close

Вход

Забыли?

вход по аккаунту

?

j.jmoldx.2018.06.009

код для вставкиСкачать
Accepted Manuscript
SM
OncoKids : A Comprehensive Next-Generation Sequencing Panel for Pediatric
Malignancies
Matthew C. Hiemenz, Dejerianne G. Ostrow, Tracy M. Busse, Jonathan Buckley,
Dennis T. Maglinte, Moiz Bootwalla, James Done, Jianling Ji, Gordana Raca, Alex
Ryutov, Xinjie Xu, Chao Jie Zhen, Jeffrey M. Conroy, Florette K. Hazard, Joshua
L. Deignan, Beverly Rogers, Amanda L. Treece, David M. Parham, Xiaowu Gai,
Alexander R. Judkins, Timothy J. Triche, Jaclyn A. Biegel
PII:
S1525-1578(18)30102-8
DOI:
10.1016/j.jmoldx.2018.06.009
Reference:
JMDI 721
To appear in:
The Journal of Molecular Diagnostics
Received Date: 16 March 2018
Revised Date:
22 May 2018
Accepted Date: 11 June 2018
Please cite this article as: Hiemenz, MC, Ostrow, DG, Busse, TM, Buckley, J, Maglinte, DT, Bootwalla,
M, Done, J, Ji, J, Raca, G, Ryutov, A, Xu, X, Zhen, CJ, Conroy, JM, Hazard, FK, Deignan, JL, Rogers,
SM
B, Treece, AL, Parham, DM, Gai, X, Judkins AR, Triche, TJ, Biegel JA, OncoKids : A Comprehensive
Next-Generation Sequencing Panel for Pediatric Malignancies, The Journal of Molecular Diagnostics
(2018), doi: 10.1016/j.jmoldx.2018.06.009.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
OncoKidsSM: A Comprehensive Next-Generation Sequencing Panel for Pediatric
Malignancies
RI
PT
Matthew C. Hiemenz,*† Dejerianne G. Ostrow,* Tracy M. Busse,* Jonathan Buckley,*† Dennis
T. Maglinte,* Moiz Bootwalla,* James Done,* Jianling Ji,*† Gordana Raca,*† Alex Ryutov,*
Xinjie Xu,‡§ Chao Jie Zhen,¶ Jeffrey M. Conroy, ||** Florette K. Hazard,†† Joshua L.
SC
Deignan,‡‡ Beverly Rogers,§§ Amanda L. Treece,¶¶ David M. Parham,*† Xiaowu Gai,*†
M
AN
U
Alexander R. Judkins,*† Timothy J. Triche,*† Jaclyn A. Biegel*†
From the Department of Pathology and Laboratory Medicine,* Children's Hospital Los Angeles,
Los Angeles, California; the Department of Pathology and Laboratory Medicine,† USC Keck
School of Medicine, Los Angeles, California; ARUP Laboratories,‡ Salt Lake City, Utah; the
Department of Pathology,§ University of Utah School of Medicine, Salt Lake City, Utah; the
TE
D
Department of Pathology,¶ University of Chicago, Chicago, Illinois; OmniSeq Inc.,|| Buffalo, New
York; Center for Personalized Medicine,** Roswell Park Cancer Institute, Buffalo, New York; the
Department of Pathology,†† Stanford University School of Medicine, Stanford, California; the
EP
Department of Pathology and Laboratory Medicine,‡‡ David Geffen School of Medicine at
UCLA, Los Angeles, California; the Department of Pathology and Laboratory Medicine,§§
AC
C
Children’s Healthcare of Atlanta, Atlanta, Georgia; the Department of Pathology and Laboratory
Medicine,¶¶ Children’s Hospital Colorado, Denver, Colorado
Corresponding Author: Matthew Hiemenz, Children's Hospital Los Angeles, 4650 Sunset
Blvd, MS#173, Los Angeles, CA 90027. E-mail: mhiemenz@chla.usc.edu
ACCEPTED MANUSCRIPT
Funding: Supported by institutional funding from the Children's Hospital Los Angeles and
Thermo Fisher Scientific, Inc.
RI
PT
Disclosures: J.M.C. is an employee of OmniSeq, Inc. (Buffalo, NY) and holds restricted stock in
OmniSeq, Inc. J.M.C is an employee of Roswell Park Comprehensive Cancer Center (Buffalo,
NY). Roswell Park Comprehensive Cancer Center is the majority shareholder of OmniSeq, Inc.
AC
C
EP
TE
D
M
AN
U
SC
Running Header: OncoKidsSM Cancer Panel for Pediatric Malignancies
ACCEPTED MANUSCRIPT
ABSTRACT
The OncoKidsSM panel is an amplification-based next-generation sequencing assay designed to
detect diagnostic, prognostic, and therapeutic markers across the spectrum of pediatric
RI
PT
malignancies, including leukemias, sarcomas, brain tumors, and embryonal tumors.
This panel uses low input amounts of DNA (20 nanograms) and RNA (20 nanograms) and is
compatible with formalin-fixed, paraffin-embedded as well as frozen tissue, bone marrow, and
SC
peripheral blood. The DNA content of this panel covers the full coding regions of 44 cancer
predisposition loci, tumor suppressor genes, and oncogenes; hotspots for mutations in 82
M
AN
U
genes; and amplification events in 24 genes. The RNA content includes 1,421 targeted gene
fusions. We describe the validation of this panel using a large cohort of 192 unique clinical
samples that included a wide range of tumor types and alterations. Robust performance was
observed for analytical sensitivity, reproducibility, and limit of detection studies. The results from
this study support the use of OncoKidsSM for routine clinical testing of a wide variety of pediatric
AC
C
EP
TE
D
malignancies.
ACCEPTED MANUSCRIPT
INTRODUCTION
Pediatric cancer is the leading cause of death by disease past infancy among children in
the United States.1 At the genetic level, pediatric cancers are unlike adult cancers. In contrast
RI
PT
with adult malignancies, pediatric cancers are more likely to be driven by gene fusions and copy
number alterations and typically harbor fewer somatic DNA single nucleotide variants (SNVs),
multi-nucleotide variants (MNVs), and insertions/deletions (InDels).2 Although individual
SC
childhood tumors often contain fewer DNA mutations than adult tumors, several key driver DNA
mutations have been described in different pediatric tumors. Examples include SMARCB1
mutations in pediatric gliomas.3-5
M
AN
U
mutations in rhabdoid tumors, ALK point mutations in neuroblastomas, and BRAF point
Most current somatic next-generation sequencing (NGS) panels are designed for adult
cancers and therefore are not optimized or do not cover the common genetic alterations found
in pediatric tumors. To address this limitation, a custom NGS panel, OncoKidsSM, was designed.
TE
D
The OncoKidsSM NGS test targets diagnostic, prognostic, and therapeutic biomarkers across the
spectrum of childhood cancers in a combined DNA and RNA sequencing assay. Pediatric tumor
types addressed by this assay include pediatric leukemias, brain tumors, sarcomas, and
EP
embryonal tumors, including neuroblastoma, retinoblastoma, Wilms tumor, and liver tumors.
OncoKidsSM is amplification-based, and covers a large number of DNA regions (3,069
AC
C
amplicons) and RNA fusions (1,421 fusion primer pairs). The panel content is principally aimed
at pediatric cancers, however, a number of gene fusions relevant to adult cancers (eg, EML4ALK in non–small-cell lung cancer) with approved targeted therapeutics are also included, as
detection of these gene fusions in pediatric malignancies might warrant use of the matched
therapy.
Amplification-based NGS panels offer several advantages for testing of pediatric
malignancies. Importantly, amplification-based assays generally have significantly lower input
requirements for DNA and RNA than hybrid capture–based assays6. This is particularly relevant
ACCEPTED MANUSCRIPT
for pediatric malignancies where smaller biopsies are often submitted to the laboratory for
testing. Additionally, amplification-based NGS assays are typically more robust than hybrid
capture assays at handling samples with degraded DNA and RNA,6 including formalin-fixed,
RI
PT
paraffin-embedded (FFPE) specimens. Amplification-based assays generate deep read depths
that can be valuable for assessment of clonal heterogeneity and emergence of treatment
resistant clones in tumors. Despite the strengths of amplification-based NGS assays, hybrid
capture assays may demonstrate better performance in certain areas such as greater
SC
uniformity, especially in larger panels or exomes.7 Although several studies have examined the
M
AN
U
use of hybrid capture NGS assays for pediatric malignancies,8,9 this assay represents a different
approach that may be advantageous for the routine testing of pediatric malignancies in the
clinical laboratory.
