PERSPECTIVES

cell death signals, and they can later acquire

OPINION
additional alterations, which lead to postnatal
malignant transformation.
The prenatal origins of cancer Some types of cancer with embryonal
histology, such as germ cell tumours,
more commonly arise in early adulthood.
Glenn M. Marshall, Daniel R. Carter, Belamy B. Cheung, Tao Liu, However, it is unclear whether these malig-
Marion K. Mateos, Justin G. Meyerowitz and William A. Weiss nancies represent cellular reversion from a
mature form to an embryonic form in adult
Abstract | The concept that some childhood malignancies arise from postnatally life or postnatal persistence of embryonal
persistent embryonal cells has a long history. Recent research has strengthened cells, because there is no evidence of a ‘rest
the links between driver mutations and embryonal and early postnatal disease’ pre-dating these cancers. Therefore,
development. This evidence, coupled with much greater detail on the cell of origin although the concept of rest cells is help-
and the initial steps in embryonal cancer initiation, has identified important ful in considering the processes that might
underlie cancer development in young
therapeutic targets and provided renewed interest in strategies for the early
children, lack of evidence for a rest disease
detection and prevention of childhood cancer. before malignant transformation has been
taken as evidence that few childhood cancers
Human cancer is a multistep process For some childhood cancers, such as arise through this mechanism. However, a
that has inexorably evolved over many retinoblastoma, infant acute lymphoblastic substantial number of childhood malignan-
years towards genomic instability and leukaemia (ALL) and malignant rhabdoid cies seem to arise prenatally. If all childhood
life-threatening cellular phenotypes. The tumours, there is evidence of an embryonal malignancies with a suspected prenatal
recognition that a minority of cancer cells cell of origin. However, it is unclear whether cell of origin are grouped together, it is
within a tumour possess potent embryo- these embryonal cancers arise from embryo- apparent that almost one-half of all child-
nal features or the plasticity to revert to nal cells in utero or as a consequence of a hood cancers might have prenatal origins
embryonal features during postnatal life has single oncogenic event in a more mature (TABLE 1). In B cell‑lineage ALL (B‑ALL),
generated the cancer stem cell hypothesis1, prenatal cell with rapid progression to transient myeloproliferative disorder (TMD)
which reflects the almost limitless hetero- genomic instability; and this model of a and myeloid leukaemia–Down syndrome
geneity that is evident in most tumours. single oncogenic event in a mature prenatal (ML–DS), driver mutations and causal gene
Cancer stem cells are not replicates of their cell is likely to be the case for mixed lineage rearrangements or balanced translocations
embryonal counterparts but instead have leukaemia (MLL) fusion genes in infant ALL are detectable in perinatal peripheral blood
some features of embryonal cells that pro- and for SMARCB1 (also known as INI1) leukocytes years before clinical presentation,
vide a survival advantage that is suited to mutations in malignant rhabdoid tumours2,3. which indicates that these childhood cancers
a particular time and place during tumour A unique characteristic of certain child- are initiated in utero9,10.
evolution1. By contrast, cancers that pre- hood cancers, such as Wilms’ tumour and Recent studies using genetically altered
sent very early in a child’s life often have neuroblastoma, is a precursor hyperplasia animal models, family linkage analyses,
embryonal features but must have moved of embryonal cells that presents in infancy large-scale expression profiling and genome-
quickly to genomic instability. This rapid but that has an overwhelming tendency to wide association studies (GWAS) have
evolution might occur when a cancer cell undergo spontaneous regression and cell identified a large number of genes that are
arises directly from an embryonal cell death4,5. More than 150 years ago, pathol­ possible drivers in prenatal cancers, and
or when a mature prenatal cell acquires ogists first suggested that postnatally per- many of these genes also have a role in
embryonal properties that favour survival sistent embryonal remnant cells (‘rest cells’) embryonal development. This Opinion arti-
in the prenatal and postnatal environments. lay dormant in normal tissues but retained cle focuses on four childhood malignancies
Prenatal tumorigenic mutations might be their capacity for growth and might later with, we argue, the strongest evidence for a
more likely to occur in stem or progenitor- undergo malignant transformation6–8. Today, prenatal cell of origin: neuroblastoma, TMD
like cells that have characteristics (such as we know that during embryogenesis many and ML–DS, B‑ALL, and medulloblastoma.
unrestrained self-renewal) that are essential more cells are produced than are required We discuss their prenatal origins in the
in utero but that are fatal in postnatal life. for organogenesis. Thus, mechanisms such context of normal tissue development,
Understanding the characteristics of these as trophic factor withdrawal are required the potential cell of origin and the evidence
cells might be important in understanding for deleting cells that are in excess after for an embryonal precancer. Although
more about childhood cancer: why it devel- organogenesis is complete. In rare instances, embryonal rest cells are often seen as ‘nor-
ops, how best to treat it and, potentially, as a potential first pathological step towards mal’ in the context of the rest cell hypoth-
how to prevent it. cancer, some embryonal cells can resist these esis, we feel that this definition limits the

NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 277

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES

Table 1 | Embryonal childhood malignancies

Malignancy Tissue of origin Evidence for prenatal origin
Pre-clinical Clinical
Neuroblastoma Neural crest • Transgenic expression of genes involved in • Young age of incidence (<5 years)12
sympathoadrenal sympathoadrenal development (MYCN, ALK • Multifocal or bilateral primary tumours in tissues
progenitors and LIN28B) recapitulate neuroblastoma20,33,43 derived from the embryonic neural crest in <5% of
(neuroblasts) • Animal modelling with Th–MYCN mice shows patients11
rest formation prior to tumour formation • Normal fetal neuroblasts are highly similar to
owing to resistance to developmental tumour-derived neuroblasts45
deletion signals15,19 • Failure of complete sympathoadrenal maturation
• Neural crest stem cell allografts generate — neuroblast rests and increased levels of
neuroblastoma with oncogene (MYCN and catecholamines (a neuroblastoma marker) in
ALKF1174L) transduction44 subclinical patients46,47,172
• Low-risk neuroblastoma and rest cells frequently
regress4,46
Wilms’ tumour Primitive Not applicable • Young median age of incidence173
metanephrogenic • Wilms’ tumour can be bilateral173,174
blastema • Inactivating mutations in kidney development gene
WT1 predispose for Wilms’ tumour173
• Nephrogenic rests adjacent to tumours174
• WT1+ nephrogenic rests found adjacent to WT1+
tumours174
• Nephrogenic rests can spontaneously regress5,175
Medulloblastoma Neural progenitors • Aberrant SHH signalling through • The SHH subtype is associated with the developmental
of the developing development recapitulates SHH subgroup disorder Gorlin syndrome111
cerebellum or medulloblastoma114,121,122 • Allelic loss on chromosome 9 (including the SHH
brainstem • SHH subgroup medulloblastoma modelling pathway suppressor PTCH1) occurs in 30% of
shows premalignant embryonal cells prior tumours113
to tumour formation owing to resistance to
developmental cell deletion signals117
• Activating mutations in β‑catenin and
aberrant WNT signalling in OLIG3+ neural
precursors through brainstem development
can recapitulate medulloblastoma137
• Animal modelling shows that cerebellar
development genes (MYCN and MYC)
can recapitulate medulloblastoma in
SHH-independent medulloblastoma126–128
Retinoblastoma Retinal Inactivation of retinal development genes • Familial cases are often bilateral, multifocal and usually
progenitors of the Rb1 and p107 in retinal progenitors causes occur in the first 2 years of postnatal life143,178
developing eye prolonged proliferation of progenitors through • Retinal progenitor cell proliferation occurs only in the
development and is sufficient to recapitulate fetal retina178
retinoblastoma176,177 • Premature baby reported with retinoblastoma in early
postnatal life179
• Patients are defined by familial and sporadic
inactivation of the retinal development gene RB1
(REF. 180)
GCTs Primordial germ Not applicable • A subset of patients with early incidence (<4 years)181,182
cells, which • Teratomas can be found prenatally183
originate in yolk • Germline genetic predisposition has earlier onset181
sac endoderm • Bilateral GCTs181
• Genetic aberrations such as loss of chromosome 1p
and gain of chromosome 1q are found in paediatric
GCT, which may reflect loss of a tumour suppressor
gene on chromosome 1p that is involved in terminal
differentiation of embryonic tissues182
• Malignant GCT component found within a ‘benign’
teratoma component, which reflects stages of
malignant transformation182
• Carcinoma in situ cells resemble primordial germ cells
in adult-onset testicular GCTs181
B‑ALL Haematopoietic Pre-leukaemic population detected in utero in • Studies of monozygotic twins with concordant
system Etv6–Runx1‑transgenic mice86,87,184 leukaemia (ETV6–RUNX1‑positive pre‑B-ALL and infant
lymphoid-primed ALL)79,81
multipotent • Studies of neonatal blood spots from twin pairs show
progenitors clonal evolution in utero of a pre-leukaemic population
expressing ETV6–RUNX181,86
• Pre-leukaemic cells detected in neonatal blood spots
for most pre‑B-ALL patients2,82

278 | APRIL 2014 | VOLUME 14 www.nature.com/reviews/cancer

© 2014 Macmillan Publishers Limited. All rights reserved

 PERSPECTIVES

Table 1 (cont.) | Embryonal childhood malignancies

Malignancy Tissue of origin Evidence for prenatal origin
Pre-clinical Clinical
TMD and ML–DS Haematopoietic GATA1s‑knock‑in mouse model of TMD shows • Tissue of aborted second trimester fetuses with Down
system (MEPs) transient hyperproliferation in megakaryocytic syndrome shows a pathological increase in MEPs in
progenitors in the embryonic yolk sac and the the fetal liver prior to the acquisition of somatic GATA1
fetal liver67 mutation54
• GATA1 mutation detected in utero66
• In utero TMD diagnosis; condition present in preterm
neonates53
• Median age of diagnosis is 3–7 days postnatally, range
0–65 days52,53
• TMD regresses prior to progression to ML–DS in 25%
of patients49
MRTs Unknown cell of Heterozygous mutations of embryonic • Fetal and neonatal MRTs are frequently identified183
origin; frequently development regulator Smarcb1 in transgenic • Concomitant primary tumours in both the kidney and
arises in the mice cause MRT predisposition and, upon loss the brain183
kidney and central of heterozygosity, these mutations manifest as • Familial or sporadic inactivation of SMARCB1 occurs in
nervous system MRTs185 >95% of tumours10,186–188
ALK, anaplastic lymphoma kinase; B‑ALL, B cell‑lineage acute lymphoblastic leukaemia; ETV6, ETS variant 6; GCT, germ cell tumour; MEPs, megakaryocyte and
erythrocyte progenitors; ML–DS, myeloid leukaemia–Down syndrome; MRT, malignant rhabdoid tumour; OLIG3, oligodendrocyte transcription factor 3; PTCH1,
patched 1; RB1, retinoblastoma 1; RUNX1, runt-related transcription factor 1; SHH, sonic hedgehog; Th, tyrosine hydroxylase; TMD, transient myeloproliferative
disorder; WT1, Wilms’ tumour 1.

