AMP Abstracts
Genetics
G01. Validation of the Ion S5 and Ion Chef for Cystic Fibrosis Mutation
Analysis
T.R. Sundin, D. Seidman, K. Slaughter, M. Aunchman, L. Grissom, J. Inman, R.
Post, A. Shean, S. Nasim
Sentara Healthcare, Norfolk, VA.
Introduction: Cystic Fibrosis (CF) is a common autosomal recessive genetic
disorder that is caused by mutations in the CFTR (cystic fibrosis transmembrane
conductance regulator) gene. There have been more than 1900 different mutations
found within the CFTR gene. Current commercially available CFTR mutation panels
are limited in the number of clinically significant mutations that can be detected. In
this study, performance of CFTR mutation analysis from whole blood using the Ion
S5 next-generation sequencing (NGS) and the Ion Chef was evaluated.
Methods: Samples (n=45) were tested utilizing a combination of retrospective whole
blood samples previously genotyped using the xTAG Cystic Fibrosis (CFTR) 39 kit
v2 (n=35) and DNA from previously characterized Coriell Institute samples (n=10).
DNA from these samples was extracted using Roche MagNA Pure 96 system with
the DNA Isolation Kit DNA/Viral NA SV 2.0. All DNA samples were quantified using
the Qubit dsDNA HS assay on the Qubit 3.0 Fluorometer. Ten nanograms of DNA
from each sample was used to prepare pooled barcoded libraries on the Ion Chef
using Ion AmpliSeq Kit for Chef DL8 with the Ion Ampliseq CFTR Panel. Pooled
libraries were combined and diluted to 50 pM and templated on the Ion Chef with the
Ion 520 & Ion 530 Kit-Chef and loaded onto an Ion 520 chip. The Ion 520 chip was
sequenced on the Ion S5. Torrent Suite software was used to analyze the data and
compare to the CFTR human genomic sequence as a reference (hg19). A filter chain
was applied to the data to analyze the CFTR gene for 97 mutations known to cause
CF. Results: Analysis NGS data from 45 specimens, revealed CFTR gene analysis
by NGS has excellent analytical sensitivity and analytical specificity [100% positive
predictive value (PPV)/ 100% negative predictive value (NPV)]. The sample set
included 25 patients with previously identified mutations in the CFTR gene and 20
patients with wild-type CFTR genes. Inter-assay variability was assessed by running
the same 3 specimens on 2 independent runs. Intra-assay variability was
determined by analyzing 3 specimens in duplicate within the same run. Analysis of
the NGS data produced the expected result for all samples. Conclusions: These
results demonstrate the benefit of CFTR mutation analysis from whole blood using
the Ion S5 and the Ion Chef. Total technologist hands-on time is 3 hours for 24
samples. The ease of the Ion Chef workflow allows NGS to become feasible for a
mid-sized clinical laboratory. Use of this NGS assay will ensure efficient, reliable and
reproducible analysis of the CFTR gene.
G02. Reinterpreting Previously Reported Genetic Variants is Clinically
Significant
J.A. SoRelle1, D.M. Thodeson1, S. Arnold1, J. Park2
1University of Texas Southwestern Medical Center, Dallas, TX; 2Children's Medical
Center and University of Texas Southwestern Medical Center, Dallas, TX.
Introduction: Next-generation sequencing (NGS) testing has become a common
diagnostic tool for evaluating pediatric patients with intractable epilepsy. NGS testing
can comprehensively identify gene variants of interest; however, determining the
clinical significance of novel variants is challenging. Advances in publically available
databases and standardized interpretation criteria have improved the clinical
interpretation of variants. However, the utility of re-interpreting previously reported
NGS epilepsy tests has not been systematically evaluated. In this study we identify
the frequency and significance of variant reclassification from previously reported
NGS epilepsy gene panels. Methods: All epilepsy NGS epilepsy gene panel reports
from patients at a tertiary-care pediatric hospital were retrospectively reviewed (July
2012 to August 2015). Reports were from one reference laboratory (GeneDx), which
only reported variants as pathogenic/likely pathogenic (P-LP), or uncertain in
significance (VUS). Previously reported variants were re-evaluated using population
(ExAC, 1000 Genomes phase 3) and clinical (HGMD, ClinVar) databases. Any
variant with a high population frequency (>1%) or a new report in a clinical database
was reinterpreted using ACMG variant classification criteria. Results: Three-hundred
and 7 patients were previously tested yielding 308 variants in 184 patients. Of the
308 variants, 19% were P-LP (n=59), and 81% were VUS (n=249). 19% (n=11) of
the P-LP variants were reclassified; variants were downgraded to VUS (n=4), or BLB (n=6), and one was upgraded to pathogenic (KCNQ2) Of the 249 VUS reported,
14 did not have sufficient variant information (transcript, genomic coordinate, other)
to be reinterpreted. Of the remaining 235 VUS, 31% (n=72) were reclassified.