A common application of targeted hybrid capture NGS assays is the capture of gene
fusions by tiling over intronic regions in a gene with one or more fusion breakpoints of interest.10
TE
D
This approach enables the use of DNA rather than RNA for gene fusion detection and is
amenable to finding unknown partner genes. However, this strategy often requires a large
number of probes to cover longer introns and higher nucleic acid input amounts than
EP
amplification-based approaches for fusion detection. Therefore, amplification-based NGS may
be beneficial for gene fusion detection when a smaller panel is desired or in the setting of limited
AC
C
nucleic acid input amounts.
Though prior studies have examined the use of NGS in a subset of pediatric oncology
patients with high-risk, recurrent, refractory, or relapsed disease,8,9 OncoKidsSM is designed for
implementation as a front-line, routine test in the clinical laboratory. Routine molecular profiling
of pediatric malignancies by NGS panels, such as OncoKidsSM, may provide a significant aid in
diagnosis. For example, pediatric soft tissue tumors, especially those within the broad category
of non-rhabdomyosarcoms, can be difficult to diagnose.11 The differential diagnosis of these
tumors typically requires immunohistochemistry and targeted molecular genetic studies.11
ACCEPTED MANUSCRIPT
However, selecting the correct targeted test can be challenging as many rare lesions with
characteristic molecular findings may be included in that differential diagnosis.12 Additionally,
poorly differentiated lesions with characteristic molecular alterations, such as synovial
RI
PT
sarcomas, may not always be sufficiently apparent by histopathology and
immunohistochemistry to prompt the selection of the correct targeted test.13,14 Therefore, testing
these tumors at the time of initial diagnosis with a broad NGS panel can aid in the identification
SC
of diagnostic markers that may have been missed with a targeted approach such as
fluorescence in situ hybridization (FISH) or reverse transcription (RT)-PCR. Further, even a
M
AN
U
defined morphologic entity such as PAX-FOXO1 positive alveolar rhabdomyosarcoma may
harbor unexpected genetic defects with potential matched targeted therapies (eg, FGFR4
mutation/amplification and FGFR kinase inhibitors) that would be missed in single-analyte
assays. As treatment protocols are increasingly focused on precise diagnosis for soft tissue
tumors/sarcomas, broad molecular profiling of pediatric malignancies will become ever more
TE
D
critical as a diagnostic tool.15
In addition to improved diagnosis, routine application of the OncoKidsSM panel can
improve therapy by identifying targetable mutations in the front-line setting. Of note, imatinib has
EP
been incorporated into current treatment algorithms for Philadelphia (Ph) chromosome–positive
acute lymphoblastic leukemia (ALL).16 Importantly, a number of promising targeted therapies
AC
C
are currently in prospective clinical trials for diverse childhood cancers including Ph-like ALL,17
BRAF V600E–positive gliomas,18 and NTRK1 fusion–positive pediatric mesenchymal tumors.19
By identifying the targetable alterations in these malignancies, OncoKidsSM enables the use of
precision therapies prospectively rather than in the setting of relapse or recurrence where they
may be less effective.
In this study, we describe the technical performance and validation of the OncoKidsSM
NGS assay. The validation set included a large number of samples harboring diverse mutations
and multiple specimen types including FFPE, frozen tissue, peripheral blood, and bone marrow
ACCEPTED MANUSCRIPT
aspirates. In addition, normal tissue samples were included to develop the copy number assay
and assess technical background, including gene fusions, associated with non-neoplastic
tissue. Sequencing of all samples was performed using the Ion S5 sequencing platform
RI
PT
(Thermo Fisher Scientific, Waltham, MA). The validation performance for analytical sensitivity,
reproducibility, and limit of detection establish OncoKidsSM as a robust method for the routine
SC
sequencing of solid and liquid tumors of childhood
MATERIALS AND METHODS
M
AN
U
Panel Design
A comprehensive catalog of genomic alterations (diagnostic, prognostic, or therapeutic)
reported in all common forms of pediatric cancer (SNVs, InDels, gene amplifications, and gene
fusions) was identified. This was accomplished through extensive literature searches, evaluation
of content from other oncology NGS panels, and expert review from subject matter experts,
TE
D
including laboratory directors, pediatric pathologists, and oncologists. AmpliSeq primers were
designed and supplied by Thermo Fisher Scientific, for both extracted DNA and cDNA produced
by reverse transcription of extracted RNA. This assay is commercially available from Thermo
EP
Fisher Scientific (Oncomine Childhood Cancer Research Assay).
AC
C
AcroMetrix Control
The AcroMetrix Oncology Hotspot Control (Thermo Fisher Scientific), a synthetic mixture of
known DNA variants and genomic DNA, was used to assess sensitivity and specificity. The
Clopper and Pearson method was used to calculate 95% confidence intervals (CI) for this and
other analyses (http://epitools.ausvet.com.au/content.php?page=CIProportion; last accessed
March 7, 2018). The synthetic DNA in this control is present at low variant allele frequencies
ACCEPTED MANUSCRIPT
(5% to 35%) to mimic somatic mutations in cancer. Of 521 variants in the synthetic DNA, 315
were in bases covered by the OncoKidsSM panel (213,825 bases covered, in total).
RI
PT
Samples with a Pathogenic Alteration
De-identified clinical samples from Children's Hospital Los Angeles (CHLA) included the
following specimen types: bone marrow aspirate, frozen tissue, peripheral blood, and FFPE. For
SC
DNA and/or RNA received from outside collaborators, the specimen type and diagnosis were
recorded. Overall, 192 clinical samples were processed, including 168 clinical samples
M
AN
U
harboring one or more pathogenic alterations and 24 non-neoplastic samples. Three cell line
samples harboring a BRAF V600E mutation were also processed. Six samples had both FFPE
and frozen tissue sections that were tested in this study. For specimens with a pathogenic
alteration, the number of samples in each specimen type included: 66 FFPE tissue samples, 64
bone marrow aspirates, 22 peripheral blood specimens, 19 frozen tissue samples, three cell line
TE
D
specimens, two cytology smears, and isolated DNA from an archival skin lesion. A wide range
of tumor types were tested including 44 B-lymphoblastic leukemias, 30 soft tissue
tumors/sarcomas, 26 myeloid neoplasms, 18 brain tumors (glial or ependymal neoplasms), and
EP
nine embryonal tumors. Cases with a previously confirmed feature of interest were chosen from
orthogonal data that included FISH, chromosomal microarray (CMA), Sanger sequencing, and
AC
C
NGS-based panel sequencing. For the DNA component of the panel, samples harboring
germline and/or somatic single nucleotide variants (SNVs), muti-nucleotide variants (MNVs),
insertions and deletions, and gene amplification were analyzed. For the RNA component,
samples with gene fusions were tested. De-identified clinical samples from CHLA with a range
of tumor cellularity (13% to 100%) were tested. All of the OncoKidsSM analyses were performed
in a blinded fashion.