discussion of the origins of some childhood
cancers. In this article, we use a broader
term, ‘embryonal precancer’, for all pre­
malignant cellular hyperplasia that is detect-
able in the perinatal period. It is possible that
embryonal precancer cells are already on the
path towards malignancy but that these cells
retain the compulsion to die in the postnatal
environment because they still require addi-
tional postnatal changes for malignant trans-
formation. The aim of suggesting this term is
to facilitate discussion and experimentation
that is directed towards identifying common
aetiological features, strategies for early diag-
nosis and therapy, and possibly the prevention
of childhood cancers of prenatal origin.
Neuroblastoma
Neuroblastoma is a malignancy of the
sympathetic nervous system that almost
exclusively occurs in infancy and early child-
hood11,12. Primary neuroblastoma arises in
the adrenal medulla or sympathetic ganglia
along the paravertebral axis. Considerable
evidence suggests that neuroblastoma is
initiated in utero during sympathoadrenal
development and that it goes through an
embryonal precancer phase.
Candidate neuroblastoma initiation factors
linked to sympathoadrenal development.
The cells of the mature adrenal medulla and
sympathetic ganglia originate in a transient
collection of neural crest progenitors that rely
on spatially and temporally regulated extrin-
sic signals for the later steps of migration,
specification, divergence and maturation13
(FIG. 1). Sympathoadrenal cells from the neu-
ral crest in the trunk region of the embryo
follow a ventral migratory pathway from the
neural crest and neural tube, and they receive
signals from somites, the ventral neural tube,
the notochord and the dorsal aorta13. During
normal sympathoadrenal development,
expression of the proto-oncogene MYCN is
high in the early post-migratory neural crest,
where it regulates the ventral migration and
expansion of neural crest cells14. MYCN pro-
tein levels gradually reduce in differentiating
sympathetic neurons14–16, which suggests that
sympathoadrenal maturation requires low or
absent MYCN expression14. Consistent with
this finding is the observation that MYCN
transduction into quiescent rat sympathetic
neurons reactivates cell cycling and blocks
cell death that is induced by nerve growth
factor (NGF) withdrawal15,16.
After sympathoadrenal specification and
expansion of the primary sympathetic gan-
glia, sympathoadrenal precursor development
diverges into neuronal or chromaffin cell
fates13. On the basis of expression patterns of
differentiation markers, the earliest cell of ori-
gin for neuroblastoma is thought to be a neu-
ral crest cell specified to the sympathoadrenal
lineage that has not received or responded
to cues that determine neuronal or chroma­
ffin cell fate (FIG. 1). Excess neural precursors
undergo apoptotic cell death at the final stage
of sympathoadrenal maturation17, and this
process is catalysed by local NGF deprivation.
In zebrafish, persistent MYCN expression in
sympathoadrenal precursor cells markedly
blocks development towards a chromaffin
cell fate and thereby leads to neuroblastoma18.
Overexpression of human MYCN under the
control of the rat tyrosine hydroxylase (Th)
promoter in Th–MYCN-transgenic mice is
sufficient to recapitulate neuroblast precancer
(discussed below in the context of humans)
and tumorigenesis15,19,20. Transformation
from precancer to cancer in Th–MYCN mice
requires even higher levels of MYCN than
are provided by the transgene. This occurs
via transgene amplification and feedforward
loops between MYCN and increased levels
of the NAD-dependent deacetylases sirtuin 1
(SIRT1) and SIRT2, which increase MYCN
stability 21,22.
This cellular tolerance for high levels of
MYCN is surprising, as supraphysio­logical
levels of MYC expression should lead to
apoptosis and senescence through the ARF
(encoded by CDKN2A)–p53 stress response
pathway 23. Neuroblast precancer cell cultures
from perinatal Th–MYCN mouse ganglia have
lower basal and induced p53 levels than wild-
type ganglia, which in turn have lower p53
levels than those of mature ganglia19. When
p53 is reactivated in precancer cell cultures,
the cells become sensitive to NGF depriva-
tion19. Inactivation of the ARF–p53 pathway
in neuro­blast precancer cells can be accom-
plished by the Polycomb complex protein
BMI1, which functions as a p53 protein E3
ubiquitin ligase in Th–MYCN mice, or by
mutant anaplastic lymphoma kinase (ALK),
through effects on PI3K and MAPK signal-
ling in zebrafish19,24. These findings indicate
that there might be an inherent susceptibility
to oncogenic stress in some embryonal cells
that lack robust p53 stress responses.
These experiments mechanistically con-
nect MYCN with the maintenance of neuro-
blast precancer cells and indicate that MYCN
is a driver of resistance to developmentally-
timed trophic factor withdrawal signals

NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 279

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES

clinically presents in conjunction with the

neural crest disorders Hirschsprung disease
Neural crest and congenital hypoventilation syndrome; and
this reinforces the connection between neuro­
Neural tube blastoma and defective neural crest develop-
Migrating neural crest cell ment. PHOX2B mutant proteins seem to be
• MYCN oncogenic in a dominant-negative manner
• MASH1?
• PHOX2B? and they promote inappropriate prolifera-
tion in neural precursors29,30. Interestingly,
Notochord neuroblast precancer cells from Th–MYCN
• PHOX2A • GATA3
• PHOX2B • MASH1 mice have high levels of PHOX2B expres-
PSG
• GATA2 • HAND2 sion, which indicates a possible indirect
BMPs downstream link to MYCN31. Future experi-
Dorsal aorta
ments should uncover the finer detail of the
Sympathetic regulatory relationships between these
ganglion cell NGF proteins in development and disease.
Paralogous RNA-binding proteins
LIN28A and LIN28B regulate the expression
of the let‑7 microRNA family during embryo-
genesis, thereby controlling developmental
Maturation timing and cell growth during neural crest
• MYCN
• PHOX2B cell lineage commitment32. Targeted expres-
Apoptosis • ALK sion of LIN28B to the developing neural crest
First hit of transgenic mice induces neuroblastoma33.
LIN28B downregulates let‑7 pre-microRNAs,
which increases MYCN expression33. LIN28B
is known to be crucial in the maintenance of
Embryonal precancer
Death resistance
cell persistence
Neuroblastoma stemness throughout embryonic develop-
Second hit Third hit ment, and this function might maintain the
Birth undifferentiated neuroblast phenotype33,34.
It will be important to determine whether
Figure 1 | Neural crest development and neuroblastoma.  Under the influence NatureofReviews
MYCN and bone
| Cancer
Lin28b‑transgenic mice recapitulate the
morphogenetic proteins (BMPs), neuroblast progenitors migrate from the neural crest and around the
neural tube to a region that is immediately lateral to the notochord and dorsal aorta. At this site, neuroblast precancer disease that occurs in
the cells undergo specification as the primary sympathetic ganglia (PSG) before divergence into neural Th–MYCN mice. The known role of LIN28B
cells of the mature sympathetic ganglia or chromaffin cells (not shown). MYCN is a ‘first hit’ by virtue in neurogenesis35 fits well with a model of
of the observations from a tyrosine hydroxylase (Th)–MYCN-transgenic mouse model, whereas muta- neuro­blastoma initiation that requires aber-
tions in anaplastic lymphoma kinase (ALK) and paired-like homeobox 2B (PHOX2B) are germline rant regulation of developmental proteins.
mutations. Local access to nerve growth factor (NGF) determines whether a normal sympathetic gan- LIN28B can be amplified and highly expressed
glion cell (blue) matures into a terminal ganglion cell or undergoes apoptotic cell death. A relatively in some neuroblastoma tumours, which cor-
common pathological state is postnatal survival of neuroblast precancer cells (purple), which requires relates with poor prognosis33. Interestingly,
the cell that is destined to become malignant to be resistant to trophic factor withdrawal before these LIN28B–let‑7 signalling through MYCN
persistent precancer cells undergo a third change to induce transformation, which presents as neuro­
seems to be important in another putatively
blastoma in early childhood. HAND2, heart and neural crest derivatives expressed 2; MASH1, murine
achaete-scute homologue 1. embryonal cancer, germ cell tumour 36.
ALK has a key role in early sympatho­
adrenal development to protect neuroblast
at neuroblastoma initiation. The marked signalling from the dorsal aorta to promote a growth in utero against nutrient depriva-
downregulation of high affinity nerve growth transcriptional programme that specifies the tion37,38. Germline and somatic mutations in
factor receptor (NTRK1; also known as neuronal and catecholaminergic properties ALK that cause constitutive kinase activation
TRKA) in MYCN-amplified neuroblastoma of the expanding primary sympathetic gan- are present in 8–10% of neuroblastoma cases
cells blocks neural differentiation. Conversely, glia13,25,26. Murine achaete-scute homologue 1 and correlate with poor prognosis39,40–42. The
neurotrophic tyrosine kinase receptor type 2 (MASH1) then promotes paired mesoderm most common and aggressive activating
(NTRK2) responds to brain-derived neuro- homeobox protein 2A (PHOX2A) expres- mutation of ALK, which results in F1174L,
trophic factor (BDNF) which has survival- sion, which, together with PHOX2B, drives is sufficient for tumour formation when
and growth-promoting actions on neurons. the expression of enzymes for catecholamine specifically expressed in the neural crest
BDNF and NTRK2 are expressed by MYCN- biosynthesis. PHOX2B initiates a secondary in transgenic mice43 and when neural
expressing neuroblastoma cells and maintain transcriptional programme, which controls crest cells expressing this mutation are
an autocrine cell survival loop that blocks terminal sympathoadrenal differentia- transplanted into nude mice44. ALK‑F1174L
differentiation and apoptosis11. tion. Heterozygous germline mutations of has been associated with MYCN amplifica-
By 5 weeks into human development, PHOX2B are associated with a subset of tion in human tumours, and co‑expression
the ventrally migrating neural crest cell familial neuroblastoma27,28. Neuroblastoma of ALK‑F1174L and MYCN synergistically
responds to bone morphogenetic protein that is driven by PHOX2B mutations often promotes tumour formation in vivo, which

280 | APRIL 2014 | VOLUME 14 www.nature.com/reviews/cancer

© 2014 Macmillan Publishers Limited. All rights reserved


suggests that these may be cooperating
events in tumour initiation43. Thus, aberrant
MYCN expression or function in collabora-
tion with the disordered function of a small
group of key neurodevelopmental regulators
are good candidates for neuroblast precancer
initiation. The exact order of these events
and the intrauterine factors that favour
neuroblast precancer are unknown.
Neuroblastoma as an embryonal precancer.
Several clinical and experimental features
link neuroblastoma to defective embryo­
genesis and pathological neuroblast pre-
cancer. The idea of an embryonal origin
for human neuroblastoma is supported by
expression profiling showing that human
fetal adrenal neuroblasts have gene signa-
tures that are remarkably similar to those
of neuroblastomas45. In some patients with
neuroblastoma, new tumours form at dif-
ferent times and sites early in the child’s
life, which suggests a ‘field’ of premalignant
lesions, thereby linking tumorigenesis to
deregulated development.
A key feature of clinical or experimental
embryonal precancer is the compulsion to
undergo cell death and spontaneous regres-
sion in the postnatal environment, thereby
mirroring features of embryonal cells
in utero that are not required for organo-
genesis. The incidence of subclinical neuro-
blastoma is much higher than the frequency
of clinical disease. A small proportion of
patients with neuroblastoma present in early
infancy with metastatic special disease (pre-
viously known as stage 4S neuroblastoma),
which undergoes regression without therapy.
Autopsies of sympathoadrenal tissues from
infants whose diagnosed cause of death
was not cancer have shown an incidence of
neuroblast precancer that is 40‑fold higher
than the incidence of the clinical disease46.
Mass infant-screening programmes in the
1990s for the neuroblastoma tumour marker
of catecholamines in urine samples found
subclinical tumours at a frequency that was
more than twofold higher than the clinical
incidence of neuroblastoma47. These mass
screening programmes aimed to prevent the
later incidence of metastatic neuroblastoma
by early tumour detection in infancy at the
localized stage, followed by surgery, but they
were unsuccessful in this regard47. One
explanation is that neuroblast precancer
cells that later progress to metastatic neuro-
blastoma might already be fundamentally
different at birth from the more common
spontaneously regressing disease. This
hypothesis is suggested on the basis of work
using Th–MYCN-transgenic mice.
NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 281
The work showed that a small number of
neuroblast precancer cells that developed
into tumours had undergone early MYCN
transgene amplification15. Moreover, recent
transgenic zebrafish models of neuroblastoma
also indicate that there is a need for a cooper-
ating signal that inactivates MYCN-induced
apoptosis at tumour initiation18. Animal mod-
els confirm that metastatic neuroblastoma
can arise from neuroblast precancer cells15.
The postnatal period of quiescent neuro­
blast precancer cells in sympathoadrenal
tissues offers a time window during early
infancy when alternative forms of screen-
ing for precancer might be envisaged on
the basis of a germline cancer susceptibil-
ity profile and recently described methods
of detecting tumour DNA in peripheral
blood48. Furthermore, it might be expected
that drugs that restore the p53 stress
response in neuroblast precancer cells,
given briefly at this crucial early embryo-
nal precancer phase, might be effective
in restoring the capacity for spontaneous
regression and thereby preventing malignant
transformation.
Another prenatal cancer with an embryo-
nal precancer phase that offers substantial
opportunities for early detection and therapy
is TMD and ML–DS.
TMD and ML–DS
TMD is a megakaryocyte (MK) lineage dis-
ease that clinically presents at birth in 5–10%
of children with DS49,50. The clinical abnor-
malities that are associated with TMD spon-
taneously resolve in early infancy in almost
all affected patients. However, approximately
20% of these children present 2–4‑years later
with ML–DS, which is an acute MK lineage
leukaemia49,51.
Glossary
Adaptive B2 cells
B2 cells that are responsible for the production of
antibodies during adaptive immunity.
Anlage
The initial clustering of embryonic cells from which an
organ or part of an organ will form.
Chromothripsis
A process whereby a single catastrophic event within the
genome leads to multiple genetic alterations across one or
more chromosomes.
Innate B1 precursor cells
B1 cells with innate sensing and responding properties.
Marginal zone B cells
B cells from the marginal zone of the spleen, which is a
unique lymphoid area located at the interface between the
circulation and the immune system.
© 2014 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES
MK development and TMD pathogenesis.
TMD develops in utero during megakary-
opoiesis. MK progenitors (MKPs) are first
produced from haematopoietic stem cells
(HSCs) in the yolk sac, the aorto–gonad–
mesonephros and thereafter in the fetal
liver 52–54. Rarely, TMD can occur in infants
without DS, but all of these cases have tri-
somy 21 mosaicism in the bone marrow,
and this supports the key role of trisomy 21
in TMD pathogenesis. Moreover, the fre-
quency of conversion to ML–DS is the same
for patients with TMD that is linked to DS
or non‑DS trisomy 21 mosaicism50,52,53,55.
Fetal liver haematopoiesis in DS shows
specific expansion of HSC and MKP com-
partments56,57. These trisomy 21‑induced
changes are likely to occur during definitive
haematopoiesis from the yolk sac or the
aorto–gonad–mesonephros56,57. Ts65Dn
mice carry an extra chromosome, with gene
copies of most of mouse chromosome 16,
which is orthologous to human chromo-
some 21 (REF. 58). Ts65Dn mice do not spon-
taneously develop TMD, but they display
abnormal megakaryopoiesis and chronic
myelofibrosis58.
MK development and differentiation is
controlled by combinations of cytokines
and other mediators that are present within
a haematopoietic vascular and endosteal
niche57,59. Of the genes on chromosome 21,
several are good candidates for leukaemo-
genesis: runt-related transcription factor 1
(RUNX1; also known as AML1), ERG,
ETS2, dual-specificity tyrosine-(Y)-phos-
phorylation regulated kinase 1A (DYRK1Α)
and GA binding protein transcription
factor‑α (GABPA). RUNX1 expression is
not increased in ML–DS60, and trisomy of
Runx1 was not required for the development
Neonatal blood spot
A card with a drop of blood collected from the heel of a
newborn baby. Neonatal blood spots are used to screen
neonates for rare but serious metabolic conditions.
Neural crest
A transient collection of multipotent embryonic progenitors
in the developing ectoderm that gives rise to a multitude of
different cell types, including melanocytes, craniofacial
chondrocytes and osteocytes, smooth muscle myocytes
and peripheral nervous system neurons.
Rostrally
Pertaining to being situated towards the oral or nasal
region, or in the case of the brain, towards the tip of the
frontal lobe.
Somites
Bilaterally paired blocks of mesoderm that form along the
anterior–posterior axis of the developing embryo.
PERSPECTIVES

