signifi -cant diff erences in the signalling gene profi les of every tumour subgroup (fi gure 2). Consistent with diff erential enrichment of lineage related genes in tumour sub groups, we noted signifi cant diff erences in expression of axonal guidance genes among the CNS PNET subgroups. Genes involved in WNT signalling were upregulated in group 1 tumours and those involved in SHH signalling were downregulated in group 2 tumours, whereas TGF-β and PTEN signalling pathway genes were specifi cally upregulated in group 3 tumours.
Immunohistochemistry and quantitative RT-PCR analyses showed that LIN28, OLIG2, and IGF2 were highly diff erentially expressed in CNS PNET group 1, group 2, and group 3 tumours, respectively (fi gure 3, appendix). IGF2 protein expression could not be reliably scored on tumour samples (appendix); however, immuno-histochemical analyses for LIN28 and OLIG2 were robust and correlated with gene-expression levels as established by arrays and quantitative RT-PCR analyses. Immunohistochemical analyses on a test cohort of 22 tumours indicated that extent of cytoplasmic LIN28 and nuclear OLIG2 immunostaining also correlated with tumour subgroup assignment based on gene-expression profi les (appendix). LIN28 was expressed at high levels and OLIG2 was expressed at low levels in group 1 tumours, whereas group 2 tumours had high OLIG2 and little LIN28 immunopositivity. LIN28 and OLIG2 protein expression was low or absent in group 3 tumours (fi gure 3).
In the LIN28 and OLIG2 immunohistochemical analyses of an additional 72 primary CNS PNETs with only FFPE samples available for analyses, 15 tumours had inconclusive immuno histochemical analyses (appendix). Overall, we were able to assign 108 of 142 primary CNS PNETs to molecular subgroups on the basis of gene expression or immunohistochemical analyses of LIN28 or OLIG2 protein expression (or both measures; tables 1, 2, appendix). We classifi ed 29 (27%) tumours as group 1, 36 (33%) as group 2, and 43 (40%) as group 3 (appendix). Group 1 tumours with high LIN28 expression generally also expressed high levels of nestin, but had little to no expression of GFAP. GFAP and synaptophysin expression varied substantially between each of the tumour groups and did not consistently correlate with LIN28 or OLIG2 expression. Notably, quantitative RT-PCR
and expression analyses suggested that expression of other neuronal diff erentiation genes also do not diff er signifi cantly among the molecular subgroups of CNS PNETs (appendix). These fi ndings collectively suggest the limitations of conven tional markers to capture the molecular diversity of CNS PNETs.
CNS PNET subgroups have distinct DNA copy number patterns. Apart from the C19MC miRNA amplicon that we previously identifi ed,12 we noted few other recurrent high level copy number gains or amplifi cation. Focal MYCN and CDK4 amplifi cation was detected in isolated tumours. Deletions centred on CDKN2A/2B were the most frequent copy number aberration noted (10 [12%] of 77 tumours; appendix). To establish whether there were characteristic copy number aberrations within the CNS PNET subgroups, we analysed the copy number patterns
Subgroup 1 Subgroup 2 Subgroup 3 p value Comparison
Age ≤4 years
n 20 9 18
Metastasis status
n 15 6 4
M0 11 5 4
M+ 4 1 0
Ratio 2·75 5·00 4·00 0·48* Groups 1 vs 2 vs 3
Status
n 20 7 13
Dead 16 6 10
Alive 4 1 3
Ratio 4·00 6·00 3·33 0·90* Groups 1 vs 2 vs 3
Survival time
n 15 6 6
Median survival, years 1·0 0·8 2·7 0·70† Groups 1 vs 2 vs 3
95% CI 0·7–1·3 0–4·8 1·9–3·5
Age >4 years
n 6 23 24
Metastasis status
n 4 14 15
M0 3 12 5
M+ 1 2 10
Ratio (analysis 1) 3·00 6·00 0·50 0·033* Group 3 vs groups 1 and 2
Ratio (analysis 2) ·· 6·00 0·50 0·014‡ Group 2 vs group 3
Status 6 19 21
Dead 5 14 10
Alive 1 5 11
Ratio (analysis 1) 5·00 2·80 0·91 0·13*
Ratio (analysis 2) ·· 2·80 0·91 0·087‡ Group 2 vs group 3
Survival time
n 5 17 17
Median survival, years 0·5 1·8 4·8 0·004 Groups 1 vs 2 vs 3
95% CI 0·0–1·0 1·5–2·2 1·6–8·0
Some patients were not included in analyses because of a lack of specifi c clinical data; details of all patients are shown in the appendix. *Pearson χ². †Log-rank (Mantel-Cox) test. ‡Fisher’s exact test.