ACCEPTED MANUSCRIPT
Nucleic Acid Extraction
Total RNA was extracted from 400 µL of blood or 300 µL of bone marrow aspirate (collected in
a purple top EDTA tube) using the commercially available Maxwell RSC simplyRNA Blood Kit
RI
PT
(Promega, Madison, WI). Total RNA was extracted from frozen tumor tissue using the RNeasy
Mini Kit (Qiagen, Hilden, Germany). Total RNA was extracted from FFPE using the Agencourt
FormaPure Kit (Beckman Coulter, Brea, CA). All RNA was DNase -treated and quantified with
SC
the Promega Quantus and the QuantiFluor RNA System (Promega). Extracted RNA was
assayed on the TapeStation RNA ScreenTape (Agilent Technologies, Santa Clara, CA) and the
M
AN
U
percentage of RNA fragments >200 nucleotides was calculated (DV200). A DV200 less than
50% was considered marginal sample quality. Genomic DNA was extracted from fice scrolls (20
µm) of either frozen or FFPE-preserved tumor tissue. The Qiagen QIAamp DNA Mini Kit or the
Gentra Puregene Tissue Kit was used for frozen tissue and the Qiagen QIAmp DNA FFPE
Tissue Kit was used for FFPE samples (all from Qiagen). A small number of FFPE samples
TE
D
were extracted with Agencourt Formapure. A volume of 50 to 200 µL of bone marrow aspirate
(collected in a purple top EDTA tube) was extracted using the Promega Maxwell RSC Blood
DNA Kit. DNA was RNAse-treated and quantified using the Promega Quantus and the
EP
QuantiFluor ONE dsDNA System (Promega). Extracted DNA was assayed on the TapeStation
Genomic DNA ScreenTape (Agilent) and the percentage DNA fragments >1000 bp was
AC
C
calculated (DV1000). For samples that did not meet the minimum-required DNA concentration,
a calculated sample volume equivalent to 20 ng of DNA was volume reduced to 7 uL (2.9 ng/uL)
using the Savant SpeedVac Concentrator.
Library Preparation for Sequencing
Molecular testing for sequence changes and abnormal gene fusions in tumor cells was
performed using the OncoKidsSM panel. A total of 20 nanograms (ng) of DNA and two highly
ACCEPTED MANUSCRIPT
multiplexed DNA primer pools were used to generate 3,069 amplicons per sample, covering the
full coding regions of 44 cancer predisposition loci, tumor suppressor genes and oncogenes;
hotspots for mutations in 82 genes; and amplification events in 24 genes (Supplemental Table
RI
PT
S1). Simultaneously, two highly multiplexed RNA primer pools and 20 ng of RNA were used to
interrogate 1,421 targeted gene fusions associated with acute myeloid and acute lymphoblastic
leukemia, childhood sarcomas, pediatric brain tumors, and soft tissue tumors. In addition to
SC
gene fusions, expression of the following genes was also captured: BCL2, BCL6, FGFR1,
FGFR4, HMBS, IGF1R, ITGB7, LRP1, MET, MYC, MYCN, TBP, and TOP2A. Library
M
AN
U
preparation for each sample was performed following the manufacturer’s protocol for the
Oncomine Comprehensive Assay using the AmpliSeq Library Kit 2.0 (Thermo Fisher Scientific),
SuperScript VILO cDNA Synthesis Kit (SuperScript III), and four pools of custom
oligonucleotides (two for DNA libraries and two for RNA libraries). Amplicon libraries with
barcodes specific to each sample were generated from DNA or cDNA and quantified using the
TE
D
TaqMan Quantitation Kit (Thermo Fisher Scientific). A library concentration less than 100 pM
indicated marginal sample performance. Libraries were diluted to 50 picomolar (pM) and four
DNA libraries and four RNA libraries were pooled at a four to one ratio (DNA: RNA). The Ion
EP
Chef (Thermo Fisher Scientific) was used for automated clonal amplification by emulsion PCR
using the Ion 540 Kit - Chef (Thermo Fisher Scientific) and automated 540 chip loading. A no
AC
C
template control (NTC) was included in each batch of library preparations. NTC libraries were
not sequenced so as to avoid diluting the emulsion PCR templating reactions.
Sequencing Data Analysis Using Ion Torrent S5 and S5XL
Sequencing was performed using 200 bp reads and the 540 chip on the Ion Torrent S5/S5XL
sequencing platform, and downstream data analysis utilized Torrent Suite 5.2.1 software
(Thermo Fisher Scientific). Default analysis parameters were used with the exception of base
calling. Additional base calling parameters were used for the BaseCaller option including -J 25, -
ACCEPTED MANUSCRIPT
-end-repair 15, and --context. –J, and --end-repair parameters rescue false negatives by
selectively forcing alignment at the 3' end of the read, and the –context parameter turns on the
option for gap penalty in homopolymers to achieve more accurate alignments. DNA library
RI
PT
reads were aligned to Human genome build 19 within Torrent Suite, and a minimum average
read depth of 1200 was required for analysis. DNA BAM files were uploaded to Ion Reporter 5.2
(Thermo Fisher Scientific) for variant calling and amplification analysis. The “AmpliSeq
SC
Childhood Cancer Research Panel - w2.1 – DNA - Single Sample” workflow was used with
some modifications. All blacklist variants were removed from the HotSpot Regions BED file and
M
AN
U
the Kmer Length parameter was also set to 15.
Ion Reporter 5.2 was used to call SNVs/MNVs and an in-house variant caller (Local Adjustment
for Background, or LAB) was developed to provide improved InDel variant calling. In brief,
observations of variant bases at a given genomic locus were evaluated by comparison to data
TE
D
from a reference set of 200 samples. After filtering out likely true positives from the reference
set, the remaining variant allele frequency (VAF) values represent the background noise (at that
locus), and the variant is only called when the observed VAF differ significantly from the
EP
background distribution.
AC
C
Variant Call Format (VCF) fiIes were generated from each variant caller and loaded separately
into a custom variant review analysis tool, Integrated Curation Environment (ICE). For each
case analyzed, the following variants were filtered/removed in ICE: variants with a population
minor allele frequency greater than one percent, synonymous variants, variants greater than five
base pairs into the introns, 3' or 5' untranslated regions, and intergenic regions. Minor allele
frequency was derived from 60,000 exomes downloaded from the Exome Aggregation
Consortium (ExAC; http://exac.broadinstitute.org/l last accessed March 7, 2018). Importantly,
filtered-out variants were also examined to ensure that no clearly pathogenic variants or variants
ACCEPTED MANUSCRIPT
present within the Catalogue of Somatic Mutations in Cancer (COSMIC) database were
incorrectly removed. For RNA libraries, 100,000 mapped reads and >20,000 gene expression
control reads was required. Fusions were analyzed in Ion Reporter 5.2 using the “AmpliSeq
RI
PT
Childhood Cancer Research Panel - w2.1 - Fusions - Single Sample” workflow with updated
reference fusion panel, and filtered to include all fusion calls with at least one supporting read.
Genes targeted for expression were also filtered to include all calls with at least one supporting
SC
read. At least 40 reads supporting a fusion were required to call it present. The dominant fusion
M
AN
U
for each sample was defined as the fusion with the highest read counts in that sample.
Copy number estimates were made using Thermo Fisher’s proprietary algorithm, the Variability
Correction Informatics Baseline (VCIB). The VCIB algorithm trains on a large number of diverse
samples to capture systematic effects and encodes them into a file (the "baseline"). The
baseline allows for assessment of corrected log2 ratios of amplicons of identified copy number
TE
D
variant (CNV) regions (usually genes) in input sample data, adjusting for known sources of
variability including pool imbalance across the two pools of amplicons, total number of reads per
amplicon, attributes of GC proportion, and length of the amplicon insert. This is followed by a
EP
correction for the percent tumor cellularity recorded for the sample to give copy number and
confidence interval data for the identified CNV regions. For the OncoKidsSM panel, the baseline
AC
C
was created using 40 well-characterized reference cell lines and the AcroMetrix reference
sample (processed in the Thermo Fisher laboratories, San Francisco, CA), 24 non-neoplastic
samples, and 58 neoplastic samples (derived from peripheral blood, bone marrow aspirate,
frozen tissue, and FFPE) run at CHLA.
Limit of Detection and Reproducibility Studies
The limit of detection (LOD) of the OncoKidsSM panel was assessed by serial dilution of nucleic
acid from a known positive sample with either DNA from a standard reference sample or RNA
ACCEPTED MANUSCRIPT
derived from a sample of the same preservation method with no pathogenic alteration. For the
DNA panel, two dilution series were performed. A sample harboring an RB1 InDel was chosen
along with a sample carrying amplification above the calling threshold of eight estimated copies
RI
PT
(MDM2) as well as amplification of MYC slightly below the cutoff (Supplemental Figure S1).