of a myeloproliferative disorder in Ts65Dn
mice58. However, the ETS transcription fac-
tor ERG can immortalize haematopoietic
progenitors61,62 and cooperate with a trun-
cated version of GATA1 to generate a
TMD-like defect in vivo63.
RUNX1 interacts with additional fac-
tors, including the transcriptional activator
GATA1 (REF. 64). GATA1 levels increase
and GATA2 levels decrease during MK dif-
ferentiation59. The GATA1 gene is mutated
in megakaryoblasts in all cases of TMD and
ML–DS, and these mutations, which occur
in exon 2 or at the exon 2–intron bound-
ary of GATA1, lead to exclusive production
of truncated GATA1, termed GATA1s65, as
early as 21 weeks of gestation66. GATA1s does
not have an amino‑terminal transactivation
domain, but it retains both DNA-binding zinc
fingers, and its expression leads to the impair-
ment of GATA1‑mediated regulation of other
transcription factors, including GATA2,
IKAROS family zinc finger 1 (IKZF1), MYB
and MYC in fetal MKs67. Paired sample
analysis by several groups indicates that the
same GATA1 mutation is present in TMD
and matched ML–DS samples53,68–70. However,
neither the mutation site nor the nature of the
mutation in GATA1 exon 2 predicts progres-
sion to ML–DS70 and, surprisingly, Kanezaki
et al.71 found an association between low
levels of GATA1s protein and a higher
probability of progression to ML–DS.
TMD as an embryonal precancer. TMD
develops in utero when blood cell production
resides in the fetal liver 52,53,72,73. Trisomy 21
creates MKP hyperplasia, and increased
MK and erythrocyte progenitor (MEP)
clonogenicity, which precedes GATA1s
expression54,56,74. Elegant mouse studies of the
effects of a GATA1s‑knock‑in allele showed
that GATA1s expression leads to the tran-
sient hyperproliferation of an embryonic
MKP that begins in the yolk sac, then moves
to the fetal liver and ultimately disappears
once haematopoiesis moves to the bone
marrow postnatally 67. Fetal liver-derived
megakaryoblasts have been recently shown
to differ from their adult counterparts by
their low type 1 interferon production: type 1
interferons are known inhibitors of MKPs.
This might help to explain why trisomy 21
and GATA1s have specific effects on fetal
MKPs75, and it also indicates that exogenous
interferon‑α might be an effective deletion
therapy for residual fetal MKPs in TMD.
Thus, to date, the evidence indicates that
the progression from normal prenatal MKP
to TMD requires trisomy 21 as the ‘first hit’
(REF. 49) and then somatic GATA1 mutation as
the ‘second hit’. We think that it is likely that
trisomy 21 and GATA1s combine to promote
TMD as an embryonal precancer in some
children with DS by causing inappropriate
MKP hyperplasia in the fetal liver. Like other
embryonal precancers, the MKP cells of TMD
retain the compulsion to die in the perinatal
period by as yet unknown mechanisms once
haematopoiesis shifts to the bone marrow.
At birth, TMD cells need a ‘third hit’ to
sustain viability in the face of death signals.
Whether these death signals also involve p53
is unknown. Transcription profiling studies
of TMD and ML–DS samples indicate that
driver mutations that activate WNT, Janus
kinase (JAK)–signal transducer and activator
of transcription (STAT) and MAPK–PI3K
signalling are good candidates for a ‘third hit’
(REF. 76) (FIG. 2). Furthermore, recent xenograft
analyses using samples from patients with
TMD show distinct subclones emerging on
serial transplantation with genomic features
of ML–DS, such as chromosome 16q dele-
tion and chromosome 1q gain, which were
present at low frequencies in the initial TMD
sample77. These observations are reminiscent
of the process of clonal selection for transgene
amplification that occurs in Th–MYCN-
transgenic mouse neuroblast precancer cells15,
and they support the idea that, even at birth,
rare leukaemogenic or tumorigenic subclones
are present among apparently regressing
embryonal precancer cell populations.
Collectively, we think that these data
suggest that embryonal precancer cells are
not normal embryonal cells that have sim-
ply persisted postnatally. Instead, postnatal
embryonal precancer cells, like cancer stem
cells, have retained embryonal features at the
early stages of a traditional pathway towards
tumorigenesis. Recent studies using next-
generation sequencing of peripheral blood
DNA from neonates with DS to look for
GATA1 mutations indicate that the incidence
of TMD could be around 30%78. These find-
ings suggest a strategy of predictive testing
for TMD in neonates with DS and pave the
way for therapeutic intervention studies of
TMD as a method of leukaemia prevention.
Another childhood cancer in which leu-
kaemia marker genes have been identified
in the perinatal period, suggesting prenatal
origins, is B‑ALL. However, the absence of
a clinical embryonal precancer phase before
B‑ALL and the rarity of the premalignant
clone in the perinatal period means that the
application of preventive disease strategies
for B‑ALL will be more difficult.
B‑ALL
ALL is the most common childhood malig-
nancy and arises in B‑lineage or T‑lineage
lymphocyte progenitor cells. We confine our
discussion to the more common B‑ALL and
B cell development, because T‑ALL gener-
ally presents in adolescence and most of the
evidence for a prenatal origin of ALL has
been obtained using samples from children
with B‑ALL.
Leukaemogenesis begins prenatally in chil‑
dren with B‑ALL. Leukaemic cells from
identical twin pairs with concordant B‑ALL
share unique, clonal, non-constitutive

Birth
MEP
GATA1s Death resistance
Foetal liver TMD persistence ML–DS
Second hit Third hit Fourth hit

Trisomy 21 Erythroid • Cell death

First hit cell ↑MEPs • Clinical resolution
TMD
Figure 2 | TMD and ML–DS.  Children with Down syndrome (DS) can mutant, truncated protein, GATA1s. All children with TMD and ML–DS have
develop a transient myeloproliferative disorder (TMD) at birth, which GATA1s and more than two copies of chromosome Nature Reviews
21. TMD | Cancer
resolves clini-
can later transform to myeloid leukaemia (ML–DS). The production of mega- cally in almost all patients but will later present as ML–DS in 25% of TMD
karyocyte and erythrocyte progenitors (MEPs) is increased in the fetal liver cases. Thus, some TMD cells at birth must survive through unique death
of children with DS and trisomy 21 as a ‘first hit’. Some megakaryocyte pre- resistance mechanisms and might undergo further changes, which lead to
cursor cells develop a ‘second hit’ mutation in GATA1, which results in a genomic instability and clinical presentation as ML–DS.

282 | APRIL 2014 | VOLUME 14 www.nature.com/reviews/cancer

© 2014 Macmillan Publishers Limited. All rights reserved

 PERSPECTIVES

• ETV6–RUNX1 Birth
Amnion • MLL–ENL
Amniotic sac Death resistance
Neural tube • BCR–ABL1
Second hit
First hit

B-ALL
Third hit
Foetal liver Bone
marrow • IKZF1
• E2A–PBX1
Aorta • CDKN2A
• CDKN2B

Pre-
HSC LMPP CLP pro-B
Yolk sac
• RUNX1 • RUNX1 • FLT3 • RUNX1 • FLT3 • RUNX1 • FLT3 • PAX5
• IKAROS • IKAROS • FOXO1 • IKAROS • FOXO1 • IKAROS • FOXO1 • EBF1
• E2A • E2A • E2A • IL-7 • E2A • IL-7

Figure 3 | The development of B‑ALL.  The presence of specific prenatal Nature

oncogenic fusion proteins in either haematopoietic stem cells (HSCs),
lymphoid-primed multipotent progenitors (LMPPs), common lymphoid pro-
genitors (CLPs) or pre-pro‑B cells combines with aberrant expression of
transcription factors that are necessary for normal B cell development and
maturation during the genesis of B cell‑lineage acute lymphoblastic leukae-
mia (B‑ALL). Fetal haematopoiesis begins in the haemogenic endothelium
formed from the yolk sac and aorto–gonad–mesonephros, then it localizes
to the fetal liver and finally resides in the bone marrow from the perinatal
period onwards. Several fusion genes (ETS variant 6 (ETV6)–runt-related
transcription factor 1 (RUNX1), mixed lineage leukaemia (MLL)–ENL and
BCR–ABL) present as clonal chromosome rearrangements at the point of
diagnosis of B‑ALL and thus represent the ‘first hit’, Reviews
as they Cancer
have |also been
identified in blood samples at birth from children who later developed
B‑ALL. We hypothesize that, like other types of embryonal precancer cells,
many of these partially transformed B‑precursor cells die in the perinatal
period, while a small number only survive because of a ‘second hit’ that leads
to death resistance. As the incidence of the ETV6–RUNX1 fusion gene is
much higher than the clinical incidence of B‑ALL, some B‑ALL precancer
cells must survive the perinatal period to undergo a ‘third hit’ that leads to
clinical presentation. CDKN2, cyclin-dependent kinase inhibitor 2; EBF1,
early B cell factor 1; FLT3, FMS-like tyrosine kinase 3; FOXO1, forkhead box
O1; IKZF1, IKAROS family zinc finger 1; IL‑7, interleukin‑7; PAX5, paired
box 5; PBX1, pre-B cell leukaemia homeobox 1.

