Summary:

The key differences between tumors arising in children and those in adults stem from the cellular origin of cancer at different ages, with adult cancers arising within aging cell hierarchies, as a consequence of accumulated damage and mutagenesis, in contrast to childhood tumors that are born in aberrantly developing tissues. A distinct biological property of childhood tumor cells—a block of developmental maturation—may hold the key to advancing the treatment of childhood cancer beyond cytotoxic strategies.

Childhood cancer is rare, yet remains the leading cause of death from disease in children in high-income nations. Although the prognosis of some tumor types has improved impressively, in others it has stalled despite decades of research efforts. For example, the once universally fatal diseases standard-risk acute lymphoblastic leukemia or Wilms tumor, a kidney cancer, can now be cured in more than 90% of children. By contrast, a different type of cancer that mostly arises in the kidney, malignant rhabdoid tumor, can be cured in only ∼40% of children with localized disease and is usually fatal in metastatic cases. The brain tumor diffuse intrinsic pontine glioma is an example of a universally fatal childhood cancer. Beyond mortality, treatment often generates considerable morbidity. Late effects from cancer treatment in childhood include organ failure, neurodisability, subfertility, other endocrinopathies, and second malignancies.

The considerable improvements in cure rates have been achieved through clinical trials that have optimized the use of conventional multimodal treatment encompassing surgery, radiotherapy, and cytotoxic chemotherapy. Given small patient cohorts for individual tumor entities, trials in pediatric oncology are typically conducted through national or international consortia. It would seem unlikely, however, that further substantial progress can be made through modification of conventional treatment schedules. In other words, variations of “more of the same” are unlikely to work.

Novel approaches, beyond conventional cell killing, are required to achieve the next leap in cure rates. These may include immunotherapy and mutation-guided treatment, which can be adopted from adult practice. In addition, one may consider ways to target unique characteristics of childhood cancer cells.

Although it is difficult to make general statements about diverse spectra of tumors, there are some unique features of childhood cancers (Fig. 1A and B). Adult cancers are mostly epithelial and typically arise as a consequence of aging and/or exposure to mutagens. Childhood cancer, by contrast, is rarely epithelial and mainly derived from mesodermal and ectodermal lineages. When the same cancer type occurs in children and adults, there are usually marked differences in terms of tumor biology and curability. For example, the spectrum of leukemia phenotypes varies across childhood, adolescence, and adulthood, associated with differences in outcome, somatic changes, and their relative significance (1).

Systematic surveys of the genetic changes of cancer have revealed specific genomic features of childhood tumors (2). One observation has been a substantial (∼10%) prevalence of pathogenic germline variants, seeding the substrate for early cancer development. The burden of somatic mutations is generally relatively low in childhood tumors, thus generating fewer potentially immunogenic neoantigens. Multiple driver events rarely occur in childhood cancer, with many examples of tumors driven by just one somatic variant. A most fundamental feature of childhood neoplasms, however, is the long-held view that their cellular origin is rooted in development. This developmental origin of childhood cancer may manifest as a specific biological property, i.e., a maturation block, which may lend itself as a target for novel treatments.

Several lines of evidence indicate that childhood cancer arises during development. Most tumor types are specific to childhood and rarely, or never, occur in adults, indicating that the cell of origin of pediatric cancers is absent in mature tissues. Furthermore, childhood tumors are rare and, as aforementioned, are typically driven by few driver events, indicating there is a limited biological window for tumor formation, alluding to the transiency of the cell of origin. When a tumor type does occur in both adults and children, it tends to exhibit unique features in children. For instance, high-grade gliomas in children are mostly driven by somatic histone 3.3 or 3.1 mutations that are rarely seen in their adult counterparts. Thus, it would appear that the oncogenic advantage that these histone mutations confer is confined to a specific period of cellular life, before brain tissue has fully matured. Another example would be testicular germ cell tumors, which vary greatly in pre- and post-pubertal cases with regard to phenotype, malignant potential, and underlying somatic genotype (3).

The incidence of many childhood tumors is tightly coupled to confined developmental periods. The early neonatal period shows a predilection for NTRK3-driven soft-tissue tumors and for spontaneously regressing neuroblastoma. Lymphoblastic leukemia driven by rearrangement of the MLL family of genes mostly occurs in infancy. Likewise, certain brain and soft-tissue tumors, such as sonic hedgehog medulloblastoma, atypical teratoid/rhabdoid tumors and extracranial malignant rhabdoid tumors as well as embryonal tumors with multilayered rosettes, typically occur within the first years of life. Wilms tumor is an example of a tumor that particularly affects toddlers and rarely occurs after the age of five years. Adolescence and the pubertal growth spurt are associated with osteosarcoma and Ewing sarcoma.