Each sample was diluted with DNA from the standard reference NA12878 cell line to the
following ratios: 50%, 25%, 12.5%, 6.25%, and 3.13%. Both samples also carried SNVs that
SC
were not shared by the NA12878 cell line. Therefore, each dilution series also assessed the
ability to detect SNVs at each dilution point. For the RNA panel, dilution series were performed
M
AN
U
for fusions in bone marrow (PML-RARA), frozen tissue (PAX3-FOXO1), and FFPE tissue
(KIAA1549-BRAF) at the following ratios: 12.5%, 6.25%, 3.13%, 1.57%, and 0.16%.
Assay repeatability (intra-run) and reproducibility (inter-run) were determined by sequencing
replicate libraries for each of three DNA samples and three RNA samples. Inter-run variability
TE
D
(two to three different library batches) and intra-run variability (for one of the inter-run batches, a
total of three libraries for a given sample were prepared simultaneously) were measured. For
DNA samples, clinically significant variants were compared across runs. Fusion reads were
AC
C
RESULTS
EP
compared across runs for all RNA samples.
Sequencing Performance
A typical sequencing run performed on the Ion S5 sequencer was composed of four paired DNA
and RNA samples for a total of eight multiplexed samples. Over 46 sequencing runs, the
average chip loading density was 89% with an average of 81 million reads per run. For the DNA
portion of the panel, there were an average of 15.4 million mapped sequencing reads per
sample at an average depth of 5,672X and a uniformity of 96.43%. Uniformity is defined as the
ACCEPTED MANUSCRIPT
percentage of bases that are covered to at least 20% of the overall mean read depth of the
sample. For the RNA component of the assay, there were an average of 1.88 million mapped
RI
PT
reads per sample and an average read depth per fusion of 160,382 reads.
Examination of Non-Neoplastic Samples for Background Fusions
Recurrent fusion RNAs have been identified in non-cancer specimens.20 Therefore, to assess
SC
the presence of potential background fusions or fusions associated with non-neoplastic tissue,
24 non-neoplastic samples were analyzed. These samples included eight peripheral blood
M
AN
U
specimens, one bone marrow aspirate, eight frozen tissue samples, and seven FFPE tissues.
Nineteen of 24 samples demonstrated one or more fusion signals. For the dominant fusion in
each sample, the fusion read count ranged from 216 to 80,339 reads. Of note, only one sample
had a targeted fusion, an FFPE sample with a FUS-CREB3L2 fusion in 216 reads. The most
common fusions detected were MYH9-BRD1 and MET-MET. MYH9-BRD1 fusions were seen
TE
D
exclusively in peripheral blood and bone marrow whereas MET-MET fusions were only seen in
frozen and FFPE solid tissue. Nine bone marrow samples with a MYH9-BRD1 fusion observed
with OncoKidsSM (read count range 1,394 to 40,668) were chosen for analysis by reverse
EP
transcription-PCR with custom primers for the MYH9-BRD1 fusion breakpoint followed by
Sanger sequencing. Of note, six of the nine samples with the highest MYH9-BRD1 fusion read
AC
C
counts (3,911 to 40,668) were confirmed with reverse transcription-PCR and Sanger
sequencing whereas three of nine samples with the lowest MYH9-BRD1 fusion read counts
(1,394 to 2,134) were not confirmed. Additionally, one bone marrow aspirate without a MYH9BRD1 fusion was analyzed; this sample was negative by reverse transcription-PCR and Sanger
sequencing.
Acrometrix Sample
ACCEPTED MANUSCRIPT
For the Acrometrix control sample, 290 of the 293 SNVs covered by the panel were correctly
called (99.0% sensitivity; 95% CI, 97.0% to 99.8%), and there were eight false positives
(>99.9% specificity; 95% CI >99.9% to 100%), for a positive predictive value (PPV) of 97.3%
RI
PT
(95% CI, 94.8% to 98.6%). All 22 InDels were called by the LAB variant caller (100% sensitivity;
95% CI, 84.6% to 100%) with two false positives (>99.9% specificity; 95% CI, >99.9% to 100%)
and a PPV of 91.7% (95% CI, 73.0% to 99.0%). By comparison, for InDel calling, Ion Reporter
SC
would have yielded five false negatives (77.3% sensitivity; 95% CI, 54.6% to 92.2%) and one
M
AN
U
false positive (>99.9% specificity, 95% CI >99.9% to 100%).
Analytical Sensitivity for SNVs, MNVs, and InDels
Extracted DNA from 82 clinical samples that had been previously characterized by Sanger
sequencing (18 samples) or NGS (64 samples) was used to test for SNVs, MNVs, and InDels.
This cohort had a total of 99 SNVs, three MNVs, and 49 InDels. InDels ranged in size from one
TE
D
to 24 bp in length. Sample types for this analysis included FFPE tissue (34 samples) and frozen
tissue (six samples) as well as peripheral blood (22 samples), bone marrow aspirate (14
samples), cell line (three samples), cytololgy smear (two samples), and extracted DNA from a
EP
skin lesion (one sample). Information regarding each sample including sample type, tumor type,
specific mutations detected, and reference method used is given in Supplemental Table S2.
AC
C
This sample set was chosen to represent a wide range of both clinically important and
analytically challenging variants.
Table 1 shows the DNA variant concordance across the sample types tested in this validation.
Overall, the variant concordance was: 98 of 99 SNVs (99.0% sensitivity; 95% CI, 94.5% to
99.9%), and 44 of 49 InDels (89.8% sensitivity; 95% CI, 77.8% to 96.7%). Additionally, three of
three MNVS were detected. It should be noted that all of the discordant InDels were observed in
ACCEPTED MANUSCRIPT
the bone marrow and peripheral blood samples; and were due to the presence of InDels in
ASXL1 and ASXL2 in acute myeloid leukemia and myelodysplastic syndrome samples as well
as a sample submitted with the diagnosis of anemia (Supplemental Table S2). Four of five
RI
PT
InDels not detected with the OncoKidsSM panel were the same ASXL1 c.1934dup (p.G646fs)
variant in four different samples (Supplemental Table S2). This variant arises from the
duplication of a single guanine (G) nucleotide within an eight G homopolymer tract. The other
SC
discordant InDel was ASXL2 c.2090dup (p.G698fs). This variant also arises from the duplication
of a single guanine nucleotide within a homopolymer tract, in this instance, the homopolymer
M
AN
U
tract is composed of five G nucleotides. All other InDel variants, including those present in FFPE
samples, were concordant. Figure 1 demonstrates an InDel called by the LAB variant caller. The
sole discordant SNV, RB1 c.1330C>T (p.Q444*), was not observed (5,887 reads) with
OncoKidsSM. In the original sample this variant was only seen in 64 of 1,213 reads (5.3% variant
allele frequency). This SNV may represent a low-level sequencing artifact in the original sample,
TE
D
however, this variant call was categorized as discordant.
Detection of Gene Fusions in Hematologic Samples with OncoKidsSM
EP
The solid tumor FFPE samples had variable RNA quality compared to the excellent RNA quality
and quantity for all hematologic samples. Due to this difference, analytical sensitivity was
AC
C
calculated separately for hematologic and solid tumor fusions. The unique fusions in the
hematologic samples are shown in Table 2 and Supplemental Table S3. Forty seven of 51
samples (92.2%; 95% CI, 81.1% to 97.8%) demonstrated concordant results. All four discordant
cases had documented ETV6-RUNX1 fusions by FISH. Currently, the OncoKidsSM RNA
sequencing assay only covers the ETV6-RUNX1 major breakpoint (ETV6 ENST00000396373
exon 5 - RUNX1 ENST00000437180 exon 3) which is estimated to account for approximately
90% of ETV6-RUNX1 fusions.21 Reverse transcription-PCR and Sanger sequencing illustrated
that all four discordant samples harbored fusions at the minor breakpoint between ETV6 exon 5
ACCEPTED MANUSCRIPT
and RUNX1 exon 4. The fusion read counts for the 47 detected fusions in the hematologic
samples ranged from 6,623 reads to 589,930 reads. It should be noted that all samples except
Detection of Gene Fusions in Solid Tumors with OncoKidsSM
RI
PT
one were bone marrow aspirate (Supplemental Table S3).