chromosome rearrangements, even when
diagnosed years apart, which indicates that
leukaemogenic molecular alterations occur
in utero79,80. Further support for the idea
of a prenatal origin came from a study of
5‑year-old monozygotic twins that aimed
to identify leukaemogenic markers81.
These patients presented with B‑ALL and
concordant translocation of ETS variant 6
(ETV6; also known as TEL) and RUNX1
genes (ETV6–RUNX1). The leukaemic
cells also showed distinct immunoglobulin
heavy chain (IgH) and immunoglobulin-κ
deleting element (IgK-Kde) gene rearrange-
ments in neonatal blood spot DNA, which
indicates that separate pre-leukaemic clones
must have evolved before birth81. Using IgH
rearrangement as a marker for leukaemic
clones, several groups have now shown
that pre-leukaemic cells can be detected
in the neonatal blood spot of most child-
hood patients with B‑ALL, which indicates
that the leukaemogenic process begins in
utero2,82 (FIG. 3).
The ETV6–RUNX1 fusion gene is a
clonal feature of B‑ALL cells at diagnosis in
more than 20% of patients, but the analyses
of neonatal blood spots indicates that it is
present at a much higher frequency than the
incidence of B‑ALL in childhood83–85. Studies
of monochorionic ETV6–RUNX1‑positive
twin pairs (one pre-leukaemic and one leu-
kaemic) have established that the presence of
ETV6–RUNX1 in B cells is sufficient to gen-
erate a population of pre-leukaemic cells and
function as a first hit mutation that results
in altered self-renewal and survival proper-
ties86. However, low ETV6–RUNX1 expres-
sion levels are only tolerated by pro‑B cells,
because these cells terminally differentiate in
the long term87. Although ETV6–RUNX1 has
been found in peripheral blood leukocyte
DNA from healthy adults (0.5–8.8%), it is
much more frequent in young adults83, and
this presumably represents pre-leukaemic
clones that originated prenatally. ETV6–
RUNX1 expression is, of course, not part
of the normal B cell developmental pro-
gramme. Thus, ETV6–RUNX1 operates as
a weak oncogenic event in prenatal HSCs or
precursor B cells to generate a pre-leukaemic
state that, like precancerous neuroblasts
and megakaryoblasts, has a short lifespan in
the postnatal environment. ETV6–RUNX1
may function as an oncogene only in
neonates owing to specific characteristics
of fetal B cells (BOX 1), thereby linking
ETV6–RUNX1‑initiated leukaemia to the
embryonal environment. Overexpression of
the erythropoietin receptor early B cell fac-
tor 1 (EBF1) and deregulated transforming
growth factor‑β responses might have a role
in the survival of ETV6–RUNX1‑initiated
pre-leukaemic cells88–90. Both the BCR–ABL1
and MLL–ENL (also known as MLLT1)
fusion genes can occur prenatally and
might initiate B-ALL91,92. However, the pre-
leukaemic clone also remains clinically silent
in the absence of additional changes, such as
an IKZF1 mutation91.
We hypothesize that B‑ALL might also
follow a traditional tumorigenic pathway that
begins in prenatal, partially transformed cells
that have adopted characteristics favoured
by the embryonal environment but not
the postnatal environment. B‑ALL might
also be preceded by clonal hyperplasia of a
small population of transient B cells that,
like neuro­blast and MK precancer cells, are
destined to die unless further changes accel-
erate the path towards malignancy. A more
detailed analysis of genetically modified ani-
mal models of B‑ALL for lymphoid hyper-
plasia and its spontaneous regression would

NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 283

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES

strengthen the hypothesis that B‑ALL begins Considerable evidence from mouse mod- Cerebellar development and medulloblas‑
as an embryonal precancer. In this regard, the els suggests that medulloblastoma passes toma tumorigenesis. The fully developed
recent description of mutant TP53 as another through an embryonal precancer phase. cerebellum comprises an outer cell-sparse
familial B‑ALL predisposition gene suggests However, like B‑ALL, there is no clinical cortical layer (molecular layer), a Purkinje
that the p53 stress response might be defec- presentation of a precancer in patients who cell layer (PCL) and an internal granu-
tive in B‑ALL precancer cells — much like develop this brain tumour. lar layer (IGL). The IGL is comprised of
neuroblasts from the Th–MYCN mice15,19,93,94. cerebellar granule cells that form excita-
However, several important facts are not Medulloblastoma tory connections with Purkinje neurons95.
known for the prenatal phase of B‑ALL: Medulloblastoma arises from the develop- Developmentally, the cerebellar anlage arises
the exact embryonal timing and tissue of ing cerebellum and adjoining structures, from the dorsal part of the anterior hind-
initiation, and the leukaemogenic order and and it is the most common malignant brain brain, which is anteriorly and posteriorly
relative interdependency of the candidate tumour of childhood. It represents four delineated by expression of the homeo­
initiator mutations with B cell homeostasis in distinct disease entities: the sonic hedgehog box proteins orthodenticle homologue 2
the prenatal environment. The susceptibility (SHH) and WNT subgroups are named (OTX2) and HOXA2, respectively 96,97. OTX2
of fetal B cells to the weak ETV6–RUNX1 on the basis of the predominant signalling is a candidate driver for some group 3
oncoprotein might be due to the different pathway that drives tumorigenesis, whereas medullo­blastomas, as it is amplified in 20%
cell-intrinsic responses of HSCs or innate B1 group 3 and group 4 medulloblastomas of these tumours and frequently overex-
precursor cells or due to the cell-extrinsic sus- show a more pleiotropic aetiology and do pressed in SHH-independent subgroups98.
ceptibilities of the fetal liver compared with not have an easily identifiable oncogenic Expression of OTX2 is abundant in the
the postnatal bone marrow (BOX 1). driver pathway (FIG. 4). developing brain and silenced in the adult
brain. OTX2 has recently been shown to
repress differentiation in medulloblastoma
Box 1 | Regulation of B cell development and B‑ALL leukaemogenesis cells, which suggests that amplification
Haemogenic endothelium from the mouse embryonic day 9 (E9) yolk sac and the para-aortic of OTX2 could generate embryonal pre-
splanchnopleura independently give rise to the first B progenitor cells that preferentially cancer 99. The observation that OTX2 also
differentiate into innate B1 cells and marginal zone B cells, but not into adaptive B2 cells150. drives proliferation and can upregulate
Thereafter, B cells are generated from haematopoietic stem cells (HSCs) in the fetal liver until B cell MYC suggests that OTX2 might also be
generation is shifted to the bone marrow from birth onwards. Fetal liver-derived B cells show some important in transforming precancer into
descriptive and functional differences from bone marrow-derived B cells, such as low terminal medulloblastoma98,100.
deoxynucleotidyl transferase expression and a restricted diversity of the immunoglobulin heavy The first germinal centre of the devel-
chain variable and hinge region repertoire151. HSCs that are isolated from fetal liver can
oping cerebellum initiates along the fourth
differentiate into B1 and B2 cells, however, B1 cell production is predominant until the bone
marrow takes over and the B2 cell postnatal programme is established152. In embryonal or adult
ventricle in the dorsomedial ventricular
haematopoietic tissues, HSCs differentiate into lymphoid-primed multipotent progenitors zone and gives rise to Purkinje cells, as
(LMPPs), which express high levels of FMS-like tyrosine kinase 3 (FLT3), lose megakaryocyte and well as several other types of cerebellar
erythrocyte progenitor (MEP) potential and retain the capacity to differentiate into lymphoid and interneurons101. Neural tube closure forms
myeloid cells153. LMPPs can differentiate into early lymphoid progenitors, which are the precursors the rhombic lip, the anterior portion of
of bone marrow-derived common lymphoid progenitors (CLPs), which give rise to pre-pro‑B cells which provides a second germinal zone
that express B cell lineage-associated genes154,155. Mice that are deficient in FLT3 ligand have of rapidly proliferating cells that express
reduced CLP levels, whereas mice that are deficient in Flt3 and interleukin‑7 receptor (Il7r) do not radial glial markers and gives rise to
develop B cells156,157. MATH1 (also known as protein atonal
A plethora of studies has revealed structural rearrangements, deletions, amplifications and point
homologue 1)-positive granule cell pro-
mutations in genes that encode regulators of B lymphocyte development in samples from patients
with B‑lineage acute lymphoblastic leukaemia (B-ALL), and many of these regulators are
genitors (GCPs)102,103. From 24 to 40 weeks
candidates for leukaemia initiation; many are transcription factors that regulate B cell of gestation in humans, the volume of the
development158. IKAROS family zinc finger 1 (Ikzf1)‑deficient HSCs in mice have a reduced developing cerebellum enlarges fivefold,
expression of IL‑7R and FLT3, and the numbers of LMPPs are reduced; and reduced expression of which corresponds to a >30‑fold increase
IKZF1 in mice results in impaired transition from the pro‑B cells to pre‑B cells. Ikzf1−/− mice in cerebellar cortex surface area104,105.
completely lack B cells159–161. IKZF1 is deleted in most BCR–ABL1‑positive B‑ALL cells, and mice that This expansion is mostly imparted by
express mutant IKZF1 develop ALL162. the rapid proliferation of GCPs, which
Conditional deletion of paired box 5 (Pax5) in mature B cells leads to dedifferentiation towards give rise to cerebellar granule cells and
uncommitted HSCs163. Somatic deletion, mutation or rearrangement of PAX5 is detectable in more almost all cerebellar cortical neurons101,106.
than 30% of children with B-ALL164. PAX5 DNA-binding and internal deletion mutants function as
Proliferation of these neuronal progeni-
hypomorphic alleles with weak competitive activity164. Hypomorphic PAX5 mutations have
recently been described in studies of two affected families with inherited B‑ALL, thereby
tors depends on SHH that is produced by
representing familial B‑ALL predisposition genes165. Purkinje neurons: inhibiting SHH signal-
E2A, which encodes the E protein transcriptional regulators E12 and E47, activates the expression ling blocks the proliferation and migration
of forkhead box O1 (FOXO1) and is required for CLP formation and B lineage specification166–168. of GCPs107. GCPs rostrally migrate along
Early B cell factor 1 (EBF1) is necessary for pre-pro‑B cells to transit to the pro‑B stage169. Forced the surface of the developing cerebel-
overexpression of EBF1 in both wild-type and E2A‑deficient HSCs leads to B cell differentiation170. lum to form the external granular layer
Runt-related transcription factor 1 (RUNX1) is indispensable for generating CLPs171, and the (EGL), where they continue to divide and
expression of PAX5, E12, E47 and EBF1 is reduced in B cell precursors that lack RUNX1 (REF. 170). eventually migrate inwards to form the
Runx1 deficiency in mice leads to a severe reduction in CLPs and to a myeloproliferative IGL, with consequent depletion of the
phenotype171.
EGL108,109. This process of GCP maturation

