The new engl and jour nal of medicine
n engl j med 373;24 nejm.org  December 10, 20152336
From the Departments of Computational
Biology (J.Z., G.W., M.N.E., D.H., X.M.,
X.Z., M.R.W., X.C., M.R., A.P., J.B.B.,
S.W.), Oncology (M.F.W., T.A.G., R.B.M.,
S.H.-D., R.N., E.Q., A.G., A.S.P., C.-H.P.,
K.E.N.), and Pathology (T.A.G., M.R.W.,
S.A.S., D.W.E., C.-H.P., J.R.D.) and the
Pediatric Cancer Genome Project (J.Z.,
M.F.W., G.W., M.N.E., T.A.G., J.E., X.M.,
D.A.Y., B.V., X.C., R.B.M., S.H.-D., R.N.,
E.Q., S.A.S., M.R., A.P., J.B.B., S.W.,
M.S.W., A.G., D.W.E., A.S.P., C.-H.P.,
K.E.N., J.R.D.), St. Jude Children’s Research
Hospital, Memphis, TN; and the Department
of Genetics and McDonnell Genome
Institute, Washington Univer­sity School
of Medicine in St. Louis, St. Louis (L.D.,
E.R.M., R.K.W.). Address reprint requests
to Dr. Downing at the Department of Pathology,
St. Jude Children’s Research
Hospital, 262 Danny Thomas Pl., Memphis,
TN 38105, or at ­james​.­downing@​
­stjude​.­org.
Drs. Zhang, Walsh, Wu, and Nichols contributed
equally to this article.
This article was published on November 18,
2015, at NEJM.org.
N Engl J Med 2015;373:2336-46.
DOI: 10.1056/NEJMoa1508054
Copyright © 2015 Massachusetts Medical Society.
BACKGROUND
The prevalence and spectrum of predisposing mutations among children and adolescents
with cancer are largely unknown. Knowledge of such mutations may improve the
understanding of tumorigenesis, direct patient care, and enable genetic counseling of
patients and families.
METHODS
In 1120 patients younger than 20 years of age, we sequenced the whole genomes (in
595 patients), whole exomes (in 456), or both (in 69). We analyzed the DNA sequences
of 565 genes, including 60 that have been associated with autosomal dominant cancerpredisposition
syndromes, for the presence of germline mutations. The pathogenicity
of the mutations was determined by a panel of medical experts with the use of cancerspecific
and locus-specific genetic databases, the medical literature, computational
predictions, and second hits identified in the tumor genome. The same approach was
used to analyze data from 966 persons who did not have known cancer in the 1000
Genomes Project, and a similar approach was used to analyze data from an autism
study (from 515 persons with autism and 208 persons without autism).
RESULTS
Mutations that were deemed to be pathogenic or probably pathogenic were identified in 95
patients with cancer (8.5%), as compared with 1.1% of the persons in the 1000 Genomes
Project and 0.6% of the participants in the autism study. The most commonly mutated genes
in the affected patients were TP53 (in 50 patients), APC (in 6), BRCA2 (in 6), NF1 (in 4), PMS2
(in 4), RB1 (in 3), and RUNX1 (in 3). A total of 18 additional patients had protein-truncating
mutations in tumor-suppressor genes. Of the 58 patients with a predisposing mutation
and available information on family history, 23 (40%) had a family history of cancer.
CONCLUSIONS
Germline mutations in cancer-predisposing genes were identified in 8.5% of the children
and adolescents with cancer. Family history did not predict the presence of an
underlying predisposition syndrome in most patients. (Funded by the American Lebanese
Syrian Associated Charities and the National Cancer Institute.)
ABSTR ACT
Germline Mutations in Predisposition Genes
in Pediatric Cancer
Jinghui Zhang, Ph.D., Michael F. Walsh, M.D., Gang Wu, Ph.D.,
Michael N. Edmonson, B.A., Tanja A. Gruber, M.D., Ph.D., John Easton, Ph.D.,
Dale Hedges, Ph.D., Xiaotu Ma, Ph.D., Xin Zhou, Ph.D.,
Donald A. Yergeau, Ph.D., Mark R. Wilkinson, B.S., Bhavin Vadodaria, B.A.,
Xiang Chen, Ph.D., Rose B. McGee, M.S., Stacy Hines‑Dowell, D.N.P.,
Regina Nuccio, M.S., Emily Quinn, M.S., Sheila A. Shurtleff, Ph.D.,
Michael Rusch, B.A., Aman Patel, M.S., Jared B. Becksfort, M.S.,
Shuoguo Wang, Ph.D., Meaghann S. Weaver, M.D., Li Ding, Ph.D.,
Elaine R. Mardis, Ph.D., Richard K. Wilson, Ph.D., Amar Gajjar, M.D.,
David W. Ellison, M.D., Ph.D., Alberto S. Pappo, M.D., Ching‑Hon Pui, M.D.,
Kim E. Nichols, M.D., and James R. Downing, M.D.​​
Original Article
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 2015 2337
Germline Mutations and Pediatric Cancer
T
he frequency of germline mutations
in cancer-predisposition genes in
children and adolescents with cancer and
the implications of such mutations are largely
unknown. Previous studies have relied mainly on
candidate-gene approaches, which are, by design,
limited. To better determine the contribution of
germline predisposition mutations to childhood
cancer, we used next-generation sequencing, including
whole-genome and whole-exome sequencing,
to analyze the genomes of 1120 children
and adolescents with cancer. We describe the
prevalence and spectrum of germline variants
among 565 cancer-associated genes, with an emphasis
on the analysis of 60 genes that have been
associated with autosomal dominant cancerpredisposition
syndromes. We also reviewed records
of patients with mutations and those without
mutations in these 60 genes for information
on family history of cancer.