Forty-one solid tumor samples with gene fusions were examined. The unique solid tumor
SC
fusions are shown in Table 3. The results from 34 of 41 (82.9%; 95% CI, 68.0% to 92.9%) solid
tumors were concordant between OncoKidsSM and orthogonal testing. A representative case is
M
AN
U
shown in Figure 2. Of note, five samples had sections preserved both as frozen tissue and
FFPE tissue (Supplemental Table S4); all of these fusions were concordant with testing by
OncoKidsSM. The fusion read counts for the 34 solid tumor fusions identified counts ranged from
73 reads to 741,156 reads. Sample types for this analysis included FFPE tissue (29 samples)
and frozen tissue (12 samples). Information regarding each sample including sample type,
Table S4.
TE
D
tumor type, specific mutations detected, and reference method used is given in Supplemental
EP
Of the discordant samples, three of seven harbored KIAA1549-BRAF fusions. All three of the
discordant samples shared the same breakpoint, KIAA1549 ENST00000242365 exon 15 and
AC
C
BRAF ENST00000288602 exon 9. This breakpoint was confirmed with reverse transcriptionPCR and Sanger sequencing in these samples. The KIAA1549 exon 15 – BRAF exon 9 fusion
was detected at low levels in two of the three frozen samples and in one of the three FFPE
samples. For exon numbering, it should be noted that KIAA1549 exon 15 based on Ensembl
transcript ID 00000242365 is also described as KIAA1549 exon 16 with RefSeq transcript IDs
NM_001164665.1 and NM_020910.2.
ACCEPTED MANUSCRIPT
Two samples (EWSR1 fusions) where the expected fusion was not identified on the OncoKidsSM
RNA sequencing panel showed marginal sample quality (samples 147 and 161, Supplemental
Table S4). Marginal sample quality was defined as those samples with a DV200 less than 50%
RI
PT
or stock library concentration less than 100 pM. In total, seven marginal quality samples were
included in this section of the validation. It should be noted that the expected fusion was
detected for the other five marginal quality samples tested (samples 145, 148, 150, 155, and
SC
160a; Supplemental Table S4). Therefore it is possible that these fusions (ie, EWSR1
rearrangements) were not detected for a different reason, such as a lack of coverage of the
these two samples are unclear.
M
AN
U
fusion breakpoints. As the reference results were generated by FISH, the exact breakpoints of
Two samples were positive for CIC gene fusions by FISH (samples 168 and 169, Supplemental
Table S4) but the expected fusion was not identified with the OncoKidsSM RNA sequencing
TE
D
panel. Reverse transcription-PCR and Sanger sequencing confirmed CIC-DUX4 fusions and
breakpoints in the 5' portion of CIC exon 20.These breakpoints have been reported previously in
a small case series of CIC–DUX4 fusion‐positive round‐cell sarcomas but they are not yet
EP
included in the panel.22
AC
C
Limit of Detection Studies
The results for both DNA dilution series are shown in Supplemental Figure S1. These results
indicate that all SNVs (11 of 11) were detected successfully at the 12.5% dilution point; this is
consistent with an SNV limit of detection of approximately 6%. The detected SNV variant allele
frequencies ranged from 5.5% to 7.5% at the 12.5% dilution level. Additionally, the RB1
c.1450_1451del InDel was detected to 2.3% variant allele frequency. This finding suggests that
InDels can be detected below the 10% variant allele frequency range used as a reporting cutoff.
Finally, two amplified genes (MDM2 and MYC) were used to assess the LOD for amplification.
ACCEPTED MANUSCRIPT
Overall, the LOD results show a steady diminution of copy number at each dilution point from an
estimated copy number of 11.4 for MDM2 and 5.7 for MYC at 100% tumor sample to an
RI
PT
essentially normal copy number state at the 6.25% and 3.13% dilution points.
Samples with known fusions were diluted to low levels to confirm the ability to detect a fusion in
a low number of reads. Supplemental Table S5 shows the LOD results for three samples with
SC
different fusions diluted to 12.5%, 6.25%, 3.13%, 1.57%, and 0.16%. For this study, two
samples with a highly expressed fusion (PML-RARA and PAX3-FOXO1) were chosen along
M
AN
U
with a sample harboring a fusion expressed at a lower level (KIAA1549-BRAF). Samples 86 and
133 with highly expressed fusions both demonstrated robust fusion detection (>10,000) reads at
the 0.16% dilution point. Sample 139 with the KIAA1549-BRAF fusion was detected down to the
1.57% dilution point where it was seen in five reads and not detected at the 0.16% dilution point.
TE
D
Reproducibility and Repeatability
To examine reproducibility and repeatability for DNA analysis, three samples with clinically
significant SNVs and/or InDels were analyzed on three separate days. On the third day, the
EP
sample was run three times on the same sequencing run to assess repeatability. The results for
this analysis are shown in Supplemental Table S6. These results indicate strong reproducibility
AC
C
and repeatability for SNVs and InDels. For RNA analysis, two samples with clinically significant
fusions were each run on three different days (Supplemental Table S7). On the fourth day, both
samples were run three times on the same sequencing run. An additional sample was run on
two different days and then run three times on the same sequencing run for the third day
(Supplemental Table S8). Each fusion was detected in each sample in the repeatability and
reproducibility analyses and showed strong concordance across replicates.
ACCEPTED MANUSCRIPT
Amplification Detection with the OncoKidsSM Platform
Data from each of the 24 copy number genes on the OncoKidsSM panel were compared to
chromosomal microarray data (either from Affymetrix Cytoscan HD or Affymetrix Oncoscan;
RI
PT
Thermo Fisher) in a blinded fashion. Data from all genes calls exceeding a hard cutoff of
greater than or equal to eight copies from either array or NGS (OncoKidsSM) are compared in
Supplemental Table S9. Across the four samples, array data showed seven copy number
SC
changes greater than or equal to eight copies; NGS detected six of the seven of these changes
at greater than or equal to eight copies. It should be noted that the discordant CNV in sample
M
AN
U
170b was seen at 4.6 copies by NGS and was therefore elevated, however, this did not meet
the specified requirement of eight copies by NGS in order to make a call. Seven of the eight
genes found to be amplified with an array were detected by NGS at greater than or equal to
eight copies (Supplemental Table S9, 87.5%; 95% CI, 47.4% to 99.7%). The specificity was 88
of 88 (100%; 95% CI, 95.8% to 100%) copy number events at less than eight copies called
TE
D
correctly (data not shown). The positive predictive value was 100% (95% CI, 59.0% to 100%).
(sample 172).
Sample Yields
EP
Figure 3 shows a copy number plot for a sarcoma with amplification of ALK, FGFR4, and MDM2
AC
C
All bone marrow and peripheral blood samples demonstrated an adequate yield of DNA and
RNA. For frozen and FFPE samples, all tested specimens had sufficient material to produce five
scrolls at 20 µm thickness. However, it is important to note that the internal samples used in this
study were provided in a de-identified form by the CHLA biorepository; these samples were preselected by the biorepository to meet the requirement for five scrolls at 20 µm. Two FFPE
samples (142 and 157a; Supplemental Table S4) yielded a DNA concentration below the input
requirement for the assay (2.7 ng/uL) following extraction. Both samples were concentrated
ACCEPTED MANUSCRIPT
prior to library preparation. Therefore, of 192 clinical samples used in this validation, 2/192
(1.0%) initially fell below the required nucleic acid concentration following extraction whereas
190/192 (99.0%) were at or above the required concentration for library preparation following
cell density, and tumor content of the sample.
SC
Tumors with FFPE and Frozen Tissue Sections
RI
PT
extraction. In clinical practice, the number of scrolls required will depend on the surface area,
Both FFPE and frozen tissue sections were available for six soft tissue tumors/sarcomas in this
M
AN
U
study. Five of these tumors harbored a gene fusion. The pathologic diagnoses and sample
types for these five tumors are given in Supplemental Table S4 (samples 157a, 157b, 158a,
158b, 159a, 159b, 160a, 160b, 162a, 162b). This FFPE/frozen tissue sample set included a
synovial sarcoma in chest wall soft tissue (samples 157a, 157b), an alveolar
rhabdomyosarcoma in a lymph node (samples 158a, 158b), an atypical Ewing's sarcoma
TE
D
excised from the scalp (samples 159a, 159b), a synovial sarcoma excised from the left hand
(samples 160a, 160b), and a congenital mesoblastic nephroma arising from the right kidney
(samples 162a, 162b). Additionally, both FFPE and frozen tissue was available for a metastatic
EP
high grade sarcoma with several gene amplification events demonstrated with chromosomal
microarray; this tumor was excised from the lung (samples 170a and 170B, Supplemental Table
AC
C
S9).