284 | APRIL 2014 | VOLUME 14 www.nature.com/reviews/cancer

© 2014 Macmillan Publishers Limited. All rights reserved


and concurrent spontaneous cell death of
excess GCPs is complete by the postnatal
age of 20 months110.
The connection between SHH signalling
and medulloblastoma was initially made in
the context of Gorlin syndrome, which is an
autosomal dominant disorder that causes
developmental defects and predisposes
individuals to several cancers, including
medulloblastoma111. Fine mapping showed
the putative gene to be highly homologous
to the Drosophila gene Patched (ptc; known
as PTCH1 in humans)112. Loss of chromo-
some 9q, which contains PTCH1, occurs
in ~30% of all tumours of the SHH sub-
group113. The PTCH1 protein is an essen-
tial negative regulator of SHH signalling;
thus, mice that are heterozygous for Ptch1
develop cerebellar medulloblastoma114.
Other germline mutations that predispose
children to SHH medulloblastoma occur
because of loss‑of‑function mutation in
Suppressor of fused homologue (SUFU),
which is a negative regulator of the SHH sig-
nalling 115. Amplification of GLI family zinc
finger 2 (GLI2), which is a positive regula-
tor of the SHH signal, is also implicated
in medulloblastoma development 116. In a
mouse model of SHH-induced medullo­
blastoma, tumour formation was preceded
by GCP hyperplasia in the first week of life,
which regressed before later emerging as a
malignant tumour 117, in a manner similar
to neuroblast hyperplasia in the Th–MYCN
mice15. However, in vivo labelling studies
that follow a precancer to cancer transfor-
mation have not yet been carried out in
either model. In vitro, the murine GCPs were
resistant to SHH withdrawal, and this feature
was partially MYCN-dependent 117. Neural
stem cells (NSCs) and GCPs have repeatedly
been shown to be the cell of origin for SHH
medulloblastoma118–122.
MYCN and its homologue MYC have
crucial roles in all medulloblastoma sub-
groups. Brain-specific germline deletion of
mouse Mycn results in cerebellar dyspla-
sia123. Expression of MYCN is essential for
SHH medulloblastoma in mouse models124,
and amplification of MYCN marks a sub-
set of poor-outcome SHH-driven human
tumours125. Moreover, tumorigenic SHH
signalling markedly increases MYCN
expression in GCPs because of effects on
MYCN protein stability 117. MYCN is also
expressed in WNT-subgroup and group 4
tumours, with targeted expression of
MYCN or MYC driving SHH-independent
medulloblastoma in transgenic mice126–128.
Group 3 tumours express MYC and have
amplification of MYC rather than MYCN129.
NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 285
PERSPECTIVES
Hb GCPs EGL
VZ
URL
VZ
LRL
SHH
• Nestin Persistent
• GFAP EGL
• MATH1
• OLIG2
Molecular
layer
Purkinje
cell layer
IGL
WNT
• BLBP
• OLIG3
Group 3: Group 4:
• MYC→MATH1 MYCN→GLT1
• Prominin 1
Brainstem
Figure 4 | Cerebellar development and embryonal origin of medulloblastoma.  The upper (ante-
Nature
rior) rhombic lip (URL) is a germinal zone of proliferating granule cell precursors Reviews
(GCPs) that| Cancer
migrate
rostrally to form the external granular layer (EGL). A second germinal centre is the ventricular zone
(VZ), which gives rise to Purkinje cells and several other types of cerebellar interneurons. Sonic hedge-
hog (SHH) medulloblastoma-subgroup tumours might arise from persistent cells of the EGL that have
not migrated to the internal granular layer (IGL), whereas WNT-subgroup tumours probably originate
from the lower rhombic lip (LRL) and embryonic dorsal brainstem. Group 3 and 4 medulloblastomas
might arise from neural stem cells of the hindbrain or the brainstem. Beige boxes indicate the identity
of the originating cell in mouse models. MYC and MYCN are driving oncogenes for group 3 and group 4
tumour models, respectively. GFAP, glial fibrillary acidic protein; Hb, hindbrain; Mb, midbrain;
OLIG, oligodendrocyte transcription factor.
Thus, there are many similarities between WNT signalling in members of the WNT
the role of MYCN in generating neuroblast signalling pathway have been described in
hyperplasia in neuroblastoma and the role sporadic medulloblastomas132–134. Recent
of MYCN in generating GCP hyperplasia in data from the Medulloblastoma Advanced
medulloblastoma. Genomics International Consortium
WNT signalling has many roles in neural showed that β‑catenin (encoded by
development, and aberrant WNT signal- CTNNB1) displayed canonical exon 3 dele-
ling in NSCs of the cerebellar ventricular tion in most (70–80%) WNT-subgroup
zone induces transient proliferation and medulloblastomas116,133,135.
impaired differentiation130. Individuals with Insight into the cell of origin for WNT-
familiar adenomatous polyposis (FAP) and subgroup medulloblastomas comes from
WNT pathway-driven colorectal cancer two mouse models, both of which involve
also have an increased risk of developing the conditional expression of degradation-
medulloblastoma131. Mutations that increase resistant β‑catenin under the control of
© 2014 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES
Blbp (also known as Fabp7) in the context
of Trp53 deletion (Blbp-cre;Ctnnb1;Trp53).
The addition of MYC expression increased
penetrance from 4% (Trp53+/−) and 15%
(Trp53−/−) to 83%, which suggests that acti-
vation of β‑catenin and WNT signalling
alone are only weak oncogenic events136,137.
These MYC-driven medulloblastomas were
evident in young mice and were formed
from BLBP+ oligodendrocyte transcription
factor 3‑positive (OLIG3+) mossy-fibre
neuron precursors that are present in the
brainstem between embryonic day 11.5
(E11.5) and E15.5, which is consistent with
the radiological observation that WNT
medulloblastomas are frequently located
within the fourth ventricle and infiltrate the
dorsal surface of the brainstem126,137,138. This
observation was consistent with the model
of OLIG3+ neural precursors representing
a brain stem precancer preceding WNT
medulloblastoma.
The pathogenesis of group 3 and group 4
medulloblastoma is less clear, although these
groups account for more than 60% of all
medulloblastoma cases139. Targeted induc-
ible expression of MYCN to the postnatal
cerebellum using the glial high affinity
glutamate transporter (Glt1; also known as
Slc1a2) promoter (Glt1−tTA;TRE−MYCN−
Luc; also known as GTML mice) leads to
tumours with a transcriptional profile of
group 3 medulloblastoma126. GTML mice
had a normal EGL, which suggests that
group 3 medulloblastoma can originate
from a cellular population that is distinct
from SHH-subgroup medulloblastoma.
However, proliferating GCPs that are isolated
Normal embryogenesis
Proliferative excess First hit
Birth
Death resistance and embryonal
Second hit
precancer cell persistence
Accelerated path to
Third hit
genomic instability
Figure 5 | A model of embryonal tumorigenesis.
We propose that eachNatureof theReviews
four embryonal
| Cancer
malignancies with evidence of a prenatal origin
(neuroblastoma, myeloid leukaemia–Down syn-
drome, B‑lineage acute lymphoblastic leukaemia
and medulloblastoma) shares common features:
a prenatal proliferative excess in the tissue of
origin, a cell-intrinsic mechanism for surviving
the hostile early postnatal environment and
an accelerated pathway towards genomic
instability.
286 | APRIL 2014 | VOLUME 14 www.nature.com/reviews/cancer
from cyclin-dependent kinase inhibitor 2C
(Cdkn2c)−/−, Trp53−/−, Math1–green fluores-
cent protein (GFP) mice generate group 3
medulloblastoma when transformed with
MYC but not MYCN127. Furthermore,
group 3 tumours have been modelled
through MYC transformation of postnatal
cerebellar stem cells marked by the expres-
sion of prominin 1 (also known as CD133)
and the lack of neuronal or glial markers128.
Thus, it seems that a single oncogene can
give rise to multiple different medullo­
blastoma tumour subtypes, depending on the
developmental age and anatomical origin of
the transformed cell, and this indicates that
susceptibility to medulloblastoma is highly
dependent on the embryonal site and stage140.
Unifying features of prenatal cancers
The work of Bishop and Varmus141 defined
the forces that drive human cancer as aberra-
tions of our normal adult self 142. Embryonal
cancer initiation seems to be an aberration
of our prenatal self. Here, we propose a new
definition of embryonal precancer cells that
includes all cancers with a prenatal origin,
whether they arise directly from an embryo-
nal cell or from a more mature prenatal cell
that has acquired pathological properties that
favour survival in the postnatal environment.
Using this new definition, neuroblastoma
and TMD seems to initiate prenatally in cells
that display resistance to cell death, numeri-
cal excess and differentiation arrest at a stage
of tissue ontogeny that is several steps distal
to the earliest progenitor cell for each organ
(FIG. 5). We hypothesize that B‑ALL cells
might also pass through this phase during
their evolution towards the clearly malignant
state. Mouse models suggest similar features
exist for some medulloblastoma subtypes.
Indeed, much of the direct evidence for an
embryonal precancer phase for these four
malignancies comes from mouse models.
Thus, these findings raise interesting ques-
tions but do not definitively indicate that
the same pathways operate in humans. We
propose that these precancer cells are groups
of abnormal embryonal cells that have under-
gone changes that allow them to persist in the
face of cell deletion signals in the postnatal
environment, but they retain the capacity for
self-destruction in the postnatal milieu unless
other tumorigenic factors intervene to cause
genomic instability — a feature that might
offer an opportunity for embryonal precancer
deletion therapy. Neuroblastoma and TMD
modelling indicates that all embryonal pre-
cancer cells are not equal at birth, with only
a small number undergoing transformation
after later exposure to unidentified postnatal
© 2014 Macmillan Publishers Limited. All rights reserved
environmental influences15,77. The properties
that permit some embryonal precancer cells
to remain dormant after birth while adjacent
cells are dying is currently unknown.
For neuroblastoma and medulloblastoma,
the timing and location of the embryonal
precancer in utero coincides with a point in
the tissue-specific maturation pathway when
rapid cell expansion is required, and this is
quickly followed by an orderly transition to
a terminally differentiated state. Substantial
cell death must occur during the final shap-
ing of the nervous system. It is resistance to
these death cues that must accompany an
inappropriate replication stimulus, such as
continued or high levels of MYCN expres-
sion. We postulate that individual variation
in factors that are required to negatively
regulate MYCN expression in specific tis-
sues during embryogenesis; to activate the
apoptosis or senescence response to aber-
rant MYCN expression; or to activate death
responses that are imposed by trophic factor
withdrawal might explain individual
susceptibility to embryonal precancer.
From early in embryogenesis, haemato­
poietic precursors are in a continual state of
replication and self-renewal, in contrast to
terminally differentiated sympathetic ganglia
and cerebellar neurons. Thus, although both
TMD and B‑ALL might not arise directly
from an embryonal cell, the initial steps in
each pathway towards postnatal malignancy
involve progenitor cells acquiring character-
istics that are favoured in the prenatal devel-
opment. For TMD (and possibly B‑ALL),
the timing of precancer initiation coincides
with a unique predominance of fetal liver
haematopoiesis. In the case of TMD, low
interferon responses might be permissive
for the survival of trisomy 21‑primed and
GATA1s mutant MKPs in the normally non-
permissive environment of the bone marrow
that exists beyond the perinatal period75.
The precancer to cancer transition occurs
in early childhood, which indicates that some
precancer cells must have the capacity for
rapid progression towards genomic instabil-
ity. One explanation for the speed of this
process is inherited germline loss‑of‑function
mutation in one allele of a tumour sup-
pressor gene143. Another recently proposed
mechanism is chromothripsis, whereby a
single catastrophic event can lead to mas-
sive genomic rearrangement that affects
more than one chromosome, rather than the
incremental acquisition of single oncogenic
mutations over decades144–146. Additional
factors that might accelerate the progression
from precancer to cancer are several recently
described feedforward loops that increase