Methods
Enrollment of the Patients
The 1120 patients included in this study represented
the major subtypes of pediatric cancer
(Fig. 1; and Table S1 in Supplementary Appendix
1, available with the full text of this article
at NEJM.org). Whole-genome, whole-exome, or
both types of sequencing data were generated
with the use of germline DNA for 595, 456, and
69 patients, respectively, as part of the St. Jude–
Washington University Pediatric Cancer Genome
Project (PCGP; www​.­ebi​.­ac​.­uk/​­ega/​­search/​­site/​
­PCGP). To verify predictions of aberrant splicing
caused by variants affecting splice junctions, we
sequenced the RNA transcripts extracted from
522 samples of tumor tissue obtained from 522
patients. The study was approved by the institutional
review board at St. Jude Children’s Research
Hospital. Written informed consent was
provided by a parent or guardian of each child
or by a patient who was 18 years of age or older.
Whole-exome sequencing data from two control
cohorts of persons without known cancer
were analyzed. The first data set, a raw wholeexome
sequencing data set from 966 unrelated
adults who were part of the 1000 Genomes
Project (http://ftp​.­1000genomes​.­ebi​.­ac​.­uk/​­vol1/​­ftp/​
­phase3), was analyzed by the same approach
that was used in our pediatric cancer cohort.
The second data set consisted of genotype files
of 515 persons with autism and 208 persons
without autism (median age, 6 years; range, 1 to
37) from the National Database for Autism Research
(https:/​­/​­ndar​.­nih​.­gov/​­study​.­html?id=307).
Analyses of this second data set did not involve
variant detection owing to a lack of access to
raw sequence data, and we excluded two cancerpredisposition
genes, NF1 and PTEN, which are
known to be associated with an autism spectrum
phenotype (Supplementary Appendix 1).
Figure 1. Frequency of Pediatric Cancer Types among Patients Younger than
20 Years of Age.
Panel A shows the distribution of pediatric cancer types on the basis of data
from the Surveillance, Epidemiology, and End Results (SEER) program.
Panel B shows the distribution of cancer types analyzed by the Pediatric
Cancer Genome Project (PCGP). ACT denotes adrenocortical tumor, CNS
central nervous system, and STS soft-tissue sarcoma.
B PCGP Cohort
A SEER Program
ACT, 1.7%
ACT, 3.5%
Other,
5.8%
Leukemia,
24.9%
Leukemia,
52.5%
Hodgkin’s lymphoma,
7.6%
Non-Hodgkin’s
lymphoma,
6.5%
CNS
tumors,
17.6%
CNS
tumors,
21.9%
Neuroblastoma, 5.0%
Neuroblastoma,
8.9%
Retinoblastoma, 2.1%
Retinoblastoma, 1.3%
Wilms’ tumor, 3.9%
Hepatoblastoma, 0.9%
Osteosarcoma, 2.9%
Osteosarcoma, 3.5%
Ewing’s sarcoma, 1.7%
Ewing’s sarcoma, 4.1%
Rhabdomyosarcoma, 2.9%
Rhabdomyosarcoma, 3.8%
Non-rhabdomyosarcoma
STS, 4.2%
Germ-cell tumor, 6.1%
Thyroid cancer, 3.2%
Melanoma, 3.0%
Melanoma, 0.4%
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 20152338
The new engl and jour nal of medicine
Cancer-Predisposition Genes Selected
for Analysis
A total of 565 genes were chosen for analysis on
the basis of review of the American College of
Medical Genetics and Genomics (ACMG) gene
list and the medical literature1-4
and were divided
into five nonoverlapping categories (Fig. 2; and
Table S3 in Supplementary Appendix 1 and Table
S2 in Supplementary Appendix 2, available at
NEJM.org). The first category included genes
that have been associated with autosomal dominant
cancer-predisposition syndromes, and it consisted
of 49 classical genes (including 23 genes
from the ACMG gene list5
) and 11 genes that
have been implicated in genetic syndromes associated
with RAS mutations (sometimes called
RASopathies; these include the cardiofaciocutaneous
syndrome, Costello’s syndrome [also called
the faciocutaneoskeletal syndrome], Noonan’s
syndrome, and the multiple lentigines syndrome).
These 60 genes were selected because of the
potential effect of germline mutations on clinical
practice, including avoidance of radiation
therapy, choice of surgical approach for tumor
resection, donor testing and selection for hematopoietic
stem-cell transplantation, possible or
proven benefits of tumor surveillance and early
cancer detection, and risk-reductive surgery.3
Variants
detected in these 60 genes were confirmed
experimentally by an independent sequencing
assay (Supplementary Appendix 1).
The second category included 29 genes that
have been associated with autosomal recessive
cancer-predisposition syndromes, with a focus
on identifying biallelic pathogenic mutations.
Variants detected in the 89 genes that have been
associated with autosomal dominant or autosomal
recessive cancer-predisposition syndromes
were reviewed by a multidisciplinary panel for
classification and reporting.5-8
An additional 476 genes were chosen for
evaluation on the basis of their recurrent somatic
mutation in cancer (http://cancer​.­sanger​
.­ac​.­uk/​­cancergenome/​­projects/​­census and www​
.­pediatriccancergenomeproject​.­org/​­site). These
genes were classified into three categories: tumorsuppressor
genes (58 genes),9
tyrosine kinase
genes (23), and other cancer genes (395). Our
analyses of the genes in these three categories
focused on known hotspot-activating mutations
in genes encoding kinases and on truncation
mutations in genes encoding tumor-suppressor
proteins and in other cancer genes.