DISCUSSION
The OncoKidsSM NGS panel was designed to capture diagnostic, prognostic, and therapeutic
markers in diverse childhood cancers including pediatric sarcomas, pediatric central nervous
system (CNS) malignancies, pediatric leukemias, and pediatric embryonal tumors. Targeted
aberrations include SNVs, MNVs, InDels, amplification (greater than or equal to eight estimated
ACCEPTED MANUSCRIPT
copies), and RNA-based gene fusions. Importantly, this assay was designed to process low
quantity DNA and RNA from FFPE which is particularly relevant for pediatric tumors that often
are submitted to the laboratory as small biopsies. This panel is currently one of the largest
RI
PT
amplification-based NGS assays available (3,069 DNA amplicons and 1,421 targeted fusions).
More importantly, it was designed with input from pediatric oncologists and pathologists in order
to encompass genomic alterations of clinical value to those involved in the care of pediatric
SC
cancer patients. The intent of this study was to address the unmet need for a pediatric cancer
focused NGS panel that included all known recurring genomic defects in these cancers (as of
M
AN
U
2017). Our local institutional experience to date suggests that we will identify clinically significant
alterations in at least 60% of patients (data not shown).
This test was validated using a diverse population of specimens that included 192
unique clinical samples. The DNA portion of the assay showed excellent technical performance
across key sequencing metrics including average sequencing depth (5,650X) and average
TE
D
uniformity (96.4%). This deep coverage and high uniformity provide confidence for calling low
variant allele frequency variants and amplification, respectively. For RNA performance, 1.88
million average mapped reads were identified; this performance facilitates the detection of
EP
fusions that may be present in a low number of reads.
To characterize the analytical sensitivity for SNV and InDel detection, two approaches
AC
C
were taken. First, an AcroMetrix control sample was used. The sensitivity for this comparison
for SNVs and InDels was 99% and 100%, respectively. Second, the sensitivity was calculated
from known mutations across 82 samples with known pathogenic alterations. It should be noted
that a wide range of clinically important variants were examined in this analysis (Supplemental
Table S2). In addition to samples with a broad spectrum of somatic mutations, several samples
with pathogenic germline alterations were analyzed (eg, RB1, SMARCB1, and TP53 pathogenic
variants). In practice, these findings would prompt targeted Sanger sequencing of a peripheral
blood sample to confirm germline status. If confirmed, these results would change clinical
ACCEPTED MANUSCRIPT
management and/or prompt further testing of at-risk family members. Table 1 shows the
analytical sensitivity for SNVs, MNVs, and InDels across the specimen types analyzed. For this
comparison, the overall sensitivity for SNVs/MNVs was >99% and 90% for InDels. It should be
RI
PT
noted that InDels were called using a custom variant caller, LAB (Local Adjustment for
Background), enabling greater sensitivity (Figure 1). Importantly, all InDels not detected with
OncoKidsSM in this study occurred within long (greater than or equal to five base pairs)
homopolymer regions. Longer homopolymer regions are well-known for creating problematic
SC
artifacts across sequencing platforms,23,24 and therefore pose a challenge for variant calling
M
AN
U
software. It should be noted that a single InDel, ASXL1 c.1934 dup, was not identified in four
different samples (Supplemental Table S2). Although an early report claimed that this variant
represents a sequencing artifact rather than a common somatic mutation,25 follow-up studies
demonstrated that it is indeed a common somatic mutation in myelodysplastic syndromes,
myeloproliferative neoplasms, myelodysplastic/myeloproliferative neoplasms (especially chronic
TE
D
myelomonocytic leukemia), and acute myeloid leukemia.26,27 As these diseases are much
more common in adults than in children, this variant is likely over-represented in this study
relative to its overall prevalence in pediatric malignancies. Accordingly, the analytical sensitivity
EP
for InDel detection is likely higher in actual practice than in this analysis.
The sensitivity and specificity for calling amplification events with OncoKidsSM at a
AC
C
conservative cut-off of eight estimated copies was 88% and 100%, respectively. However, it
should be noted that a small number of samples were tested for this analysis (Supplemental
Table S9). As more samples are processed using both OncoKidsSM and chromosomal
microarray, a more quantitative measure of amplification may be assessed. This assay was not
designed to assess copy number loss, especially exon-level deletions. One limitation of
OncoKidsSM is that common deletions in genes such as IKZF1, which is frequently mutated in
high risk acute lymphoblastic leukemia and often lacks multiple exons, 28,29 will not be detected.
ACCEPTED MANUSCRIPT
A wide range of hematologic and solid tumor samples harboring pathologic fusions were
used to test the analytical sensitivity of RNA fusion detection. In contrast to the solid tumor
samples tested, all hematologic samples demonstrated high quality RNA. However, the overall
RI
PT
performance was excellent for both hematologic malignancy–associated fusions and solid
tumor–associated fusions (Supplemental Tables S3 and S4, respectively). It should be noted
that all discrepant results were due to either non-coverage of a rare breakpoint (ETV6-RUNX1
SC
minor breakpoint, CIC-DUX4), suboptimal primer binding (primer for KIAA1549 exon 15), or
marginal sample quality (EWSR1-rearranged samples 147 and 161 in Supplemental Table S4).
M
AN
U
Additional primers for the ETV6-RUNX1 minor breakpoint and a new KIAA1549 exon 15 primer
have been incorporated into the commercially available version of the assay (Oncomine
Childhood Cancer Research Assay). The informatics pipeline has also been modified to include
four additional hotspot amplicons (in CCND1, FGFR1, MYC, and MYCN) and four additional
genes for amplification (ABL2, JAK1, JAK2, and JAK3) in the DNA component of the assay
TE
D
(Supplemental Table S1).
Tables 2 and 3 show the unique hematologic malignancy–associated fusions and unique
solid tumor–associated fusions, respectively, identified with OncoKidsSM. A large number of
EP
actionable fusions in hematologic malignancies were identified with OncoKidsSM including BCRABL1, PML-RARA, FIP1L1-PDGFRA as well as numerous Ph-like fusions. Of note, Ph-like
AC
C
fusions observed with OncoKidsSM include those targetable with tyrosine kinase inhibitors such
as dasatinib (ABL1, ABL2, and PDGFRB kinase fusions), JAK2 inhibitors (CRLF2 and JAK2
kinase fusions), and tropomyosin kinase inhibitors such as larotrectinib (NTRK3 kinase
fusions).30 Supplemental Table S3 contains information on all Ph-like fusions identified in this
study (samples 99-117). Potentially targetable solid tumor–associated fusions included GOPCROS1, CCDC6-RET, EML4-ALK, and ETV6-NTRK3. Taken together, these results demonstrate
the potential to identify actionable gene fusions across a wide variety of solid and liquid tumors
ACCEPTED MANUSCRIPT
of childhood using OncoKidsSM. In clinical practice, novel fusions will be confirmed with reverse
transcription-PCR and Sanger sequencing prior to reporting.
Notably, NGS testing with OncoKidsSM demonstrates several advantages over FISH or
RI
PT
reverse transcription-PCR testing for pediatric gene fusions in the first-line setting. In tumor
types with multiple actionable or prognostically relevant fusions, such as B-ALL, the use of FISH
or reverse transcription-PCR to comprehensively test for these fusions requires a laboratory to
SC
run and interpret many individual targeted assays. The use of a single assay to cover both
standard of care fusions (eg, ETV6-RUNX1, BCR-ABL1, KMT2A fusions) and other actionable
fusions (eg, Ph-like fusions) dramatically simplifies the clinical workflow for fusion testing. In
M
AN
U
addition, unexpected fusions for a given tumor type are readily identified with OncoKidsSM.