the expression of MYCN to very high levels
through the effects of SIRT1, SIRT2 and
aurora kinase A on MYCN protein stabil-
ity 21,22,147. We propose that additional factors
favouring rapid malignant transformation of
precancer cells are those properties of growth
and self-renewal that are inherent in some
embryonal cells.
Although chemotherapy, surgery and
radiotherapy have improved the cure rates
for childhood cancer, this has come at a sub-
stantial cost in terms of the severe acute and
long-term side effects of therapy, as well as
considerable consumption of the paediatric
health economy. Early evidence from screen-
ing trials in children at increased risk of
cancer development owing to germline p53
mutations has revived interest in strategies
for the prevention of childhood cancer 148,149.
Although the early detection of neuro-
blastoma using increased urine catechola-
mines failed to detect all neuroblastoma,
we envisage greater success when genomic
susceptibility markers are incorporated
into predictive algorithms for embryonal
pre­cancer. Therapy that aims to transiently
restore p53‑mediated or interferon-mediated
cell death responses or to block MYCN, ALK,
PI3K or MAPK signals in embryonal pre­
cancer cells is an exciting strategy for child-
hood cancer prevention, which must first be
demonstrated to be effective in mouse mod-
els of embryonal cancer. It is apparent that,
with current models and a clearer under-
standing of the prenatal missteps that initiate
embryonal precancer, these novel strategies
can be contemplated for the first time.
Glenn M. Marshall and Marion K. Mateos are at the
Kids Cancer Centre, Sydney Children’s Hospital,
Randwick 2031, New South Wales, Australia; and the
Children’s Cancer Institute Australia for Medical
Research, Lowy Cancer Centre, University of
New South Wales, Randwick 2031, Australia.
Daniel R. Carter, Belamy B. Cheung and Tao Liu are at
the Children’s Cancer Institute Australia for Medical
Research, Lowy Cancer Centre, University of
New South Wales, Randwick 2031, Australia.
Justin G. Meyerowitz and William A. Weiss are at the
Department of Neurology, Helen Diller Family
Comprehensive Cancer Center, University of California,
San Francisco, California 94158, USA.
Correspondence to G.M.M.  26. Schneider, C., Wicht, H., Enderich, J., Wegner, M. &
e‑mail: g.marshall@unsw.edu.au
doi:10.1038/nrc3679
Published online 6 March 2014
1. Valent, P. et al. Cancer stem cell definitions and
terminology: the devil is in the details. Nat. Rev.
Cancer 12, 767–775 (2012).
2. Greaves, M. F., Maia, A. T., Wiemels, J. L. &
Ford, A. M. Leukemia in twins: lessons in natural
history. Blood 102, 2321–2333 (2003).
3. Eaton, K. W., Tooke, L. S., Wainwright, L. M.,
Judkins, A. R. & Biegel, J. A. Spectrum of SMARCB1/
INI1 mutations in familial and sporadic rhabdoid
tumors. Pediatr. Blood Cancer 56, 7–15 (2011).
NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 287
4. Nuchtern, J. G. et al. A prospective study of expectant
observation as primary therapy for neuroblastoma in
young infants: a Children’s Oncology Group study.
Ann. Surg. 256, 573–580 (2012).
5. Beckwith, J. B. Precursor lesions of Wilms tumor:
clinical and biological implications. Med. Pediatr.
Oncol. 21, 158–168 (1993).
6. Cohnheim, J. Congenitales, quergestreiftes
Muskelsarkom der Nieren. Virchows Archiv 65,
64–69 (in German) (1875).
7. Durante, F. Nesso fisio-patologico tra la struttura dei
nei materni e la genesi di alcuni tumori maligni. Arch.
Memo. Observ. Chir Prat 11, 217–226 (in Italian)
(1874).
8. Virchow, R. Die multiloculäre, ulcerirende
Echinokokkengeschwulst der Leber. Verhandlungen
der Physicalisch-Medicinischen Gesellschaft, 6,
84–95 (in German) (1855).
9. Gruhn, B. et al. Prenatal origin of childhood acute
lymphoblastic leukemia, association with birth weight
and hyperdiploidy. Leukemia 22, 1692–1697
(2008).
10. Pine, S. R. et al. Incidence and clinical implications of
GATA1 mutations in newborns with Down syndrome.
Blood 110, 2128–2131 (2007).
11. Brodeur, G. M. Neuroblastoma: biological insights into
a clinical enigma. Nat. Rev. Cancer 3, 203–216
(2003).
12. Maris, J. M. Recent advances in neuroblastoma. New
Engl. J. Med. 362, 2202–2211 (2010).
13. Huber, K. The sympathoadrenal cell lineage:
specification, diversification, and new perspectives.
Dev. Biol. 298, 335–343 (2006).
14. Zimmerman, K. A. et al. Differential expression of myc
family genes during murine development. Nature 319,
780–783 (1986).
15. Hansford, L. M. et al. Mechanisms of embryonal
tumor initiation: distinct roles for MycN expression
and MYCN amplification. Proc. Natl Acad. Sci. USA
101, 12664–12669 (2004).
16. Wartiovaara, K., Barnabe-Heider, F., Miller, F. D. &
Kaplan, D. R. N‑myc promotes survival and induces
S‑phase entry of postmitotic sympathetic neurons.
J. Neurosci. 22, 815–824 (2002).
17. Yuan, J. & Yankner, B. A. Apoptosis in the nervous
system. Nature 407, 802–809 (2000).
18. Zhu, S. et al. Activated ALK collaborates with MYCN in
neuroblastoma pathogenesis. Cancer Cell 21,
362–373 (2012).
19. Calao, M. et al. Direct effects of Bmi1 on p53 protein
stability inactivates oncoprotein stress responses in
embryonal cancer precursor cells at tumor initiation.
Oncogene 32, 3616–3626 (2013).
20. Weiss, W. A., Aldape, K., Mohapatra, G.,
Feuerstein, B. G. & Bishop, J. M. Targeted expression
of MYCN causes neuroblastoma in transgenic mice.
EMBO J. 16, 2985–2995 (1997).
21. Marshall, G. M. et al. SIRT1 promotes N‑Myc
oncogenesis through a positive feedback loop
involving the effects of MKP3 and ERK on N‑Myc
protein stability. PLoS Genet. 7, e1002135 (2011).
22. Liu, P. Y. et al. The histone deacetylase SIRT2
stabilizes Myc oncoproteins. Cell Death Differ. 20,
503–514 (2013).
23. Murphy, D. J. et al. Distinct thresholds govern Myc’s
biological output in vivo. Cancer Cell 14, 447–457
(2008).
24. Berry, T. et al. The ALKF1174L mutation potentiates the
oncogenic activity of MYCN in neuroblastoma. Cancer
Cell 22, 117–130 (2012).
25. Reissmann, E. et al. Involvement of bone
morphogenetic protein‑4 and bone morphogenetic
protein‑7 in the differentiation of the adrenergic
phenotype in developing sympathetic neurons.
Development 122, 2079–2088 (1996).
Rohrer, H. Bone morphogenetic proteins are required
in vivo for the generation of sympathetic neurons.
Neuron 24, 861–870 (1999).
27. Mosse, Y. P. et al. Germline PHOX2B mutation in
hereditary neuroblastoma. Am. J. Hum. Genet. 75,
727–730 (2004).
28. Trochet, D. et al. Germline mutations of the paired-like
homeobox 2B (PHOX2B) gene in neuroblastoma. Am.
J. Hum. Genet. 74, 761–764 (2004).
29. Bourdeaut, F. et al. Germline mutations of the paired-
like homeobox 2B (PHOX2B) gene in neuroblastoma.
Cancer Lett. 228, 51–58 (2005).
30. Reiff, T. et al. Neuroblastoma phox2b variants stimulate
proliferation and dedifferentiation of immature
sympathetic neurons. J. Neurosci. 30, 905–915 (2010).
© 2014 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES
31. Alam, G. et al. MYCN promotes the expansion of
Phox2B‑positive neuronal progenitors to drive
neuroblastoma development. Am. J. Pathol. 175,
856–866 (2009).
32. Rybak, A. et al. A feedback loop comprising lin‑28 and
let‑7 controls pre-let‑7 maturation during neural stem-
cell commitment. Nature Cell Biol. 10, 987–993
(2008).
33. Molenaar, J. J. et al. LIN28B induces neuroblastoma
and enhances MYCN levels via let‑7 suppression.
Nature Genet. 44, 1199–1206 (2012).
34. Yu, J. et al. Induced pluripotent stem cell lines derived
from human somatic cells. Science 318, 1917–1920
(2007).
35. Balzer, E., Heine, C., Jiang, Q., Lee, V. M. &
Moss, E. G. LIN28 alters cell fate succession and acts
independently of the let‑7 microRNA during
neurogliogenesis in vitro. Development 137,
891–900 (2010).
36. Murray, M. J. et al. LIN28 Expression in malignant
germ cell tumors downregulates let‑7 and increases
oncogene levels. Cancer Res. 73, 4872–4884
(2013).
37. Reiff, T. et al. Midkine and Alk signaling in sympathetic
neuron proliferation and neuroblastoma
predisposition. Development 138, 4699–4708
(2011).
38. Cheng, L. Y. et al. Anaplastic lymphoma kinase spares
organ growth during nutrient restriction in Drosophila.
Cell 146, 435–447 (2011).
39. Mosse, Y. P. et al. Identification of ALK as a major
familial neuroblastoma predisposition gene. Nature
455, 930–935 (2008).
40. Janoueix-Lerosey, I. et al. Somatic and germline
activating mutations of the ALK kinase receptor in
neuroblastoma. Nature 455, 967–970 (2008).
41. Chen, Y. Y. et al. Oncogenic mutations of ALK kinase in
neuroblastoma. Nature 455, 971–974 (2008).
42. De Brouwer, S. et al. Meta-analysis of neuroblastomas
reveals a skewed ALK mutation spectrum in tumors
with MYCN amplification. Clin. Cancer Res. 16,
4353–4362 (2010).
43. Heukamp, L. C. et al. Targeted expression of mutated
ALK induces neuroblastoma in transgenic mice. Sci.
Transl. Med. 4, ra91 (2012).
44. Schulte, J. H. et al. MYCN and ALKF1174L are
sufficient to drive neuroblastoma development from
neural crest progenitor cells. Oncogene 21,
1059–1065 (2012).
45. De Preter, K. et al. Human fetal neuroblast and
neuroblastoma transcriptome analysis confirms
neuroblast origin and highlights neuroblastoma
candidate genes. Genome Biol. 7, 17 (2006).
46. Beckwith, J. B. & Perrin, E. V. In situ neuroblastomas:
a contribution to the natural history of neural crest
tumors. Am. J. Pathol. 43, 1089–1104 (1963).
47. Woods, W. G. et al. A population-based study of the
usefulness of screening for neuroblastoma. Lancet
348, 1682–1687 (1996).
48. Murtaza, M. et al. Non-invasive analysis of acquired
resistance to cancer therapy by sequencing of plasma
DNA. Nature 497, 108–112 (2013).
49. Roy, A., Roberts, I., Norton, A. & Vyas, P. Acute
megakaryoblastic leukaemia (AMKL) and transient
myeloproliferative disorder (TMD) in Down syndrome:
a multi-step model of myeloid leukaemogenesis. Br.
J. Haematol. 147, 3–12 (2009).
50. Brodeur, G. M., Dahl, G. V., Williams, D. L.,
Tipton, R. E. & Kalwinsky, D. K. Transient leukemoid
reaction and trisomy 21 mosaicism in a phenotypically
normal newborn. Blood 55, 691–693 (1980).
51. Hasle, H. et al. A pediatric approach to the WHO
classification of myelodysplastic and myeloproliferative
diseases. Leukemia 17, 277–282 (2003).
52. Massey, G. V. et al. A prospective study of the natural
history of transient leukemia (TL) in neonates with
Down syndrome (DS): Children’s Oncology Group
(COG) study POG‑9481. Blood 107, 4606–4613
(2006).
53. Klusmann, J. H. et al. Treatment and prognostic
impact of transient leukemia in neonates with Down
syndrome. Blood 111, 2991–2998 (2008).
54. Tunstall-Pedoe, O. et al. Abnormalities in the myeloid
progenitor compartment in Down syndrome fetal liver
precede acquisition of GATA1 mutations. Blood 112,
4507–4511 (2008).
55. Polski, J. M. et al. Acute megakaryoblastic leukemia
after transient myeloproliferative disorder with clonal
karyotype evolution in a phenotypically normal
neonate. J. Pediatr. Hematol. Oncol. 24, 50–54
(2002).
PERSPECTIVES
56. Maclean, G. A. et al. Altered hematopoiesis in trisomy
21 as revealed through in vitro differentiation of
isogenic human pluripotent cells. Proc. Natl Acad. Sci.
USA 109, 17567–17572 (2012).
57. Chou, S. T. et al. Trisomy 21‑associated defects in
human primitive hematopoiesis revealed through
induced pluripotent stem cells. Proc. Natl Acad. Sci.
USA 109, 17573–17578 (2012).
58. Kirsammer, G. et al. Highly penetrant
myeloproliferative disease in the Ts65Dn mouse model
of Down syndrome. Blood 111, 767–775 (2008).
59. Dore, L. C. & Crispino, J. D. Transcription factor
networks in erythroid cell and megakaryocyte
development. Blood 118, 231–239 (2011).
60. Bourquin, J. P. et al. Identification of distinct molecular
phenotypes in acute megakaryoblastic leukemia by