Table 2: Clinical and molecular characteristics of children with CNS primitive neuro-ectodermal brain tumours, by molecular subgroup and age
Articles
844 www.thelancet.com/oncology Vol 13 August 2012
of a subset of 59 tumours that could be subgrouped on the basis of LIN28 and OLIG2 gene or protein expression (appendix). Copy number analyses showed that, in addition to chr19q13·41 amplifi cation and chromosome 2 gains, group 1 tumours had frequent gains of chromo-some 3. Group 2 tumour copy number aberration profi les were characterised by more frequent gains of chromosome 8p (p=0·027), 13 (p=0·009), and 20 (p=0·039) compared with group 1 and 2 tumours. Notably, only group 2 and 3 tumours had frequent chromosome 9p loss centred on the CDKN2A/2B locus. Moreover, group 3 tumours showed frequent loss of chromosome 14 (p=0·009). Thus, CNS PNET subgroups correlate with distinct gene expression as well as genomic profi les.
To determine the clinical signifi cance of CNS PNET molecular subgroups, we examined whether subgroups diff ered in patient characteristics and outcome. Of 108 patients for which tumour subgrouping could be established, demographic data on sex, age, survival, and tumour stage were available for 107, 100, 58 and 66 cases, respectively (tables 1 and 2; appendix). Molecular subgroups were associated with distinct clinical pheno-types. Sex and age distribution diff ered between the three molecular CNS PNET subgroups. Group 1 tumours were
more often noted in female patients than were group 2 and group 3 tumours (fi gure 4, table 1). Patients with group 1 and group 2 tumours had bimodal age distributions with peak incidence at opposite age spectra, whereas patients with group 3 tumours had a single peak between 4–8 years. Patients with group 1 tumours were younger than were those with group 2 or group 3 tumours (table 1). 47 (47%) of 100 patients with data for age were 4 years old or younger; however young patients were signifi cantly over-represented in group 1 as compared with group 2 and group 3 (p=0·001; fi gure 4, table 1).
Molecular subgroups of CNS PNET also had signifi cant diff erences in incidence of tumour metastases. Patients with group 3 tumours had the highest incidence of disseminated disease at diagnosis (fi gure 4, table 1). Although metastatic disease is reportedly more frequently in younger children with embryonal brain tumours, analyses done with stratifi cation for age (≤4 years vs >4 years) showed the incidence of tumour metastases diff ered signifi cantly in CNS PNET subgroups diagnosed in older children (fi gure 4, table 2). More patients aged older than 4 years at diagnosis in group 3 presented with metastatic disease than did those in group 1 or group 2 (table 2). The proportion of
0
10
20
30
40
50
60
70
80
0–2 2–4 4–6 6–10 >10Pa
tient
s (%
)
Age range (years)
0
20
40
60
80
100
1 2 3
9
23
18
24
20
6
0
20
40
60
80
100
16
20
16
26
18
11
1 2 3
Patie
nts (
%)
Tumour subgroup Tumour subgroup
0
20
40
60
80
100
3
17
9
10
5
14
Patie
nts (
%)
1 2 3
4 1 1 2 9
11 5 4 3 12 6
1 2 3 1 2 3Tumour subgroup Tumour subgroup Tumour subgroup
Patie
nts (
%)
0
20
40
60
80
100
Patie
nts (
%)
0
20
40
60
80
100
Patie
nts (
%)
BA
Group 1Group 2Group 3
Male Female p=0·04 >4 years ≤4 years p=0·001
D Metastasis and age: ≤4 yearsC Metastasis
M0 M+ p=0·037 M0 M+ p=0·48 M0 M+ p=0·033
Metastasis and age: >4 years
Figure 4: Clinical phenotypes of molecular subgroups of CNS PNETs(A) Sex-specifi c and (B) age-specifi c correlations with tumour subgroup in 108 primary CNS PNET tumours (tables 1, 2, appendix). (C) Metastatic status at diagnosis (58 patients); p value from the two-sided Fisher’s exact test (group 1 vs groups 2 and 3). (D) Metastatic status at diagnosis, stratifi ed by age (58 patients); p values from Pearson’s χ² (group 1 vs group 2 vs group 3). PNET=primitive neuro-ectodermal brain tumour.