Examples from this study include a FIP1L1-PDGFRA fusion in a T-ALL and a GOPC-ROS1
fusion in a low-grade glioma (sample 84 in Supplemental Table S3 and sample 137 in
Supplemental Table S4, respectively). More broadly, in pediatric tumors where the diagnosis is
TE
D
uncertain, OncoKidsSM enables comprehensive fusion assessment without the need to choose
the correct FISH probe or reverse transcription-PCR primers.
As this panel uses an opposing primer strategy for fusion detection, both partners in the
EP
gene fusion must be defined within an approximate 100 to 200 base pair region. Therefore, one
limitation of the panel is that fusion detection requires PCR primers against both partners in a
AC
C
fusion. For this reason, the common IGH/IGK/IGL rearrangements will not be detected as
OncoKidsSM lacks PCR primers against IGH/IGK/IGL. Several non-Hodgkin's lymphomas, such
as Burkitt lymphoma, are driven by immunoglobulin (IG) gene rearrangements. Although the
OncoKidsSM assay cannot reliably capture IG gene rearrangements for diagnosis of lymphomas,
these translocation-mediated fusions are easily detected in clinical practice by karyotype
analysis and FISH. Overall, it should be noted that a negative fusion result does not preclude
the presence of a gene fusion involving one of the driver fusion partners in this assay
(Supplemental Table S1). Possible reasons for false negative fusion results involving driver
ACCEPTED MANUSCRIPT
fusion partners include the presence of an uncovered gene partner, poor nucleic
quality/quantity, or suboptimal primer binding to a gene fusion partner.
A high level of reproducibility was observed in the precision analysis for both DNA
RI
PT
variants (ie, SNVs, InDels) as well as RNA fusions (Supplemental Tables S6, S7, S8). The limit
of detection analysis for SNVs and InDels showed consistent detection to the 12.5% dilution
point; this is consistent with a limit of approximately 6% variant allele frequency (Supplemental
SC
Figure S1). Although the InDel dilution series showed detection down to 2.4% variant allele
frequency, conservative reporting cutoffs of 6% VAF for SNVs and 10% VAFs for InDels have
M
AN
U
been established to minimize the risk of false positives.
Routine molecular profiling of pediatric malignancies by NGS panels, such as
OncoKidsSM, is expected to provide a significant improvement in the diagnosis and care of
children with pediatric malignancies. For diagnosis, up-front testing can obviate the need to
select specific reverse transcription-PCR assays or FISH probes. Additionally, by identifying
TE
D
targetable alterations in pediatric malignancies in the first-line setting, OncoKidsSM enables the
use of precision therapies prospectively where they are most likely to be effective. For these
reasons, broad molecular profiling of all pediatric malignancies by tests such as OncoKidsSM
EP
should become standard of care in the near future.
In conclusion, the OncoKidsSM assay demonstrates robust performance characteristics
AC
C
across several dimensions including analytical sensitivity, detection sensitivity, and
reproducibility. Features of this assay include coverage of biomarkers across a wide range of
pediatric tumor types along with the ability to process FFPE tissue. Taken together, these
results support the analytical validity and value of routine clinical testing of both solid and liquid
tumors from pediatric patients with the OncoKidsSM assay.
ACCEPTED MANUSCRIPT
ACKNOWLEDGEMENTS
We thank the childhood cancer panel team at Thermo Fisher Scientific for excellent support; the
following individuals for kindly providing samples or assisting in the procurement of samples: Dr.
RI
PT
Jennifer Cotter (Children's Hospital Los Angeles), Andrea Fischbeck (Children's Hospital Los
Angeles), Monica Mendez (Children's Hospital Los Angeles), Dr. Shalini Reshmi (Nationwide
Children's Hospital), Dr. Margaret Macy (Children’s Hospital of Colorado), Dr. Lindsey Hoffman
SC
(Children’s Hospital Colorado), Dr. Julie Yin (Children’s Healthcare of Atlanta), Dr. Phillipe
Szankasi (ARUP), Dr. Larissa Furtado (ARUP), Dr. Julia Bridge (University of Nebraska Medical
M
AN
U
Center), David Puskas (Nationwide Children’s Hospital), Dr. Jennifer Morrissette (Hospital of the
University of Pennsylvania), Dr. Amanda Oran (Hospital of the University of Pennsylvania), Dr.
Robyn Sussman (Hospital of the University of Pennsylvania), and David Lieberman (Hospital of
the University of Pennsylvania); Cindy Fong, Jennifer Han, and Chern-Yu Yen for technical
AC
C
EP
TE
D
assistance; and Dr. Ryan Schmidt for assistance with figures.
ACCEPTED MANUSCRIPT
REFERENCES
statistics, 2014. CA: CA Cancer J Clin 2014, 64:83-103
RI
PT
1. Ward E, DeSantis C, Robbins A, Kohler B, Jemal A. Childhood and adolescent cancer
2. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Kinzler KW. Cancer
Genome Landscapes. Science 2013, 339:1546-58
SC
3. Roberts C, Biegel J. The role of SMARCB1/INI1 in the development of rhabdoid tumors.
Cancer Biol Ther 2009, 8:412-6
M
AN
U
4. Carpenter EL, Mossé YP. Targeting ALK in neuroblastoma—preclinical and clinical
advancements. Nature Rev Clin Oncol 2012, 9:391-9
5. Schiffman JD, Hodgson JG, VandenBerg SR, Flaherty P, Polley MY, Yu M, Fisher PG,
Rowitch DH, Ford JM, Berger MS, Ji H. Oncogenic BRAF mutation with CDKN2A
inactivation is characteristic of a subset of pediatric malignant astrocytomas. Cancer Res
TE
D
2010, 70:512-9
6. Nikiforova MN, Laframboise WA, Nikiforov YE. Clinical Genomics: A Guide to Clinical Next
Generation Sequencing. Edited by Kulkarni S and Pfeifer J. Waltham, MA, Academic Press,
EP
2015, pp. 57-66
7. Samorodnitsky E, Jewell BM, Hagopian R, Miya J, Wing MR, Lyon E, Damodaran S,
AC
C
Bhatt D, Reeser JW, Datta J, Roychowdhury S. Evaluation of Hybridization Capture Versus
Amplicon-Based Methods for Whole-Exome Sequencing. Hum Mutat 2015, 36:903-14
8. Oberg JA, Bender JL, Sulis ML, Pendrick D, Sireci AN, Hsiao SJ, Turk AT, Cruz FS,
Hibshoosh H, Remotti H, Zylber RJ. Implementation of next generation sequencing into
pediatric hematology-oncology practice: moving beyond actionable alterations. Genome
Med 2016, 23:133
ACCEPTED MANUSCRIPT
9. Harris MH, DuBois SG, Bender JL, Kim A, Crompton BD, Parker E, Dumont IP, Hong AL,
Guo D, Church A, Stegmaier K. Multicenter feasibility study of tumor molecular profiling to
inform therapeutic decisions in advanced pediatric solid tumors: The Individualized Cancer
RI
PT
Therapy (iCat) Study. JAMA Oncol 2016, 2:608-15
10. Abel HJ, Al-Kateb H, Cottrell CE, Bredmeyer AJ, Pritchard CC, Grossman AH, Wallander
ML, Pfeifer JD, Lockwood CM, Duncavage EJ. Detection of gene rearrangements in
SC
targeted clinical next-generation sequencing. J Mol Diag 2014, 16:405-17
11. Sangkhathat S. Current management of pediatric soft tissue sarcomas. World J Clin Pediatr
M
AN
U
2015, 8:94
12. Arva NC. Diagnostic Challenges in Pediatric Round Blue Cell Tumors of the Bone. Diagn
Pathol Open 2016, 1:2
13. Vohra HA, Davies S, Vohra H, Rosin MD, Snead DR. Primary synovial sarcoma of the
pleura: beware of misdiagnosis. European J Intern Med 2004, 15:465-6
TE
D
14. Cheng Y, Sheng W, Zhou X, Wang J. Pericardial synovial sarcoma, a potential for
misdiagnosis: clinicopathologic and molecular cytogenetic analysis of three cases with
literature review. American J Clin Path 2012, 137:142-9
EP
15. Steppan A, Pratilas CA, Loeb DM. Targeted therapy for soft tissue sarcomas in adolescents
and young adults. Adolesc Health Med Ther 2017, 8:41-55.