gene expression profiling. Proc. Natl Acad. Sci. USA
103, 3339–3344 (2006).
61. Stankiewicz, M. J. & Crispino, J. D. ETS2 and ERG
promote megakaryopoiesis and synergize with
alterations in GATA‑1 to immortalize hematopoietic
progenitor cells. Blood 113, 3337–3347 (2009).
62. Salek-Ardakani, S. et al. ERG is a megakaryocytic
oncogene. Cancer Res. 69, 4665–4673 (2009).
63. Birger, Y. et al. Perturbation of fetal hematopoiesis in
a mouse model of Down syndrome’s transient
myeloproliferative disorder. Blood 122, 988–998
(2013).
64. Elagib, K. E. et al. RUNX1 and GATA‑1 coexpression
and cooperation in megakaryocytic differentiation.
Blood 101, 4333–4341 (2003).
65. Gurbuxani, S., Vyas, P. & Crispino, J. D. Recent
insights into the mechanisms of myeloid
leukemogenesis in Down syndrome. Blood 103,
399–406 (2004).
66. Taub, J. W. et al. Prenatal origin of GATA1 mutations
may be an initiating step in the development of
megakaryocytic leukemia in Down syndrome. Blood
104, 1588–1589 (2004).
67. Li, Z. et al. Developmental stage-selective effect of
somatically mutated leukemogenic transcription factor
GATA1. Nature Genet. 37, 613–619 (2005).
68. Hitzler, J. K., Cheung, J., Li, Y., Scherer, S. W. &
Zipursky, A. GATA1 mutations in transient leukemia
and acute megakaryoblastic leukemia of Down
syndrome. Blood 101, 4301–4304 (2003).
69. Rainis, L. et al. Mutations in exon 2 of GATA1 are early
events in megakaryocytic malignancies associated with
trisomy 21. Blood 102, 981–986 (2003).
70. Alford, K. A. et al. Analysis of GATA1 mutations in
Down syndrome transient myeloproliferative disorder
and myeloid leukemia. Blood 118, 2222–2238
(2011).
71. Kanezaki, R. et al. Down syndrome and GATA1
mutations in transient abnormal myeloproliferative
disorder: mutation classes correlate with progression
to myeloid leukemia. Blood 116, 4631–4638 (2010).
72. Zipursky, A. Transient leukaemia—a benign form of
leukaemia in newborn infants with trisomy 21. Br.
J. Haematol. 120, 930–938 (2003).
73. Muramatsu, H. et al. Risk factors for early death in
neonates with Down syndrome and transient
leukaemia. Br. J. Haematol. 142, 610–615 (2008).
74. Roy, A. et al. Perturbation of fetal liver hematopoietic
stem and progenitor cell development by trisomy 21.
Proc. Natl Acad. Sci. USA 109, 17579–17584 (2012).
75. Woo, A. J. et al. Developmental differences in IFN
signaling affect GATA1s‑induced megakaryocyte
hyperproliferation. J. Clin. Invest. 123, 3292–3304
(2013).
76. Nikolaev, S. I. et al. Exome sequencing identifies
putative drivers of progression of transient
myeloproliferative disorder to AMKL in infants with
Down syndrome. Blood 122, 554–561 (2013).
77. Saida, S. et al. Clonal selection in xenografted TAM
recapitulates the evolutionary process of myeloid
leukemia in Down syndrome. Blood 121, 4377–4387
(2013).
78. Roberts, I. et al. GATA1‑mutant clones are frequent
and often unsuspected in babies with Down syndrome:
identification of a population at risk of leukemia.
Blood 122, 3908–3917 (2013).
79. Ford, A. M. et al. In utero rearrangements in the
trithorax-related oncogene in infant leukaemias.
Nature 363, 358–360 (1993).
80. Wiemels, J. L. et al. Prenatal origin of acute
lymphoblastic leukaemia in children. Lancet 354,
1499–1503 (1999).
81. Teuffel, O. et al. Prenatal origin of separate evolution
of leukemia in identical twins. Leukemia 18,
1624–1629 (2004).
288 | APRIL 2014 | VOLUME 14 www.nature.com/reviews/cancer
82. Taub, J. W. et al. High frequency of leukemic clones in
newborn screening blood samples of children with
B‑precursor acute lymphoblastic leukemia. Blood 99,
2992–2996 (2002).
83. Eguchi-Ishimae, M. et al. Breakage and fusion of the
TEL (ETV6) gene in immature B lymphocytes induced
by apoptogenic signals. Blood 97, 737–743 (2001).
84. McHale, C. M. et al. Prenatal origin of ETV6–
RUNX1‑positive acute lymphoblastic leukemia in
children born in California. Genes Chromosomes
Cancer 37, 36–43 (2003).
85. Olsen, M. et al. Preleukemic TEL–AML1‑positive
clones at cell level of 10(-3) to 10(-4) do not persist
into adulthood. J. Pediatr. Hematol.Oncol. 28,
734–740 (2006).
86. Hong, D. et al. Initiating and cancer-propagating cells
in TEL–AML1‑associated childhood leukemia. Science
319, 336–339 (2008).
87. Tuszuki, S. & Seto, M. TEL (ETV6)-AML1 (RUNX1)
initiates self-renewing fetal pro‑B cells in association
with a transcriptional program shared with embryonic
stem cells in mice. Stem Cells 31, 236–247 (2013).
88. Torrano, V., Procter, J., Cardus, P., Greaves, M. &
Ford, A. M. ETV6‑RUNX1 promotes survival of early B
lineage progenitor cells via a dysregulated
erythropoietin receptor. Blood 118, 4910–4918
(2011).
89. van der Weyden, L. et al. Modeling the evolution of
ETV6‑RUNX1‑induced B‑cell precursor acute
lymphoblastic leukemia in mice. Blood 118,
1041–1051 (2011).
90. Ford, A. M. et al. The TEL–AML1 leukemia fusion gene
dysregulates the TGF-β pathway in early B lineage
progenitor cells. J. Clin. Invest. 119, 826–836 (2009).
91. Cazzaniga, G. et al. Developmental origins and impact
of BCR‑ABL1 fusion and IKZF1 deletions in
monozygotic twins with Ph+ acute lymphoblastic
leukemia. Blood 118, 5559–5564 (2011).
92. Chuk, M. K., McIntyre, E., Small, D. & Brown, P.
Discordance of MLL-rearranged (MLL‑R) infant acute
lymphoblastic leukemia in monozygotic twins with
spontaneous clearance of preleukemic clone in
unaffected twin. Blood 113, 6691–6694 (2009).
93. Powell, B. C. et al. Identification of TP53 as an acute
lymphocytic leukemia susceptibility gene through
exome sequencing. Pediatr. Blood Cancer 60, E1–E3
(2013).
94. Holmfeldt, L. et al. The genomic landscape of
hypodiploid acute lymphoblastic leukemia. Nature
Genet. 45, 242–252 (2013).
95. Ito, M. Control of mental activities by internal models
in the cerebellum. Nature Rev. Neurosci. 9, 304–313
(2008).
96. Simeone, A., Acampora, D., Gulisano, M.,
Stornaiuolo, A. & Boncinelli, E. Nested expression
domains of four homeobox genes in developing rostral
brain. Nature 358, 687–690 (1992).
97. Wingate, R. J. & Hatten, M. E. The role of the rhombic
lip in avian cerebellum development. Development
126, 4395–4404 (1999).
98. Adamson, D. C. et al. OTX2 is critical for the
maintenance and progression of Shh-independent
medulloblastomas. Cancer Res. 70, 181–191 (2010).
99. Bai, R. Y., Staedtke, V., Lidov, H. G., Eberhart, C. G. &
Riggins, G. J. OTX2 represses myogenic and neuronal
differentiation in medulloblastoma cells. Cancer Res.
72, 5988–6001 (2012).
100. Bunt, J. et al. OTX2 directly activates cell cycle genes
and inhibits differentiation in medulloblastoma cells.
Int. J. Cancer 131, E21–E32 (2012).
101. Hatten, M. E. & Heintz, N. Mechanisms of neural
patterning and specification in the developing
cerebellum. Annu. Rev. Neurosci. 18, 385–408
(1995).
102. Morales, D. & Hatten, M. E. Molecular markers of
neuronal progenitors in the embryonic cerebellar
anlage. J. Neurosci. 26, 12226–12236 (2006).
103. Spassky, N. et al. Primary cilia are required for
cerebellar development and Shh-dependent expansion
of progenitor pool. Dev. Biol. 317, 246–259 (2008).
104. Chang, C. H., Chang, F. M., Yu, C. H., Ko, H. C. &
Chen, H. Y. Assessment of fetal cerebellar volume
using three-dimensional ultrasound. Ultrasound Med.
Biol. 26, 981–988 (2000).
105. Volpe, J. J. Cerebellum of the premature infant: rapidly
developing, vulnerable, clinically important. J. Child
Neurol. 24, 1085–1104 (2009).
106. Andersen, B. B., Korbo, L. & Pakkenberg, B. A
quantitative study of the human cerebellum with
unbiased stereological techniques. J. Comparative
Neurol. 326, 549–560 (1992).
© 2014 Macmillan Publishers Limited. All rights reserved
107. Dahmane, N. & Ruiz i Altaba, A. Sonic hedgehog
regulates the growth and patterning of the
cerebellum. Development 126, 3089–3100 (1999).
108. Lee, A. et al. Isolation of neural stem cells from the
postnatal cerebellum. Nature Neurosci. 8, 723–729
(2005).
109. Wang, V. Y., Rose, M. F. & Zoghbi, H. Y. Math1
expression redefines the rhombic lip derivatives and
reveals novel lineages within the brainstem and
cerebellum. Neuron 48, 31–43 (2005).
110. Raaf, J. & Kernohan, J. W. A study of the external
granular layer in the cerebellum. The disappearance of
the external granular layer and the growth of the
molecular and internal granular layers in the
cerebellum. Am. J. Anat. 75, 151–172 (1944).
111. Gailani, M. R. et al. Developmental defects in Gorlin
syndrome related to a putative tumor suppressor gene
on chromosome 9. Cell 69, 111–117 (1992).
112. Hahn, H. et al. Mutations of the human homolog of
Drosophila patched in the nevoid basal cell carcinoma
syndrome. Cell 85, 841–851 (1996).
113. Northcott, P. A. et al. Medulloblastoma comprises four
distinct molecular variants. J. Clin. Oncol. 29,
1408–1414 (2011).
114. Goodrich, L. V., Milenkovic, L., Higgins, K. M. &
Scott, M. P. Altered neural cell fates and
medulloblastoma in mouse patched mutants. Science
277, 1109–1113 (1997).
115. Brugieres, L. et al. Incomplete penetrance of the
predisposition to medulloblastoma associated with
germ-line SUFU mutations. J. Med. Genet. 47,
142–144 (2010).
116. Northcott, P. A. et al. Subgroup-specific structural
variation across 1,000 medulloblastoma genomes.
Nature 488, 49–56 (2012).
117. Thomas, W. D. et al. Patched1 deletion increases
N‑Myc protein stability as a mechanism of
medulloblastoma initiation and progression.
Oncogene 28, 1605–1615 (2009).
118. Hallahan, A. R. et al. The SmoA1 mouse model reveals
that Notch signaling is critical for the growth and
survival of sonic hedgehog-induced medulloblastomas.
Cancer Res. 64, 7794–7800 (2004).
119. Mao, J. et al. A novel somatic mouse model to survey
tumorigenic potential applied to the Hedgehog
pathway. Cancer Res. 66, 10171–10178 (2006).
120. Weiner, H. L. et al. Induction of medulloblastomas in
mice by sonic hedgehog, independent of Gli1. Cancer
Res. 62, 6385–6389 (2002).
121. Schuller, U. et al. Acquisition of granule neuron
precursor identity is a critical determinant of
progenitor cell competence to form Shh-induced
medulloblastoma. Cancer Cell 14, 123–134 (2008).
122. Yang, Z. J. et al. Medulloblastoma can be initiated by
deletion of Patched in lineage-restricted progenitors or
stem cells. Cancer Cell 14, 135–145 (2008).
123. Knoepfler, P. S., Cheng, P. F. & Eisenman, R. N. N‑myc
is essential during neurogenesis for the rapid
expansion of progenitor cell populations and the
inhibition of neuronal differentiation. Genes Dev. 16,
2699–2712 (2002).
124. Browd, S. R. et al. N‑myc can substitute for insulin-like
growth factor signaling in a mouse model of sonic
hedgehog-induced medulloblastoma. Cancer Res. 66,
2666–2672 (2006).
125. Korshunov, A. et al. Biological and clinical
heterogeneity of MYCN-amplified medulloblastoma.
Acta Neuropathol. 123, 515–527 (2012).
126. Swartling, F. J. et al. Pleiotropic role for MYCN in
medulloblastoma. Genes Dev. 24, 1059–1072
(2010).
127. Kawauchi, D. et al. A mouse model of the most
aggressive subgroup of human medulloblastoma.
Cancer Cell 21, 168–180 (2012).
128. Pei, Y. et al. An animal model of MYC-driven
medulloblastoma. Cancer Cell 21, 155–167 (2012).
129. Kool, M. et al. Molecular subgroups of
medulloblastoma: an international meta-analysis of
transcriptome, genetic aberrations, and clinical data of
WNT, SHH, Group 3, and Group 4 medulloblastomas.
Acta Neuropathol. 123, 473–484 (2012).
130. Pei, Y. et al. WNT signaling increases proliferation and
impairs differentiation of stem cells in the developing
cerebellum. Development 139, 1724–1733 (2012).
131. Kropilak, M., Jagelman, D. G., Fazio, V. W., Lavery, I. L.
& McGannon, E. Brain tumors in familial adenomatous
polyposis. Diseases Colon Rectum 32, 778–782
(1989).
132. Taylor, M. D. et al. Failure of a medulloblastoma-
derived mutant of SUFU to suppress WNT signaling.
Oncogene 23, 4577–4583 (2004).