Data Analysis
The sequencing data were analyzed for the presence
of single-nucleotide variants and small insertions
and deletions10
and for evidence of
germline mosaicism (Supplementary Appendix 1).
Germline copy-number variations and structural
variations were identified with the use of the
Copy Number Segmentation by Regression Tree
in Next Generation Sequencing (CONSERTING)11
and Clipping Reveals Structure (CREST)12
algoFigure
2. Categories of the 565 Cancer Genes Analyzed for Germline Mutations.
The number of genes in each category is shown in parentheses. Genes that have overlapping categories are listed
only once. Gene names in the other categories are shown in Figure S9 in Supplementary Appendix 1. RASopathies
are genetic syndromes that include the cardiofaciocutaneous syndrome, Costello’s syndrome (also called the faciocutaneoskeletal
syndrome), Noonan’s syndrome, and the multiple lentigines syndrome.
Cancer Predisposition RASopathy
BRAF
CBL
HRAS
KRAS
MAP2K1
MAP2K2
NRAS
PTPN11
RAF1
SHOC2
SOS1
ALK
APC
BAP1
BMPR1A
BRCA1
BRCA2
CDC73
CDH1
CDK4
CDKN1C
CDKN2A
CEBPA
DICER1
EPCAM
FH
GATA2
MAX
MEN1
MLH1
MSH2
MSH6
NF1
NF2
PALB2
PAX5
PHOX2B
PMS2
PRKAR1A
PTCH1
PTEN
RB1
RET
RUNX1
SDHA
SDHAF2
SDHB
SDHC
SDHD
SMAD4
SMARCA4
SMARCB1
STK11
SUFU
TMEM127
TP53
TSC1
TSC2
VHL
WT1
Total
(565)
Autosomal recessive
cancer-predisposition
genes (29)
Autosomal dominant
cancer-predisposition
genes (60)
Tumor-suppressor
genes (58)
Tyrosine
kinase genes
(23)
Other
cancer genes
(395)
=++ ++
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 2015 2339
Germline Mutations and Pediatric Cancer
rithms. Common germline structural variations
and structural variations that did not affect coding
exons were excluded from the analysis.
Nonsilent coding variants that passed qualitycontrol
and minor-allele population frequency
checks were classified as pathogenic, probably
pathogenic, of uncertain significance, probably
benign, or benign. Classification criteria included
information from curated databases, computational
predictions of mutational effect on protein
function, and recent ACMG guidelines for
interpretation.13
Full details of the data analysis
and interpretation are provided in Figure S1 in
Supplementary Appendix 1.
Results
Characteristics of the Patients
The PCGP cohort included 588 children and adolescents
with leukemia (52.5%), 245 with central
nervous system (CNS) tumors (21.9%), and 287
with non-CNS solid tumors (25.6%) (Fig. 1, and
Table S1 in Supplementary Appendix 1). The
median age of the patients was 6.9 years (range,
8 days to 19.7 years). The cancers that were selected
for sequencing included those that have
been associated with a poor clinical outcome
(e.g., hypodiploid leukemia)14
and those without
a clearly defined oncogenic cause (e.g., diffuse
intrinsic pontine glioma).15
Our cohort included
more patients with leukemia and adrenocortical
tumors than was expected on the basis of the
population in the Surveillance, Epidemiology, and
End Results (SEER) program (http://seer​.­cancer​
.­gov/​­iccc) (Fig. 1). Lymphoma, Wilms’ tumor,
germ-cell tumors, non-rhabdomyosarcoma softtissue
sarcoma, and hepatoblastoma were not
included because of an inadequate number of
samples for high-risk subtypes.
Germline Mutations in Cancer-Predisposition
Genes
In the 60 genes that have been associated with
autosomal dominant cancer-predisposition syndromes,
we identified 633 nonsilent germline
variants, of which 78 (12%) were deemed to be
pathogenic, 17 (3%) probably pathogenic, 226
(36%) of uncertain significance, 273 (43%) probably
benign, and 39 (6%) benign (Table S4A in
Supplementary Appendix 2). The 95 variants that
were deemed to be pathogenic or probably pathogenic
included 54 missense mutations, 14 nonsense
mutations, 12 frameshift mutations, 9 splicesite
mutations, and 1 in-frame deletion, as well
as 5 copy-number alterations (Fig. S2 in Supplementary
Appendix 1).
The 95 variants that were deemed to be pathogenic
or probably pathogenic were detected in 21
of the 60 genes in 94 patients (Fig. 3A, and Fig.
S3 in Supplementary Appendix 1). TP53 was most
commonly involved (in 50 patients), followed by
APC (in 6), BRCA2 (in 6), NF1 (in 4), PMS2 (in 4),
RB1 (in 3), and RUNX1 (in 3). One patient (Patient
HGG111) with café au lait spots and a high-grade
glioma had 2 distinct PMS2 truncation mutations,
which indicated a diagnosis of biallelic
mismatch-repair deficiency that was corroborated
by the somatic hypermutation observed in
the genome of the high-grade glioma.15
The most
common cancer types that were associated with
germline TP53 mutations included adrenocortical
tumors (in 27 of 39 patients [69%]), hypodiploid
acute lymphoblastic leukemia (in 9 of 47
[19%]), and choroid plexus carcinoma (in 1 of 4
[25%]) — findings that were consistent with
those in previous reports.14
As anticipated, the
tumors from all 37 of these patients had a loss
of heterozygosity at the TP53 locus (Table S4 in
Supplementary Appendix 2), including 1 patient
who had a germline deletion of 8.7 kb that removed
TP53 exons 2 through 5 (Fig. S2 in Supplementary
Appendix 1).
Four germline mutations were mosaic, with
the detected level of the mutant allele less than
a single copy (mutant allele fraction, 0.08 to 0.30).