Articles
www.thelancet.com/oncology Vol 13 August 2012 845
non-metastatic to metastatic tumours in this age group diff ered signifi cantly in comparisons of group 3 to a combined cohort of group 1 and group 2 patients and to group 2 patients alone (fi gure 4, table 2).
Log-rank analysis of all tumour age groups showed that overall survival for patients in group 1 was signifi cantly shorter than it was for patients in group 2 and group 3 (fi gure 5, table 1). With the exception of two longer term survivors, all patients in group 1 were deceased within 4·2 years of diagnosis. Because most group 1 tumours arise in younger children who are often treated hetero-geneously with radiation-sparing thera peutic approaches due to worries of neurocognitive damage,3,21 we examined whether the poor prognostic association of LIN28 expression in group 1 tumours held true for older children who are conventionally prescribed intensifi ed treatment regimens with higher dose craniospinal irradiation. Moreover, because most infant brain tumour protocols enrol patients up to 3–4 years of age,3,4,21 we stratifi ed patients by age at the cutoff of 4 years, to remove age and potential treatment biases on survival. Although overall survival for all young patients was similarly poor, children older than 4 years of age with LIN28 group 1 tumours fared signifi cantly worse (median survival of 0·5 years, 95% CI 0·0–1·0; p=0·004) than did patients older than 4 years of age in group 2 (1·8 years, 1·5–2·2) and group 3 (4·8 years, 1·6–8·0). These fi ndings suggest that immunopositivity for LIN28 identifi es a particularly high risk group of CNS PNET across ages.
DiscussionAdvances in treatment for childhood CNS PNET have been diffi cult because of the low incidence of the disease,9 incomplete understanding of the clinical and biological spectra of disease, and an absence of specifi c markers to aid histopathological diagnoses (panel).8,10 In this study, we aimed to integrate gene expression, copy number, and immuno histochemical analyses to characterise 142 primary hemispheric CNS PNETs. Diff erential expres sion of cell lineage markers, LIN28 and OLIG2, distinguishes three molecular subgroups of CNS PNET and identifi es CNS PNET subgroups at very high risk of metastases and treatment failures and distinct demographic features. Primitive neural group 1 tumours, with frequent C19MC amplifi cation and high LIN28 expression, are distinctly aggressive tumours arising in young children. Oligoneural group 2 tumours, which have high OLIG2 expression, arise in older children and are frequently localised. Mesenchymal group 3 tumours, which have low LIN28 and OLIG2 expression, are associated with a high incidence of metastases and occur at all ages. Group 1 tumours were more frequently in females, whereas group 2 and group 3 tumours arose more frequently in males. These markers are promising molecular identifi ers for childhood CNS PNET that could be applied to refi ne tumour diagnosis, classifi cation, and treatment risk stratifi cation.