AC
C
16. Chung C, Ma H. Driving Toward Precision Medicine for Acute Leukemias: Are We There
Yet? Pharmacotherapy 2017, 37:1052-1072
17. Santiago R, Vairy S, Sinnett D, Krajinovic M, Bittencourt H. Novel therapy for childhood
acute lymphoblastic leukemia. Expert Opin Pharmacother 2017, 18:1081-1099
18. Ryall S, Tabori U, Hawkins C. A comprehensive review of paediatric low-grade diffuse
glioma: pathology, molecular genetics and treatment. Brain Tumor Pathol 2017, 34:51-61
19. Khotskaya YB, Holla VR, Farago AF, Shaw KR, Meric-Bernstam F, Hong DS. Targeting
TRK family proteins in cancer. Pharmacol Ther 2017, 173:58-66
ACCEPTED MANUSCRIPT
20. Babiceanu M, Qin F, Xie Z, Jia Y, Lopez K, Janus N, Facemire L, Kumar S, Pang Y, Qi Y,
Lazar IM. Recurrent chimeric fusion RNAs in non-cancer tissues and cells. Nucleic Acids
Res. 2016, 44:2859-72.
RI
PT
21. Jin Y, Wang X, Hu S, Tang J, Li B, Chai Y. Determination of ETV6‐RUNX1 genomic
breakpoint by next‐generation sequencing. Cancer Med 2016, 5:337-51
22. Gambarotti M, Benini S, Gamberi G, Cocchi S, Palmerini E, Sbaraglia M, Donati D, Picci P,
SC
Vanel D, Ferrari S, Righi A. CIC–DUX4 fusion‐positive round‐cell sarcomas of soft tissue
and bone: a single‐institution morphological and molecular analysis of seven cases.
M
AN
U
Histopathology 2016, 69:624-34
23. Quail MA, Smith M, Coupland P, Otto TD, Harris SR, Connor TR, Bertoni A, Swerdlow HP,
Gu Y. A tale of three next generation sequencing platforms: comparison of Ion Torrent,
Pacific Biosciences and Illumina MiSeq sequencers. BMC Genomics 2012, 13:341
24. Laehnemann D, Borkhardt A, McHardy AC. Denoising DNA deep sequencing data—high-
TE
D
throughput sequencing errors and their correction. Brief Bioinform 2015, 17:154-79
25. Abdel-Wahab O, Kilpivaara O, Patel J, Busque L, Levine RL. The most commonly reported
variant in ASXL1 (c. 1934dupG; p.Gly646TrpfsX12) is not a somatic alteration. Leukemia
EP
2010, 24:1656
26. Jankowska AM, Makishima H, Tiu RV, Szpurka H, Huang Y, Traina F, Visconte V, Sugimoto
AC
C
Y, Prince C, O'Keefe C, Hsi ED. Mutational spectrum analysis of chronic myelomonocytic
leukemia includes genes associated with epigenetic regulation: UTX, EZH2, and DNMT3A.
Blood 2011, 118:3932-41
27. Gelsi-Boyer V, Brecqueville M, Devillier R, Murati A, Mozziconacci MJ, Birnbaum D.
Mutations in ASXL1 are associated with poor prognosis across the spectrum of malignant
myeloid diseases. J Hematol Oncol 2012, 5:12
28. Yokota, T, Kanakura Y. Genetic abnormalities associated with acute lymphoblastic
leukemia. Cancer Sci 2016, 107:721-5.
ACCEPTED MANUSCRIPT
29. Mullighan CG, Miller CB, Radtke I, Phillips LA, Dalton J, Ma J, White D, Hughes TP, Le
Beau MM, Pui CH, Relling MV, Shurtleff SA, Downing JR. BCR-ABL1 lymphoblastic
leukemia is characterized by the deletion of Ikaros. Nature 2008, 453:110-14
RI
PT
30. Roberts KG, Li Y, Payne-Turner D, Harvey RC, Yang YL, Pei D, McCastlain K, Ding L, Lu C,
Song G, Ma J. Targetable kinase-activating lesions in Ph-like acute lymphoblastic leukemia.
AC
C
EP
TE
D
M
AN
U
SC
N Engl J Med 2014, 371:1005-15
ACCEPTED MANUSCRIPT
Figure Legends
Figure 1. Example of InDel called by the in-house variant caller, Local Adjustment for
Background (LAB). Note the insertion variant (represented as a purple bar and indicated with
RI
PT
the red arrow) in addition to the deletion artifact (represented as a horizontal bar and indicated
with the green arrow). This insertion variant (ATRX c.4744dup) was not called by the Ion
Reporter 5.2 variant caller likely due to the long homopolymer tract at this position.
SC
Figure 2. Detection of gene fusions with OncoKidsSM. The fusion (C11orf95-RELA) is
M
AN
U
visualized with an in-house software. The 3’ portion of C11orf95 exon three is highlighted in
green whereas the 5’ portion of RELA exon two is highlighted in red. Forward reads are in dark
blue and reverse reads are in light blue. Yellow cells indicate bases that differ from the
reference sequence, cells with a horizontal bar represent deletion variants, and pink vertical
TE
D
lines represent insertions.
Figure 3. Detection of gene amplification with OncoKidsSM. The estimated copy number of
genes is visualized with an in-house software. In this example, ALK, FGFR4, and MDM2
AC
C
EP
demonstrate an estimated copy number greater than eight fold (red dotted line).
ACCEPTED MANUSCRIPT
Peripheral Blood
Bone Marrow
Frozen Tissue
FFPE Tissue
Cell Line
Extracted DNA
from Skin Sample
SNV
Concordance
33/34 (97%)
30/30 (100%)
2/2 (100%)
30/30 (100%)
3/3 (100%)
N/A
MNV
Concordance
N/A
N/A
N/A
3/3 (100%)
N/A
InDel
Concordance
6/8 (75%)
12/15 (80%)
4/4 (100%)
21/21 (100%)
N/A
M
AN
U
Sample Type
SC
NGS assays and OncoKidsSM across sample types.
RI
PT
Table 1. Concordance of SNVs, MNVs, and InDels between Sanger sequencing or clinical
N/A
1/1 (100%)
This sample set included 79 clinical samples as well as three cell lines. Discordant InDels in
peripheral blood and bone marrow samples were long (greater than five base pairs)
AC
C
EP
TE
D
homopolymeric mutations in ASXL1 or ASXL2.
ACCEPTED MANUSCRIPT
SC
M
AN
U
TE
D
EP
AC
C
ATF7IP-JAK2
BCR-ABL1
BCR-JAK2
EBF1-PDGFRB
ETV6-ABL1
ETV6-JAK2
ETV6-NTRK3
ETV6-RUNX1
FIP1L1-PDGFRA
FOXP1-ABL1
KMT2A-AFF1
KMT2A-MLLT3
NUP214-ABL1
NUP98-NSD1
P2RY8-CRLF2
PAG1-ABL2
PAX5-JAK2
PML-RARA
RANBP2-ABL1
RBM15-MKL1
RCSD1-ABL1
RCSD1-ABL2
SSBP2-JAK2
STIL-TAL1
TERF2-JAK2
ZC3HAV1-ABL2
ZEB2-PDGFRB
ZMIZ1-ABL1
RI
PT
Table 2. Unique fusions detected with the OncoKidsSM panel in hematologic
malignancies.
ACCEPTED MANUSCRIPT
Table 3. Unique solid tumor associated fusions identified with the OncoKidsSM panel.
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
C11orf95-RELA
CCDC6-RET
EML4-ALK
ETV6-NTRK3
EWSR1-ETV1
EWSR1-FLI1
EWSR1-NR4A3
EWSR1-WT1
FUS-CREB3L2
FUS-DDIT3
GOPC-ROS1
KIAA1549-BRAF
NPM1-ALK
PAX3-FOXO1
SS18-SSX1
SS18-SSX2
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
C11orf95 exon 3
RELA exon 2
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
Документ
Категория
Без категории
Просмотров
0
Размер файла
4 994 Кб
Теги
2018, jmoldx, 009
1/--страниц
Пожаловаться на содержимое документа