The morphologic overlap between some childhood tumors and embryonal tissues points to a fetal origin of tumors, such as the histologic overlap between hepatoblastoma and the developing liver. For some tumors, specific fetal cellular correlates have been identified. Examples include fetal hepatic megakaryocyte precursors that cause transient leukemia of Down syndrome through somatic acquisition of GATA1 mutations (4). A major piece of evidence suggesting a fetal origin for childhood lymphoblastic leukemia was the discovery that in some cases leukomogenic fusions could be detected in peripheral blood samples obtained at birth, several years prior to presentation (5).

Biological readouts of cell states, including transcription, enhancer states, and DNA methylation, substantiate the view that childhood cancer cells represent “arrested” development. Methylation signatures of childhood brain tumors have evolved over recent years as a highly specific and sensitive classification tool (6). Multiple lines of evidence suggest that methylation signatures encode the developmental state in which cancer cells have arrested. Similarly, the enhancer landscape of some childhood brain tumors and neuroblastoma has been linked to specific fetal developmental cell states (7–10).

A further pertinent observation pointing toward a maturation block in childhood tumors is that some tumors are capable of differentiating into harmless tissue, either spontaneously or following treatment with cytotoxic chemotherapy or maturation agents. The prototypical tumor type in this regard is neuroblastoma (11), which in infants frequently regresses spontaneously, even in the case of metastatic disease. Maturation of neuroblastoma can also be observed in resection specimens of high-risk tumors following an intense induction regimen of cytotoxic chemotherapy. At resection, previously undifferentiated, highly malignant cells are often replaced by differentiated, nonmalignant neuronal tissue. In addition, the vitamin A derivative 13-cis retinoic acid, thought to function as a maturation agent, has been shown to improve survival in neuroblastoma and is thus included in most treatment schedules of high-risk tumors. Another retinoid, all-trans retinoic acid (ATRA), is used to drive differentiation of malignant promyeloblasts in acute promyelocytic leukemia, which is caused by an oncogenic fusion of the retinoic acid receptor gene RARA. Treatment with ATRA is thought to overcome the maturation block maintained by the RARA fusion.

More recently, single-cell mRNA profiling of childhood cancer cells and relevant normal tissues has correlated cancer cells with specific developmental cell types. For example, cells of the childhood kidney cancer Wilms tumor were shown to correspond to cells of the developing nephron (12). In certain childhood brain tumors, single-cell sequencing of hindbrain cells has corroborated and further refined cellular hierarchies (13, 14) that had previously been unraveled by cross-species genomics and transgenic mouse modeling in the developing brain (15, 16).

A key challenge in targeting maturation lies in replacing the approximate notion of developmental arrest with data, with exact quantitative readouts of cell states and developmental trajectories. This may now be achievable in the context of the rise of single-cell transcriptomics and epigenetics, as a surrogate of cellular identity and function. The Human Cell Atlas and related initiatives aim to characterize by single-cell mRNA sequencing all human cell types, including fetal cells. Such data will provide the quantitative reference that enables precise cancer-to-normal-cell comparisons.

We may expect such analyses to point at fetal cells or developmental processes that childhood cancer cells represent (Fig. 1C). This may in turn reveal therapeutic vulnerabilities, be it a modifiable transcriptional state or specific fetal transcripts as targets for immunotherapy or small molecule–based interference. A precedent in this regard is the fetal antigen GD2, which is expressed by neuroblastoma, among other tumor types, yet by very few mature tissues. Antibodies against GD2 are routinely used in the treatment of high-risk neuroblastoma. Our enthusiasm is perhaps dampened by the possibility that the fetalness of childhood cancer may confer unprecedented plasticity, enabling cells to differentiate in unforeseen ways to evade therapeutic efforts.

Childhood cancer cells have distinct properties that we cannot target by simply adopting strategies from adult practice. The rise of quantitative descriptive single-cell technologies provides an opportunity to tease out aberrant embryology that underpins pediatric tumors. This knowledge may form the basis to develop novel treatments in order to achieve the next leap in cure rates of childhood cancer. Such efforts will require dedicated funding to support the development of economically unviable drugs, perhaps provided by private sources, which is a long-standing tradition in childhood cancer research.

S. Behjati reports grants from Wellcome Trust, the Little Princess Trust, and St. Baldrick's Foundation during the conduct of the study. S.M. Pfister reports grants from IMI-2 project together with various companies outside the submitted work. No disclosures were reported by the other author.