0 4·2 8·3 12·5 16·7 20·8
0 4·2 8·3 12·5
0 4·2 8·3 12·5 16·6 20·8Years
Cum
ulat
ive
surv
ival
1·0
0·8
0·6
0·4
0·2
0
Cum
ulat
ive
surv
ival
1·0
0·8
0·6
0·4
0·2
0
Cum
ulat
ive
surv
ival
1·0
0·8
0·6
0·4
0·2
0
Number at riskGroup 1 20 2 1 1Group 2 23 5 1 Group 3 23 8 2 1 1 1
Number at riskGroup 1 15 1 1 1Group 2 6 2 1 1 Group 3 6 2 0
Number at riskGroup 1 5 0Group 2 17 3 1 Group 3 17 6 2 1 1 1
B Survival and age: ≤4 years
A Survival
C Survival and age: >4 years
Group 1Group 2Group 3Censored
p=0·070
p=0·004
p=0·019
Figure 5: Survival of molecular subgroups of primary CNS primitive neuro-ectodermal brain tumours(A) Overall survival (66 patients). Overall survival, stratifi ed by age 4 years or younger (B; 27 patients) and age older than 4 years (C; 39 patients). p values from the log-rank test (group 1 vs group 2 vs group 3).
Articles
846 www.thelancet.com/oncology Vol 13 August 2012
Comprehensive clinical and biological data for a large cohort of primary CNS PNETs are not available. Previous molecular studies have often included a spectrum of CNS PNETs including rare variants, tumours arising in diff erent anatomical locations such as pineoblastomas, and medulloblastoma.8 Because the biological relation of CNS PNET arising in diff erent anatomical locations is unclear, we restricted our study to tumours arising in the cerebral hemispheres, making up most childhood CNS PNET.
Although CNS PNETs are generally regarded as mainly a disease of younger children, analysis of our data suggests that more than 50% of CNS PNETs arise in older children (>4 years) and that CNS PNETs have an age-dependent distribution of molecular subgroups. The strong association of lineage-specifi c gene-expression signatures and age with specifi c tumour subgroups suggests molecular subgroups of hemispheric CNS PNET might derive from diff erent precursor cell stages or type. Specifi cally, transcriptional signatures of group 1
tumours (enriched for CD133, CRABP1, LIN28, and ASCL1) and group 2 tumours (enriched for OLIG1/2) suggest that the former arose from early neural progenitors and the latter from oligoneural progenitors. The possible cellular origin of group 3 CNS PNETs, which are enriched for mesenchymal diff erentiation genes including ZIC222 and LHX2,23 is less clear.
Studies of human brain tumours and brain tumour models suggest that cell lineage-related gene-expression signatures often correlate with and underlie clinical and biological heterogeneity in a spectrum of CNS tumours including malignant gliomas,24 ependymoma,25 and medulloblastoma.26 In addition to age and sex, we noted signifi cant diff erences in survival and metastatic tendency between the three CNS PNET subgroups. We noted poorest survival in the primitive neural group 1 tumours identifi ed on the basis of LIN28 expression, irrespective of age or metastatic status. Together with previous fi ndings that link C19MC amplifi cation with a distinctly aggressive CNS PNET phenotype,12,13 our study further emphasises CNS PNET with C19MC amplifi cation and/or LIN28 expression as a unique clinicopathological entity and suggests LIN28 immuno-histochemistry could be an important, new diagnostic tool for this distinct group of embryonal brain tumours.
Overall survival for group 2 and group 3 tumours, which more commonly presented in older children, did not diff er signifi cantly (p=0·087), but there was some suggestion that children older than 4 years with group 3 mesenchymal lineage tumours had better survival (11 [52%] of 21 children were alive at last assessment) than did those with group 2 tumours (fi ve [26%] of 19 were alive at last assessment; table 2). This fi nding is surprising, because group 3 tumours had the highest incidence of metastases at diagnosis, which is linked to poorer outcomes in medulloblastoma and other embryonal brain tumours. These observations suggest greater sensitivity of group 3 tumours to medulloblastoma-type drugs, which are usually prescribed to older children with CNS PNET, and might suggest greater biological relatedness of group 3 CNS PNETs to medulloblastoma. However, treatment designed for high-risk medullo-blastoma with higher-dose craniospinal irradiation, which is usually prescribed to older children with CNS PNET, might not off er additional therapeutic benefi t for most group 2 CNS PNETs. Our fi ndings underscore the clinical hetero geneity of CNS PNET arising in the cerebral hemisphere and suggest that group 2 and group 3 CNS PNET need diff erent therapeutic approaches tailored to their specifi c biology.