One patient with retinoblastoma had a mosaic
RB1 mutation, and three patients with hypodiploid
acute lymphoblastic leukemia had a mosaic
TP53 mutation (Table S4A in Supplementary Appendix
2). The mutant allele fraction in matching
tumor specimens ranged from 0.76 to 0.90,
a finding that is consistent with the presence of a
second hit within the tumors. Validation by means
of deep sequencing at more than 2000× coverage
verified the mutant allele fraction within the
germline and tumor samples in each patient
(Fig. 4).
In the first control data set, from the 1000
Genomes Project, we identified 11 pathogenic or
probably pathogenic mutations in the 60 genes
that have been associated with autosomal dominant
cancer-predisposition syndromes; mutations
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 20152340
The new engl and jour nal of medicine
26/588
20/245
4/100
5/46
3/43
7/39
27/39
2/15
TP53
APC
BRCA2
N
F1
PM
S2
RB1
RUN
X1
KRAS
N
F2
RET
VH
L
ALK
BRCA1
CD
H
1
M
SH
2
M
SH
6
N
RAS
PALB2
PTCH
1
SD
H
A
SD
H
B
No.ofPatients
50
40
45
35
30
20
15
5
25
10
0
C Mutations in Cancer-Susceptibility Genes
B Mutation Frequency in 21 Genes, According to Cancer Subtype
A Mutations in 21 Genes Associated with Autosomal Dominant Cancer-Predisposition Syndromes
Patients(%)
80
40
60
20
0
No.ofPatients
100
80
90
70
60
40
30
10
50
20
0
D Mutations in Other Gene Categories
No.ofPatients
100
80
90
70
60
40
30
10
50
20
0
RetinoblastomaCNS tumorLeukemia
Osteosarcoma Ewing’s sarcomaRhabdomyosarcoma
ACT
Neuroblastoma
Leukemia
Autosomal
Dominant
Autosomal
Recessive
Tumor-
Suppressor
Gene
Tyrosine Kinase
Gene
Other Cancer
Gene
CNS
Tumor
Neuroblastoma Ewing’s
Sarcoma
Rhabdo-
myosarcoma
Osteosarcoma Retino-
blastoma
ACT
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 2015 2341
Germline Mutations and Pediatric Cancer
were found in APC (in one person), BRCA1 (in
one), BRCA2 (in four), MSH6 (in one), SDHA (in
one), SDHB (in one), and TP53 (in two) (Table S6
in Supplementary Appendix 1). The prevalence of
mutations was 1.1%, which was significantly
lower than the 8.4% prevalence observed in the
PCGP cohort (P = 5.9 × 10−16
by Fisher’s exact test).
A similar trend was observed in the second control
set, which involved participants from the
autism study (frequency, 0.6%; P = 7.4 × 10−16
for
the comparison with the PCGP cohort) (Table S7
in Supplementary Appendix 1).
The PCGP cohort included a greater-thanexpected
proportion of patients with hypodiploid
acute lymphoblastic leukemia and those with
adrenocortical tumors (Fig. 1). However, after
these two subtypes were excluded, the prevalence
of germline mutations of 5.6% was still significantly
higher than the prevalence in the two
control cohorts (P<10−7
by Fisher’s exact test for
both comparisons).
In our analysis of 29 autosomal recessive
cancer-predisposition genes, we observed only one
instance of biallelic pathogenic mutations in 1 patient
(Table S8 in Supplementary Appendix 1).
Combining data from this single patient, who
had ataxia telangiectasia caused by biallelic mutations
in ATM (Fig. S4 in Supplementary Appendix
1), with data from the 94 patients who had
pathogenic mutations in the 60 autosomal dominant
cancer-predisposition genes, we observed
an 8.5% prevalence (95 of 1120 patients) of germline
mutations that were pathogenic or probably
pathogenic in the sample we analyzed. A total of
61 of the 93 patients (66%) with monoallelic
germline mutations had a second hit within the
tumor genome (Table S4 in Supplementary Appendix
2), as shown by loss of heterozygosity (in
57 patients) or mutational inactivation of the
second allele (in 4). These data are available on
our pediatric cancer data portal (http://pecan​
.­stjude​.­org) (Figs. S5 and S6 in Supplementary
Appendix 1)
Prevalence of Germline Mutations across
Tumor Types
The prevalence of germline mutations that were
pathogenic or probably pathogenic was greatest
among patients with non-CNS solid tumors (48
of 287 patients [16.7%]), followed by those with
CNS tumors (21 of 245 [8.6%], including the
patient with biallelic loss of ATM) or leukemia
(26 of 588 [4.4%]) (Fig. 3B). The prevalence of
germline mutations varied among patients with
different subtypes of non-CNS solid tumors,
such as adrenocortical tumor (69.2%), osteosarcoma
(17.9%), retinoblastoma (13.3%), Ewing’s
sarcoma (10.9%), rhabdomyosarcoma (7.0%), and
neuroblastoma (4.0%) (Fig. 3B). The histologic
subtypes of CNS tumor that were most often associated
with germline mutations included choroid
plexus carcinoma (in 1 of 4 patients [25%]),
medulloblastoma (in 5 of 37 [13.5%]), high-grade
glioma (in 9 of 99 [9.1%]), low-grade glioma (in
3 of 38 [7.9%]), and ependymoma (in 4 of 67
[6.0%]). Overall, patients with leukemia had the
lowest prevalence of germline mutations (4.4%),
despite the inclusion of patients with hypodiploid
acute lymphoblastic leukemia, a subtype with
a high frequency of germline mutation.14
Correlation between Germline Genotype
and Tumor Phenotype
The correlation of patient genotype with tumor
phenotype revealed several known associations
as well as some new ones. The known associaFigure
3 (facing page). Distribution of Germline Mutations
in Different Gene Categories and Cancer Subtypes.