S. Behjati acknowledges funding from Wellcome, the St. Baldrick's Foundation, and the Little Princess Trust, R.J. Gilbertson acknowledges funding from CRUK, and S.M. Pfister acknowledges the BRAIN-MATCH grant from the ERC.

1.
Tarlock
K
,
Zhong
S
,
He
Y
,
Ries
R
,
Severson
E
,
Bailey
M
, et al
Distinct age-associated molecular profiles in acute myeloid leukemia defined by comprehensive clinical genomic profiling
.
Oncotarget
2018
;
9
:
26417
30
.
2.
Grobner
SN
,
Worst
BC
,
Weischenfeldt
J
,
Buchhalter
I
,
Kleinheinz
K
,
Rudneva
VA
, et al
The landscape of genomic alterations across childhood cancers
.
Nature
2018
;
555
:
321
7
.
3.
Ulbright
TM
. 
Germ cell tumors of the gonads: a selective review emphasizing problems in differential diagnosis, newly appreciated, and controversial issues
.
Mod Pathol
2005
;
18
(
Suppl 2
):
S61
79
.
4.
Roberts
I
,
Izraeli
S
. 
Haematopoietic development and leukaemia in Down syndrome
.
Br J Haematol
2014
;
167
:
587
99
.
5.
Greaves
M
. 
A causal mechanism for childhood acute lymphoblastic leukaemia
.
Nat Rev Cancer
2018
;
18
:
471
84
.
6.
Capper
D
,
Jones
DTW
,
Sill
M
,
Hovestadt
V
,
Schrimpf
D
,
Sturm
D
, et al
DNA methylation-based classification of central nervous system tumours
.
Nature
2018
;
555
:
469
74
.
7.
Durbin
AD
,
Zimmerman
MW
,
Dharia
NV
,
Abraham
BJ
,
Iniguez
AB
,
Weichert-Leahey
N
, et al
Selective gene dependencies in MYCN-amplified neuroblastoma include the core transcriptional regulatory circuitry
.
Nat Genet
2018
;
50
:
1240
6
.
8.
Erkek
S
,
Johann
PD
,
Finetti
MA
,
Drosos
Y
,
Chou
HC
,
Zapatka
M
, et al
Comprehensive analysis of chromatin states in atypical teratoid/rhabdoid tumor identifies diverging roles for SWI/SNF and polycomb in gene regulation
.
Cancer Cell
2019
;
35
:
95
110
e8
.
9.
Lin
CY
,
Erkek
S
,
Tong
Y
,
Yin
L
,
Federation
AJ
,
Zapatka
M
, et al
Active medulloblastoma enhancers reveal subgroup-specific cellular origins
.
Nature
2016
;
530
:
57
62
.
10.
Mack
SC
,
Pajtler
KW
,
Chavez
L
,
Okonechnikov
K
,
Bertrand
KC
,
Wang
X
, et al
Therapeutic targeting of ependymoma as informed by oncogenic enhancer profiling
.
Nature
2018
;
553
:
101
5
.
11.
Brodeur
GM
,
Bagatell
R
. 
Mechanisms of neuroblastoma regression
.
Nat Rev Clin Oncol
2014
;
11
:
704
13
.
12.
Young
MD
,
Mitchell
TJ
,
Vieira Braga
FA
,
Tran
MGB
,
Stewart
BJ
,
Ferdinand
JR
, et al
Single-cell transcriptomes from human kidneys reveal the cellular identity of renal tumors
.
Science
2018
;
361
:
594
9
.
13.
Filbin
M
,
Monje
M
. 
Developmental origins and emerging therapeutic opportunities for childhood cancer
.
Nat Med
2019
;
25
:
367
76
.
14.
Vladoiu
MC
,
El-Hamamy
I
,
Donovan
LK
,
Farooq
H
,
Holgado
BL
,
Sundaravadanam
Y
, et al
Childhood cerebellar tumours mirror conserved fetal transcriptional programs
.
Nature
2019
;
572
:
67
73
.
15.
Gibson
P
,
Tong
Y
,
Robinson
G
,
Thompson
MC
,
Currle
DS
,
Eden
C
, et al
Subtypes of medulloblastoma have distinct developmental origins
.
Nature
2010
;
468
:
1095
9
.
16.
Johnson
RA
,
Wright
KD
,
Poppleton
H
,
Mohankumar
KM
,
Finkelstein
D
,
Pounds
SB
, et al
Cross-species genomics matches driver mutations and cell compartments to model ependymoma
.
Nature
2010
;
466
:
632
6
.