Our study confi rms the documented poor overall outcome of CNS PNETs across age groups and emphasises the need to seek new treatment strategies for this aggressive disease. Pathway enrichment analyses suggested that the non-canonical WNT pathway pre-dominates in group 1 tumours and thus could be an attractive target for treatment. Group 1 tumours also
Panel: Research in context
Systematic reviewWe searched PubMed and Google Scholar for molecular studies of childhood CNS primitive neuro-ectodermal brain tumours (PNETs) published in English between Jan 1, 1985, and Dec 31, 2011, with the search terms “childhood PNET”, “CNS-PNET”, “supratentorial PNET”, and “embryonal brain tumours” (because CNS PNETs were often included in molecular studies of medulloblastoma).
InterpretationPresent treatment strategies for CNS PNETs are largely designed based on the cancer’s close histological similarities to medulloblastoma, although such therapy is not as effi cacious in CNS PNETs. A tailored treatment for CNS PNET is needed to exploit their distinct biology; however, molecular studies of CNS PNETs have been restricted by rare disease incidence and scarcity of robust diagnostic markers. Apart from two recent studies12,14 by our group and a previous review,8 molecular studies of CNS PNETs have been restricted to small cohorts of CNS PNETs from various anatomical sites and diagnoses made on various histopathological criteria. Our integrated genomic analyses of primary CNS PNET tumour samples were restricted to the cerebral hemispheres and tumours that met the 2007 WHO CNS classifi cation criteria for CNS PNET and were confi rmed with genetic methods to be non-rhabdoid tumours. In keeping with previous clinical observations, we show that CNS PNET makes up a heterogeneous spectrum of tumours and defi nes three molecular subtypes of CNS PNETs with distinct survival and metastatic features. Our study provides the fi rst molecular prognostic markers for CNS PNET, and is a substantial advance towards biology-driven treatment strategies for this disease.
Articles
www.thelancet.com/oncology Vol 13 August 2012 847
showed signifi cant upregulation of the SHH signalling pathway, suggesting that novel SHH pathway inhibitors in clinical trials27 might be attractive new drugs for this subgroup. Of note, CRABP1, a retinoid-binding protein known to change retinoic metabolism and confer all-trans retinoic acid resistance,28 is very highly expressed in the group 1 CNS PNET. Thus retinoic acid, which is being tested in high-risk medulloblastoma and CNS PNET cooperative group clinical trials for older children, might be of little therapeutic benefi t in group 1 CNS PNETs.
By contrast with group 3 tumours that showed upregu-lation of several canonical pathways, the oligoneural group 2 tumours showed downregulation of both SHH and WNT signalling pathways. Although we noted higher expression of PDGFRA and ERBB3 in this subgroup (appendix), pathway analyses did not show signifi cant global enrichment of receptor tyrosine sig nalling pathways. However, these potential therapeutic pathways might emerge with studies of larger cohorts, and further delineation of CNS PNET subgroups. Consistent with the higher incidence of metastases noted in group 3 tumours, we identifi ed substantial activation of semaphorin signalling genes in this tumour group. In addition to activated TGFβ signalling, group 3 tumours showed upregulation of PTEN signalling and IGF2 expression, thus making these pathways and genes attractive targets for potential subgroup specifi c therapies.