Panels A and B include only mutations that were deemed
to be pathogenic or probably pathogenic and that affect
genes that have been associated with autosomal dominant
cancer-predisposition syndromes, according to
tumor subtype. Panel A shows the distribution of mutations
in each gene among patients with various cancers
included in the PCGP cohort. Panel B shows the
prevalence of the mutations in each cancer subtype.
Five patients with melanoma without mutations are
not shown, and one patient (HGG027) who had a CNS
tumor with biallelic mutation in an autosomal recessive
gene (ATM) is not included in the summary. Panel C
shows the number of patients who had germline mutations
considered to be pathogenic or probably pathogenic
in genes that have been associated with autosomal
dominant (60 genes) and autosomal recessive (29) cancer
susceptibility, according to cancer subtype. Panel D
shows the total number of patients who had truncation
mutations in tumor-suppressor genes, tyrosine kinase
genes, and other cancer genes, according to cancer
subtype.
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 20152342
The new engl and jour nal of medicine
Figure 4. Distinguishing Mosaicism from Tumor Contamination.
Panel A shows that in the tumor-contaminated germline sample of Patient 1 (E2A019), most somatic mutations
were observed at a lower frequency in the germline than in the tumor. Nine genes were selected to show this point.
Panel B shows that in the case of mosaicism in Patient 2 (HYPO055), only one TP53 mutation was observed at a
lower frequency in the germline than in the tumor. Other somatic mutations in the tumor were absent in the
matched germline sample. Panel C shows that MiSeq sequencing confirmed that the mutant allele fraction (MAF)
of TP53 c.C374G (coding for p.T125R) in the germline sample of Patient 2 was still low (0.20; only 487 reads of 2383
reads had the mutation), a finding that is consistent with germline mosaicism. Two minor peaks supporting C and
G alleles (arrows) were seen in the Sanger-sequencing chromatograph.
MAF(%)
Tumor
100
50
0
C TP53 c.C374G in Patient 2
A Patient 1
Mutant allele Wild-type allele
TP53
LCN
8
SI
W
IPF1
TEX15ZN
F365
SELPRAD
21L1
M
UC16
MAF(%)
Germline
100
50
0
TP53
LCN
8
SI
W
IPF1
TEX15ZN
F365
SELPRAD
21L1
M
UC16
MAF(%)
Tumor
100
50
0
B Patient 2
TP53UG
T2A1
ABCB5CH
ST14
FCRL4
TPCN
1
O
D
Z4
UBR4
RELN
MAF(%)
Germline
100
50
0
TP53UG
T2A1
ABCB5CH
ST14
FCRL4
TPCN
1
O
D
Z4
UBR4
RELN
Platform
Whole-genome sequencing
MiSeq
22
4143
27
4628
0.81
0.90
4
487
37
2383
0.11
0.20
Tumor
no. of
mutant reads
no. of
total reads MAF
no. of
mutant reads
no. of
total reads MAF
Germline
Tumor
Germline
G G G C A A C T G A C C C T G C A A G T C A C A G
G G G C A A C T G A C C G T G C A A G T C A C A G
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 2015 2343
Germline Mutations and Pediatric Cancer
tions included the association of TP53 mutations
with classic Li–Fraumeni syndrome–associated
component cancers such as rhabdomyosarcomas,
osteosarcomas, adrenocortical tumors, CNS
tumors, and leukemia; NF1 mutations with CNS
tumors; RB1 mutations with retinoblastoma and
osteosarcoma; and ALK mutations with neuroblastoma
(Fig. 3A). New associations included
the association of germline TP53, PMS2, and
RET mutations with Ewing’s sarcoma; APC and
SDHB mutations with neuroblastoma; and a
variety of mutations (APC, VHL, CDH1, PTCH1, or
SDHA) with leukemia.
A total of eight children had germline mutations
in the adult-onset cancer–predisposition
genes BRCA1, BRCA2, and PALB2. The spectrum
of cancers observed in these children included
leukemia, CNS tumors, neuroblastoma, osteosarcoma,
and rhabdomyosarcoma. Although biallelic
mutations of BRCA1/2 and PALB2 are known
to cause Fanconi’s anemia,16-19
there were no
germline mutations or deletions involving the
second alleles of these genes in any of the affected
patients.
Medical and Family History
Medical records were available for review for 75 of
the 95 patients with mutations that were deemed
to be pathogenic or probably pathogenic. The
records showed that only 12 patients had undergone
clinical genetic testing previously. Clinical
testing did not identify the predisposing genetic
lesions in 2 patients. Of these 2 patients, 1 had
an adrenocortical tumor tested for TP53 (TP53
p.I332F in Patient ACT001) and 1 had retinoblastoma
that was tested for RB1 (mosaic RB1 p.R445*
in Patient RB002); both lesions were identified
by means of the next-generation sequencing approaches
used in this study.
A total of 58 of the 75 records (77%) contained
information regarding family history, and
only 23 of 58 records (40%) indicated a family
history of cancer (defined here as one or more
first- or second-degree relatives with cancer)
(Fig. S7 in Supplementary Appendix 1). Furthermore,
among these 23 patients, only 13 (57%)
had a history that was consistent with the underlying
genetic syndrome, including 8 patients with
TP53 mutations (and thus the Li–Fraumeni syndrome),
2 with APC mutations (familial adenomatous
polyposis), 2 with BRCA2 mutations
(hereditary breast and ovarian cancer; the pedigrees
are shown in Fig. S8 in Supplementary
Appendix 1), and 1 with PMS2 mutations (hereditary
nonpolyposis colorectal cancer, also known
as the Lynch syndrome). The 8 patients with the
Li–Fraumeni syndrome all met the revised Chompret
criteria regarding family history.20
We completed a similar analysis of a comparison
cohort of 100 randomly selected patients
who did not have germline mutations in the 60
autosomal dominant cancer-predisposition genes.