133. Robinson, G. et al. Novel mutations target distinct
subgroups of medulloblastoma. Nature 488, 43–48
(2012).
134. Barker, N. et al. The chromatin remodelling factor
Brg‑1 interacts with β-catenin to promote target gene
activation. EMBO J. 20, 4935–4943 (2001).
135. Pugh, T. J. et al. Medulloblastoma exome sequencing
uncovers subtype-specific somatic mutations. Nature
488, 106–110 (2012).
136. Momota, H., Shih, A. H., Edgar, M. A. & Holland, E. C.
c‑Myc and β-catenin cooperate with loss of p53 to
generate multiple members of the primitive
neuroectodermal tumor family in mice. Oncogene 27,
4392–4401 (2008).
137. Gibson, P. et al. Subtypes of medulloblastoma have
distinct developmental origins. Nature 468,
1095–1099 (2010).
138. Takebayashi, H. et al. Non-overlapping expression of
Olig3 and Olig2 in the embryonic neural tube. Mech.
Dev. 113, 169–174 (2002).
139. Taylor, M. D. et al. Molecular subgroups of
medulloblastoma: the current consensus. Acta
Neuropathol. 123, 465–472 (2012).
140. Swartling, F. J. et al. Distinct neural stem cell
populations give rise to disparate brain tumors in
response to N‑MYC. Cancer Cell 21, 601–613 (2012).
141. Stehelin, D., Varmus, H. E., Bishop, J. M. & Vogt, P. K.
DNA related to the transforming gene(s) of avian
sarcoma viruses is present in normal avian DNA.
Nature 260, 170–173 (1976).
142. Siddhartha, M. The Emperor of All Maladies: A
Biography of Cancer (Scribner, 2010).
143. Knudson, A. G. Jr Mutation and cancer: statistical
study of retinoblastoma. Proc. Natl Acad. Sci. USA 68,
820–823 (1971).
144. Stephens, P. J. et al. Massive genomic rearrangement
acquired in a single catastrophic event during cancer
development. Cell 144, 27–40 (2011).
145. Forment, J. V., Kaidi, A. & Jackson, S. P.
Chromothripsis and cancer: causes and consequences
of chromosome shattering. Nat. Rev. Cancer 12,
663–670 (2012).
146. Molenaar, J. J. et al. Sequencing of neuroblastoma
identifies chromothripsis and defects in neuritogenesis
genes. Nature 483, 589–593 (2012).
147. Otto, T. et al. Stabilization of N‑Myc is a critical
function of Aurora A in human neuroblastoma. Cancer
Cell 15, 67–78 (2009).
148. Masciari, S. et al. F18‑fluorodeoxyglucose-positron
emission tomography/computed tomography
screening in Li‑Fraumeni syndrome. JAMA 299,
1315–1319 (2008).
149. Custodio, G. et al. Impact of neonatal screening and
surveillance for the TP53 R337H mutation on early
detection of childhood adrenocortical tumors. J. Clin.
Oncol. 31, 2619–2626 (2013).
150. Yoshimoto, M. et al. Embryonic day 9 yolk sac and
intra-embryonic hemogenic endothelium
independently generate a B-1 and marginal zone
progenitor lacking B-2 potential. Proc. Natl Acad. Sci.
USA 108, 1468–1473 (2011).
151. Hardy, R. R. B-1 B cell development. J. Immunol. 177,
2749–2754 (2006).
152. Montecino-Rodriguez, E. & Dorshkind, K. Formation
of B-1 B cells from neonatal B-1 transitional cells
exhibits NF‑kB redundancy. J. Immunol. 187,
5712–5719 (2011).
153. Adolfsson, J. et al. Identification of Flt3+ lympho-
myeloid stem cells lacking erythro-megakaryocytic
potential a revised road map for adult blood lineage
commitment. Cell 121, 295–306 (2005).
154. Li, Y. S., Wasserman, R., Hayakawa, K. & Hardy, R. R.
Identification of the earliest B lineage stage in
mouse bone marrow. Immunity 5, 527–535 (1996).
NATURE REVIEWS | CANCER VOLUME 14 | APRIL 2014 | 289
155. Rumfelt, L. L., Zhou, Y., Rowley, B. M., Shinton, S. A. &
Hardy, R. R. Lineage specification and plasticity in
CD19– early B cell precursors. J. Exp. Med. 203,
675–687 (2006).
156. Sitnicka, E. et al. Complementary signaling through
flt3 and interleukin‑7 receptor α is indispensable for
fetal and adult B cell genesis. J. Exp. Med. 198,
1495–1506 (2003).
157. Sitnicka, E. et al. Key role of flt3 ligand in regulation of
the common lymphoid progenitor but not in
maintenance of the hematopoietic stem cell pool.
Immunity 17, 463–472 (2002).
158. Nutt, S. L. & Kee, B. L. The transcriptional regulation
of B cell lineage commitment. Immunity 26, 715–725
(2007).
159. Kirstetter, P., Thomas, M., Dierich, A., Kastner, P. &
Chan, S. Ikaros is critical for B cell differentiation
and function. Eur. J. Immunol. 32, 720–730
(2002).
160. Wang, J. H. et al. Selective defects in the
development of the fetal and adult lymphoid system
in mice with an Ikaros null mutation. Immunity 5,
537–549 (1996).
161. Yoshida, T., Ng, S. Y., Zuniga-Pflucker, J. C. &
Georgopoulos, K. Early hematopoietic lineage
restrictions directed by Ikaros. Nature Immunol. 7,
382–391 (2006).
162. Mullighan, C. G. et al. BCR‑ABL1 lymphoblastic
leukaemia is characterized by the deletion of Ikaros.
Nature 453, 110–114 (2008).
163. Cobaleda, C., Jochum, W. & Busslinger, M. Conversion
of mature B cells into T cells by dedifferentiation to
uncommitted progenitors. Nature 449, 473–477
(2007).
164. Mullighan, C. G. et al. Genome-wide analysis of
genetic alterations in acute lymphoblastic leukaemia.
Nature 446, 758–764 (2007).
165. Shah, S. et al. A recurrent germline PAX5 mutation
confers susceptibility to pre‑B cell acute
lymphoblastic leukemia. Nature Genet. 45,
1226–1231 (2013).
166. Bain, G. et al. Both E12 and E47 allow commitment
to the B cell lineage. Immunity 6, 145–154
(1997).
167. Borghesi, L. et al. E47 is required for V(D)J
recombinase activity in common lymphoid
progenitors. J. Exp. Med. 202, 1669–1677
(2005).
168. Welinder, E. et al. The transcription factors E2A and
HEB act in concert to induce the expression of FOXO1
in the common lymphoid progenitor. Proc. Natl Acad.
Sci. USA 108, 17402–17407 (2011).
169. Kikuchi, K., Kasai, H., Watanabe, A., Lai, A. Y. &
Kondo, M. IL‑7 specifies B cell fate at the common
lymphoid progenitor to pre-proB transition stage by
maintaining early B cell factor expression. J. Immunol.
181, 383–392 (2008).
170. Seo, W., Ikawa, T., Kawamoto, H. & Taniuchi, I.
Runx1‑Cbfβ facilitates early B lymphocyte
development by regulating expression of Ebf1. J. Exp.
Med. 209, 1255–1262 (2012).
171. Growney, J. D. et al. Loss of Runx1 perturbs adult
hematopoiesis and is associated with a
myeloproliferative phenotype. Blood 106, 494–504
(2005).
172. Tsubono, Y. & Hisamichi, S. A halt to neuroblastoma
screening in Japan. New Engl. J. Med. 350,
2010–2011 (2004).
173. Breslow, N., Olshan, A., Beckwith, J. B. & Green, D. M.
Epidemiology of Wilms tumor. Med. Pediatr. Oncol.
21, 172–181 (1993).
174. Coppes, M. J., Haber, D. A. & Grundy, P. E. Genetic
events in the development of Wilms’ tumor. N. Engl.
J. Med. 331, 586–590 (1994).
© 2014 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES
175. Park, S. et al. Inactivation of WT1 in nephrogenic
rests, genetic precursors to Wilms’ tumour. Nature
Genet. 5, 363–367 (1993).
176. Chen, D. et al. Cell-specific effects of RB or RB/p107
loss on retinal development implicate an intrinsically
death-resistant cell‑of‑origin in retinoblastoma. Cancer
Cell 5, 539–551 (2004).
177. Zhang, J., Schweers, B. & Dyer, M. A. The first
knockout mouse model of retinoblastoma. Cell Cycle
3, 952–959 (2004).
178. Federico, S., Brennan, R. & Dyer, M. A. Childhood
cancer and developmental biology a crucial
partnership. Curr. Top. Dev. Biol. 94, 1–13
(2011).
179. Abramson, D. H. et al. Rapid growth of retinoblastoma
in a premature twin. Arch. Ophthalmol. 120,
1232–1233 (2002).
180. Lohmann, D. R. & Gallie, B. L. Retinoblastoma:
revisiting the model prototype of inherited cancer. Am.
J. Med. Genet. C. Semin. Med. Genet. 129C, 23–28
(2004).
181. Looijenga, L. H. & Oosterhuis, J. W. Pathogenesis of
testicular germ cell tumours. Rev. Reprod. 4, 90–100
(1999).
182. Bussey, K. J. et al. Chromosome abnormalities of
eighty-one pediatric germ cell tumors: sex-, age-, site-,
and histopathology-related differences—a Children’s
Cancer Group study. Genes Chromosomes Cancer 25,
134–146 (1999).
183. Isaacs, H. Jr. Perinatal (fetal and neonatal) germ
cell tumors. J. Pediatr. Surg. 39, 1003–1013
(2004).
184. Schindler, J. W. et al. TEL–AML1 corrupts
hematopoietic stem cells to persist in the bone
marrow and initiate leukemia. Cell Stem Cell 5,
43–53 (2009).
185. Klochendler-Yeivin, A. et al. The murine SNF5/INI1
chromatin remodeling factor is essential for embryonic
development and tumor suppression. EMBO Rep. 1,
500–506 (2000).
186. Versteege, I. et al. Truncating mutations of hSNF5/INI1
in aggressive paediatric cancer. Nature 394,
203–206 (1998).
187. Biegel, J. A. et al. Germ-line and acquired mutations of
INI1 in atypical teratoid and rhabdoid tumors. Cancer
Res. 59, 74–79 (1999).
188. Sevenet, N. et al. Constitutional mutations of the
hSNF5/INI1 gene predispose to a variety of cancers.
Am. J. Hum. Genet. 65, 1342–1348 (1999).
Acknowledgements
G.M.M., B.B.C., D.R.C. and T.L. are supported by grants from
the Australian National Health and Medical Research Council
(NHMRC), the Cancer Institute New South Wales, Australia,
the Cancer Council New South Wales, Australia, and the
Steven Walter Children’s Cancer Foundation. T.L. is an
Australian Research Council (ARC) Research Fellow. M.K.M.
is supported by a fellowship from the Kids Cancer Project.
J.G.M. and W.A.W. are supported by US National Institutes
of Health (NIH) grants F30CA174154, U01CA176287,
R 01 CA 1 3 3 0 91 , R 01 CA 10 2 3 21 , R 01 CA 1 4 8 6 9 9 ,
R01CA159859, U54CA163155 and P01CA081403, as well
as the Katie Dougherty, Pediatric Brain Tumor and Samuel G.
Waxman Foundations. Sydney Children’s Hospital, Randwick,
New South Wales, Australia, and the Children’s Cancer
Institute Australia for Medical Research, Randwick, New
South Wales, Australia, are affiliated with the University of
New South Wales, Sydney, Australia. The authors apologize
to the numerous colleagues whose important contributions
could not be included in this Opinion article owing to space
limitations.
Competing interests statement
The authors declare no competing interests.