Our data show that hemispheric tumours diagnosed as CNS PNET in children can be diff erentiated into subgroups with distinct survival and metastatic charac-teristics on the basis of lineage markers, LIN28 and OLIG2. Because our study was restricted to hemispheric CNS PNETs, assessment of the signifi cance of these molecular groupings to non-hemispheric CNS PNETs, such as pineoblastoma, will be important. Our study was limited by sample size and the absence of an independent validation cohort due to the rare incidence of CNS PNET. Nonetheless, we anticipate that LIN28 and OLIG2, which are the fi rst molecular markers reported for CNS PNET to date, will help identify high-risk group 1 tumours for new therapies, allow tailoring of chemoradiotherapy for patients with group 2 and group 3 tumours (which diff er strikingly in metastatic potential), and will help establish a working classifi cation of CNS PNETs. Our report under scores the importance of concerted, collaborative eff orts to study large retrospective cohorts of tumours and patients to accelerate biological and ultimately therapeutic studies of rare tumours.
ContributorsAH, DP, RGG designed the study. AH and RGG procured fi nancial
support. CH, RGG, AG, SMP, EB, AK, JC, LL-C, CE, HT, JF, S-KK, Y-SR,
TVM, CCL, SLP, H-KN, CJ, SCC, JTH, XF, KMM, HN, and AH provided
study materials or patients. DP, AH, SM, CH, and PZH analysed and
interpreted the data. DP, AH and SM wrote the report; AH, SM, SMP
and RGG reviewed and edited the fi nal report. All authors approved the
fi nal manuscript.
Confl icts of interestWe declare that we have no confl icts of interest.
AcknowledgmentsFunding was received from the Canadian Institute of Health Research
(grant number 102684) and Brainchild (AH), and the Samantha Dickson
Brain Tumour Trust, grant number 17/53 (RG). We are grateful for the
assistance of clinicians from the Children’s Cancer Leukaemia Group
centres and Biological studies committee, the neuropathological review
by Keith Robson and James Lowe, Jennifer Ward for FISH studies
(Children’s Brain Tumour Research Centre, University of Nottingham,
Nottingham, UK), statistical consultations from Derek Stephen
(Statistical Support Unit, Hospital for Sick Children, Toronto, ON,
Canada), and technical help from Jonathon Torchia (Hospital for Sick,
Children, Toronto, ON, Canada).
References1 Kohler BA, Ward E, McCarthy BJ, et al. Annual report to the nation
on the status of cancer, 1975–2007, featuring tumors of the brain and other nervous system. J Natl Cancer Inst 2011; 103: 714–36.
2 Turner CD, Rey-Casserly C, Liptak CC, et al. Late eff ects of therapy for pediatric brain tumor survivors. J Child Neurol 2009; 24: 1455–63.
3 Timmermann B, Kortmann RD, Kuhl J, et al. Role of radiotherapy in supratentorial primitive neuroectodermal tumor in young children: results of the German HIT-SKK87 and HIT-SKK92 trials. J Clin Oncol 2006; 24: 1554–60.
4 Pizer BL, Weston CL, Robinson KJ, et al. Analysis of patients with supratentorial primitive neuro-ectodermal tumours entered into the SIOP/UKCCSG PNET 3 study. Eur J Cancer 2006; 42: 1120–28.
5 Northcott PA, Korshunov A, Witt H, et al. Medulloblastoma comprises four distinct molecular variants. J Clin Oncol 2011; 29: 1408–14.
6 Cho YJ, Tsherniak A, Tamayo P, et al. Integrative genomic analysis of medulloblastoma identifi es a molecular subgroup that drives poor clinical outcome. J Clin Oncol 2011; 29: 1424–30.
7 Gajjar A, Chintagumpala M, Ashley D, et al. Risk-adapted craniospinal radiotherapy followed by high-dose chemotherapy and stem-cell rescue in children with newly diagnosed medulloblastoma (St Jude Medulloblastoma-96): long-term results from a prospective, multicentre trial. Lancet Oncol 2006; 7: 813–20.
8 Li MH, Bouff et E, Hawkins CE, et al. Molecular genetics of supratentorial primitive neuroectodermal tumors and pineoblastoma. Neurosurg Focus 2005; 19: E3.
9 Louis DN, Wiestler OD. WHO classifi cation of tumours of the central nervous system, 4th edn. Lyon, France: International Agency for Research on Cancer, 2007.