We observed that the percentage of patients with
a family history of cancer (42%; 18 of 43 records
with family-history information) was similar to
that observed among persons with germline
mutations (40%; 23 of 58 records).
Germline Mutations in Other CancerAssociated
Genes
We identified 4348 nonsilent coding mutations
in the remaining 476 genes that were analyzed.
These included 114 heterozygous truncation mutations,
in 109 patients, that involved tumorsuppressor
genes, tyrosine kinase genes, or other
cancer genes (Fig. 3D, and Table S5 in Supplementary
Appendix 1 and Table S4 in Supplementary
Appendix 2). The most commonly affected
tumor-suppressor genes included CHEK2 (in 4 patients),
PML (in 4), and BUB1B (in 3). A total of
18 patients who did not have pathogenic mutations
in genes that have been associated with
cancer-predisposition syndromes had proteintruncating
mutations in tumor-suppressor genes.
Two known hotspots of somatic activating mutations
in EGFR, T790 and H773, were identified
once each in the germline of 2 patients with
leukemia (Fig. S6 in Supplementary Appendix 1).
Discussion
In this study involving 1120 children and adolescents
with cancer, we found that 8.5% of the
patients had predisposing gene mutations. Sequence
coverage exceeded 10× for more than
95% of the coding exons and 20× for more than
85% of the coding exons in the genes of interest
(Table S3 in Supplementary Appendix 1), which
was sufficient for genotype accuracy.21
However,
the prevalence may be underestimated. First, we
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 20152344
The new engl and jour nal of medicine
included mutations that were pathogenic or probably
pathogenic in 60 genes that have been associated
with clinically relevant autosomal dominant
cancer-predisposition syndromes, and we
did not include other genes that, when mutated,
may contribute toward a patient’s susceptibility
to cancer. In this regard, we observed that an
additional 38 patients (3.4%) had heterozygous
mutations that were pathogenic or probably
pathogenic in 29 genes that are known to be
associated with autosomal recessive cancerpredisposition
syndromes (Table S4 in Supplementary
Appendix 2). Moreover, 109 children (9.7%)
had germline-truncating mutations in other
cancer-associated genes, although non-hotspot
missense mutations in these genes were not fully
characterized, some of which may eventually be
considered to be cancer-susceptibility genes.
Second, among the 226 variants of uncertain
significance that were identified in the 60 genes
that have been associated with autosomal dominant
cancer-predisposition syndromes, 119
(52.7%) were predicted to be deleterious by at
least two computational methods, and some of
these could, in fact, confer susceptibility to cancer.
Third, as sequencing depth increases, additional
mosaic germline mutations will be discovered.
Finally, as we learn more about how
certain genetic alterations (e.g., structural variations,
changes in noncoding regions, and epigenetic
modifications) influence cellular function,
new cancer-predisposing lesions in these
and other newly discovered cancer-associated
genes will be identified.
We found several unexpected germline mutations
in patients with Ewing’s sarcoma, neuroblastoma,
osteosarcoma, or leukemia. Although
Ewing’s sarcoma has been recognized as a second
cancer in children who have been treated for
retinoblastoma,22-25
it has not, to our knowledge,
been associated with other cancer-predisposition
syndromes.26
We found that six patients with
Ewing’s sarcoma had pathogenic germline mutations
in TP53 (in four patients), PMS2 (in one),
or RET (in one), although we cannot state with
certainty that each mutation had a bearing on
the patient’s cancer. Additional studies are needed
to determine the role, if any, that these germline
mutations played in the development of
Ewing’s sarcoma or these other tumors.
Eight patients had heterozygous mutations in
BRCA1, BRCA2, or PALB2. These genes are not
normally examined in children because they are
thought to be predisposition genes for adult
cancer. Magnusson et al. described a high prevalence
of childhood cancer in families with germline
BRCA2 mutations,27
and Brooks et al. reported
20 cases of pediatric cancer among 379 families,
members of which had a mutation in either
BRCA1 or BRCA2.28
These reports suggest that
pathogenic mutations in BRCA1 and BRCA2 are
more common in pediatric cancer than has been
recognized previously and that they potentially
underpin a broader spectrum of cancer pheno-
types.28-38
Family history is commonly used to identify
persons with a possible heritable predisposition,
especially within the pediatric cancer popula-
tion.39
However, only 40% of our patients with
germline mutations that were pathogenic or
probably pathogenic and that could be evaluated
had a family history of cancer. In addition, only
half of those had a family history that was consistent
with a known cancer-predisposition syndrome.
This low frequency probably resulted
from multiple factors, including incomplete information
on family history, de novo mutations,
and incomplete penetrance. Furthermore, parents
and other first- or second-degree relatives of our
pediatric patients are often young, and cancer
may not have developed yet. A review of 100
randomly selected patients who did not have
germline cancer-predisposition gene mutations
revealed that 42% had a family history of cancer.
Conceivably, some of these patients have mutations
in genes that were not analyzed in the
current study. Nonetheless, on the basis of these
observations, family history cannot be the sole
indication used to guide the provision of genetic
testing.
This study has several limitations. First, several
subtypes of pediatric cancers were not examined.
In addition, our cohort included greaterthan-expected
proportions of patients with
hypodiploid acute lymphoblastic leukemia and
those with adrenocortical tumors. However, after
excluding these two subtypes, the prevalence of
germline mutations of 5.6% was still significantly
higher than the prevalence in two control
cohorts (P<10−7
by Fisher’s exact test for both
comparisons). Observation of pathogenic mutations
in the controls may indicate uncharacterized
cancer phenotypes in the people enrolled,
rather than refuting the pathogenicity of these
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 2015 2345
Germline Mutations and Pediatric Cancer
mutations in cancer predisposition (Tables S6
and S7 in Supplementary Appendix 1). Moreover,
by adjusting for the distribution of cancer subtypes
observed in the SEER program and by applying
previously reported mutation frequencies
for those not included in this study, we predicted
an overall mutation prevalence of 7.3 to 9.8% in
the SEER pediatric cancer population (Table S10
in Supplementary Appendix 2).
Second, we did not study the parents or relatives
of the patients in our cohort and hence
could not assess whether variants were new or
segregated with a cancer phenotype among family
members. This information could have augmented
the evidence of pathogenicity in some
cases. Nonetheless, the discovery of four mosaic
germline mutations in TP53 and RB1 indicates
that a fraction of the mutations that were identified
in this study were de novo. Third, although
the 60 autosomal dominant cancer-predisposition
genes that were the focus of this report have
been well characterized, information regarding
the penetrance of many mutations that were
identified within these genes is lacking. Additional
family, epidemiologic, and functional studies
are warranted to better understand the cancer
risks associated with each of these variants.
In conclusion, germline mutations in cancerpredisposing
genes were identified in 8.5% of
the children and adolescents with cancer who
participated in this study. Family history did not
predict the presence of an underlying predisposition
syndrome in most patients. The germline
mutations identified in this study may provide
insights into the causes of cancer. Knowledge of
their presence may influence clinical management
by directing cancer care, enabling presymptomatic
genetic testing of relatives, guiding
family-planning measures, and facilitating the
institution of potentially lifesaving measures for
cancer prevention and surveillance.
Supported by funding from the American Lebanese Syrian
Associated Charities to St. Jude Children’s Research Hospital
and by a Cancer Center Support (Core) grant (CA21765) from
the National Cancer Institute to St. Jude Children’s Research
Hospital.
Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
We thank the staff of the Pediatric Cancer Genome Project
Laboratory and the Molecular Pathology Laboratory at St. Jude’s
Children’s Research Hospital for next-generation sequencing
and subsequent validation of identified pathogenic variants.
References
1.	 Futreal PA, Coin L, Marshall M, et al.
A census of human cancer genes. Nat Rev
Cancer 2004;​4:​177-83.
2.	 Lindor NM, McMaster ML, Lindor CJ,
Greene MH. Concise handbook of familial
cancer susceptibility syndromes —
second edition. J Natl Cancer Inst Monogr
2008;​(38):​1-93.
3.	 Pizzo PA, Poplack DG, eds. Principles
and practice of pediatric oncology. 6th ed.
Philadelphia:​Wolters Kluwer–Lippincott
Williams & Wilkins Health, 2011.
4.	 Pui C-H. Childhood leukemias. 3rd ed.
Cambridge, United Kingdom:​Cambridge
University Press, 2012.
5.	 Green RC, Berg JS, Grody WW, et al.
ACMG recommendations for reporting of
incidental findings in clinical exome and
genome sequencing. Genet Med 2013;​15:​
565-74.
6.	 Bombard Y, Offit K, Robson ME. Risks
to relatives in genomic research: a duty to
warn? Am J Bioeth 2012;​12:​12-4.
7.	 Offit K, Groeger E, Turner S, Wadsworth
EA, Weiser MA. The “duty to warn”
a patient’s family members about hereditary
disease risks. JAMA 2004;​292:​1469-
73.
8.	 Storm C, Agarwal R, Offit K. Ethical
and legal implications of cancer genetic
testing: do physicians have a duty to warn
patients’ relatives about possible genetic
risks? J Oncol Pract 2008;​4:​229-30.
9.	 Zhao M, Sun J, Zhao Z. TSGene: a Web
resource for tumor suppressor genes.
Nucleic Acids Res 2013;​41:​D970-D976.
10.	Edmonson MN, Zhang JH, Yan CH,
Finney RP, Meerzaman DM, Buetow KH.
Bambino: a variant detector and alignment
viewer for next-generation sequencing
data in the SAM/BAM format. Bioinformatics
2011;​27:​865-6.
11.	Chen X, Gupta P, Wang J, et al.
­CONSERTING: integrating copy-number
analysis with structural-variation detection.
Nat Methods 2015;​12:​527-30.
12.	 Wang J, Mullighan CG, Easton J, et al.
CREST maps somatic structural variation
in cancer genomes with base-pair resolution.
Nat Methods 2011;​8:​652-4.
13.	 Richards S, Aziz N, Bale S, et al. Standards
and guidelines for the interpretation
of sequence variants: a joint consensus
recommendation of the American
College of Medical Genetics and Genomics
and the Association for Molecular Pathology.
Genet Med 2015;​17:​405-24.
14.	 Holmfeldt L, Wei L, Diaz-Flores E, et al.
The genomic landscape of hypodiploid
acute lymphoblastic leukemia. Nat Genet
2013;​45:​242-52.
15.	Wu G, Diaz AK, Paugh BS, et al. The
genomic landscape of diffuse intrinsic
pontine glioma and pediatric non-brainstem
high-grade glioma. Nat Genet 2014;​
46:​444-50.
16.	Alter BP, Kupfer G. Fanconi anemia.
Seattle:​University of Washington, 2002.
17.	Howlett NG, Taniguchi T, Olson S,
et al. Biallelic inactivation of BRCA2 in
Fanconi anemia. Science 2002;​297:​606-9.