10 Burger PC. Supratentorial primitive neuroectodermal tumor (sPNET). Brain Pathol 2006; 16: 86.
11 Jackson EM, Sievert AJ, Gai X, et al. Genomic analysis using high-density single nucleotide polymorphism-based oligonucleotide arrays and multiplex ligation-dependent probe amplifi cation provides a comprehensive analysis of INI1/SMARCB1 in malignant rhabdoid tumors. Clin Cancer Res 2009; 15: 1923–30.
12 Li M, Lee KF, Lu Y, et al. Frequent amplifi cation of a chr19q13·41 microRNA polycistron in aggressive primitive neuroectodermal brain tumors. Cancer Cell 2009; 16: 533–46.
13 Korshunov A, Remke M, Gessi M, et al. Focal genomic amplifi cation at 19q13·42 comprises a powerful diagnostic marker for embryonal tumors with ependymoblastic rosettes. Acta Neuropathol 2010; 120: 253–60.
14 Miller S, Rogers HA, Lyon P, et al. Genome-wide molecular characterization of central nervous system primitive neuroectodermal tumor and pineoblastoma. Neuro Oncol 2011; 13: 866–79.
15 Rogers HA, Miller S, Lowe J, et al. An investigation of WNT pathway activation and association with survival in central nervous system primitive neuroectodermal tumours (CNS PNET). Br J Cancer 2009; 100: 1292–302.
16 Brunet JP, Tamayo P, Golub TR, et al. Metagenes and molecular pattern discovery using matrix factorization. Proc Natl Acad Sci USA 2004; 101: 4164–69.
17 Skottman H, Mikkola M, Lundin K, et al. Gene expression signatures of seven individual human embryonic stem cell lines. Stem Cells 2005; 23: 1343–56.
18 Ligon KL, Fancy SP, Franklin RJ, et al. Olig gene function in CNS development and disease. Glia 2006; 54: 1–10.
Articles
848 www.thelancet.com/oncology Vol 13 August 2012
19 Jacquet BV, Muthusamy N, Sommerville LJ, et al. Specifi cation of a Foxj1-dependent lineage in the forebrain is required for embryonic-to-postnatal transition of neurogenesis in the olfactory bulb. J Neurosci 2011; 31: 9368–82.
20 Houzelstein D, Auda-Boucher G, Cheraud Y, et al. The homeobox gene Msx1 is expressed in a subset of somites, and in muscle progenitor cells migrating into the forelimb. Development 1999; 126: 2689–701.
21 Fangusaro J, Finlay J, Sposto R, et al. Intensive chemotherapy followed by consolidative myeloablative chemotherapy with autologous hematopoietic cell rescue (AuHCR) in young children with newly diagnosed supratentorial primitive neuroectodermal tumors (sPNETs): report of the Head Start I and II experience. Pediatr Blood Cancer 2008; 50: 312–18.
22 Elms P, Siggers P, Napper D, et al. Zic2 is required for neural crest formation and hindbrain patterning during mouse development. Dev Biol 2003; 264: 391–406.
23 Hoch RV, Rubenstein JL, Pleasure S. Genes and signaling events that establish regional patterning of the mammalian forebrain. Semin Cell Dev Biol 2009; 20: 378–86.
24 Phillips HS, Kharbanda S, Chen R, et al. Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell 2006; 9: 157–73.
25 Taylor MD, Poppleton H, Fuller C, et al. Radial glia cells are candidate stem cells of ependymoma. Cancer Cell 2005; 8: 323–35.
26 Gibson P, Tong Y, Robinson G, et al. Subtypes of medulloblastoma have distinct developmental origins. Nature 2010; 468: 1095–99.
27 Rudin CM, Hann CL, Laterra J, et al. Treatment of medulloblastoma with hedgehog pathway inhibitor GDC-0449. N Engl J Med 2009; 361: 1173–78.
28 Boylan JF, Gudas LJ. The level of CRABP-I expression infl uences the amounts and types of all-trans-retinoic acid metabolites in F9 teratocarcinoma stem cells. J Biol Chem 1992; 267: 21486–91.