18.	Reid S, Schindler D, Hanenberg H,
et al. Biallelic mutations in PALB2 cause
Fanconi anemia subtype FA-N and predispose
to childhood cancer. Nat Genet
2007;​39:​162-4.
19.	 Sawyer SL, Tian L, Kahkonen M, et al.
Biallelic mutations in BRCA1 cause a new
Fanconi anemia subtype. Cancer Discov
2015;​5:​135-42.
20.	Bougeard G, Sesboüé R, Baert-Desurmont
S, et al. Molecular basis of the Li–
Fraumeni syndrome: an update from the
French LFS families. J Med Genet 2008;​45:​
535-8.
21.	Meldrum C, Doyle MA, Tothill RW.
Next-generation sequencing for cancer
diagnostics: a practical perspective. Clin
Biochem Rev 2011;​32:​177-95.
22.	Cope JU, Tsokos M, Miller RW. Ewing
sarcoma and sinonasal neuroectodermal
tumors as second malignant tumors after
retinoblastoma and other neoplasms. Med
Pediatr Oncol 2001;​36:​290-4.
23.	Mittal R, Al Awadi S, Sahar O, Behbehani
AM. Ewing’s sarcoma as second
malignant neoplasm after retinoblastoma:
a case report. Med Princ Pract 2008;​
17:​84-5.
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.
n engl j med 373;24 nejm.org  December 10, 20152346
Germline Mutations and Pediatric Cancer
24.	 Spunt SL, Rodriguez-Galindo C, Fuller
CE, et al. Ewing sarcoma-family tumors
that arise after treatment of primary
childhood cancer. Cancer 2006;​107:​201-6.
25.	Tahasildar N, Goni V, Bhagwat K,
Tripathy SK, Panda BB. Ewing’s sarcoma
as second malignancy following a short
latency in unilateral retinoblastoma. J Orthop
Traumatol 2011;​12:​167-71.
26.	Randall RL, Lessnick SL, Jones KB,
et al. Is there a predisposition gene for
Ewing’s sarcoma? J Oncol 2010;​2010:​
397632. 10.1155/2010/397632
27.	 Magnusson S, Borg A, Kristoffersson
U, Nilbert M, Wiebe T, Olsson H. Higher
occurrence of childhood cancer in families
with germline mutations in BRCA2,
MMR and CDKN2A genes. Fam Cancer
2008;​7:​331-7.
28.	Brooks GA, Stopfer JE, Erlichman J,
Davidson R, Nathanson KL, Domchek SM.
Childhood cancer in families with and
without BRCA1 or BRCA2 mutations ascertained
at a high-risk breast cancer
clinic. Cancer Biol Ther 2006;​5:​1098-
102.
29.	 Miki Y, Swensen J, Shattuck-Eidens D,
et al. A strong candidate for the breast
and ovarian cancer susceptibility gene
BRCA1. Science 1994;​266:​66-71.
30.	 Wooster R, Bignell G, Lancaster J, et al.
Identification of the breast cancer susceptibility
gene BRCA2. Nature 1995;​378:​789-92.
31.	 Armes JE, Trute L, White D, et al. Distinct
molecular pathogeneses of earlyonset
breast cancers in BRCA1 and BRCA2
mutation carriers: a population-based
study. Cancer Res 1999;​59:​2011-7.
32.	 Cavalli LR, Singh B, Isaacs C, Dickson
RB, Haddad BR. Loss of heterozygosity in
normal breast epithelial tissue and benign
breast lesions in BRCA1/2 carriers
with breast cancer. Cancer Genet Cytogenet
2004;​149:​38-43.
33.	Dworkin AM, Spearman AD, Tseng
SY, Sweet K, Toland AE. Methylation not
a frequent “second hit” in tumors with
germline BRCA mutations. Fam Cancer
2009;​8:​339-46.
34.	 Johnston JJ, Rubinstein WS, Facio FM,
et al. Secondary variants in individuals
undergoing exome sequencing: screening
of 572 individuals identifies high-penetrance
mutations in cancer-susceptibility
genes. Am J Hum Genet 2012;​91:​97-108.
35.	 Lucas AL, Shakya R, Lipsyc MD, et al.
High prevalence of BRCA1 and BRCA2
germline mutations with loss of heterozygosity
in a series of resected pancreatic
adenocarcinoma and other neoplastic lesions.
Clin Cancer Res 2013;​19:​3396-403.
36.	Osorio A, de la Hoya M, RodriguezLopez
R, et al. Loss of heterozygosity
analysis at the BRCA loci in tumor samples
from patients with familial breast
cancer. Int J Cancer 2002;​99:​305-9.
37.	Staff S, Isola JJ, Johannsson O, Borg
A, Tanner MM. Frequent somatic loss of
BRCA1 in breast tumours from BRCA2
germ-line mutation carriers and vice versa.
Br J Cancer 2001;​85:​1201-5.
38.	Willems AJ, Dawson SJ, Samaratunga
H, et al. Loss of heterozygosity at the
BRCA2 locus detected by multiplex ligation-dependent
probe amplification is common
in prostate cancers from men with a
germline BRCA2 mutation. Clin Cancer
Res 2008;​14:​2953-61.
39.	Knapke S, Nagarajan R, Correll J,
Kent D, Burns K. Hereditary cancer risk
assessment in a pediatric oncology follow-up
clinic. Pediatr Blood Cancer 2012;​
58:​85-9.
Copyright © 2015 Massachusetts Medical Society.	
The New England Journal of Medicine
Downloaded from nejm.org at Masarykova Univerzita v Brne on February 1, 2016. For personal use only. No other uses without permission.
Copyright © 2015 Massachusetts Medical Society. All rights reserved.