Variants were downgraded to B-LB (n=65) or upgraded to P-LP
(n=7, MECP2, TBC1D24, TSC1, PRICKLE1, and SCN8A). Overall, 66 of 307
patients (21.5%) were impacted by variant reclassification. Reclassification rate
decreased from tests initially reported in 2012 (37%) to 2015 (25%) (m= -4%/yr,
R2=0.99, p=0.007). Conclusions: Based on our single institution experience, a large
proportion of variant classifications were revised (26.9%; 83/308). These
reclassifications impacted 21% of patients (66 /307). A 3% gain in diagnostic yield
was obtained by reinterpreting previously reported VUS. In addition, many
reclassified VUS were downgraded in clinical significance. Overall, these findings
indicate that all P-LP and VUS variants from previously reported NGS tests should
be routinely reviewed and reinterpreted.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
G03. Hypertrophic Epicardial Adipose Tissue is a Source of EPAC Proteins
Directly Associated to ST2 Production and Heart Dilation and may be a
Potential Index of Heart Remodeling in CVDs Patients
M.M. Corsi Romanelli1 , E. Dozio2 , L. Tacchini2 , G. Schmitz3 , E. Vianello2
1University of Milan and IRCC Policlinico San Donato, Milan, Italy; 2 University of
Milan, Milan, Italy; 3University Hospital Regensburg, Regensburg, Germany.
Introduction: Epicardial adipose tissue (EAT) is a myocardial fat from which
released molecules can directly reach the heart. In pathological conditions,
the ubiquitously tissue hypertrophy mediators are the exchange proteins directly
activated by cAMP (EPACs), named EPAC1 and EPAC2. In the heart the main
protective signalling against detrimental remodelling is the ST2 and IL33 molecules.
ST2 exists both as transmembrane receptor (ST2L) and soluble (sST2) form and
in case of physiological stretch ST2L bind IL33 promoting anti-fibrotic signals.
Contrarily in CVDs, sST2 is up regulated functioning as scavenger of IL33 and
promoting heart dilatation. Interesting is that ST2 can be also produced by adipose
tissue in normal condition. Due to EPACs properties to induce hyperplasia, our
hypothesis is that larger EAT cells may also produce sST2 that can local amplify its
detrimental role on myocardium. For these reasons we wanted to verify in CVDs
patients first if larger EAT cells are able to up-regulate EPAC proteins and second if
EPACs may be associate to sST2 EAT production and heart
dilatation. Methods: Fifty CVD patients are enrolled and stratified according to EAT
median thickness (8mm). Plasma and EAT biopsies were collected during surgery.
Indexed left ventricular mass (hLVM), end-diastolic posterior wall (EDPW), relative
wall thickness (RWT), left ventricular mass (LVM) values were used as cardiac
dilatation indexes commonly approved in clinical practice. Gene expression and
protein assays were performed to investigate EPACs, ST2, IL33 mRNA and protein
production. Results: Our data demonstrated that CVDs patients with EAT >8mm
have significantly positive correlation with RWT and they also presented higher
EPAC1 and ST2 mRNA and protein levels than CVDs patients <8mm; contrarily IL33
protein is down regulated in CVDs >8mm and up-regulated in CVDs patients < 8mm.
CVDs patients with hypertrophic EAT both EPAC1 and EPAC2presented positive
correlation with hLVM, EDPW, LVM indexes and ST2 mRNA
levels. Conclusions: Our results demonstrated in EAT, EPACs are directly
associated to EAT hyperplasia and sST2 local production suggesting their
implication in detrimental heart hyperplasia due to the positive correlation between
EPAC1 and EPAC2 versus cardiac indexes and ST2. From these results we can
suggest that EAT thickness can be a potential newer parameter of detrimental heart
remodelling in the prevention of CVDs complications.
G04. Discovery of a Novel, Accurate Tagging SNP for HLA-B*15:02 Screening
Before Carbamazepine Therapy in the Multiethnic United States Population
H. Fang, X. Xu, K. Kaur, M. Dedek, G. Zhu, B. Riley, F. Espin, A. Del Tredici, T.A.
Moreno
Millennium Health, San Diego, CA.
Introduction: The gene HLA-B*15:02 is strongly associated with life-threatening
severe skin hypersensitivity reactions such as Stevens-Johnson Syndrome and toxic
epidermal necrolysis in patients treated with carbamazepine (CBZ) and structurallyrelated medications. FDA-approved labeling information for CBZ recommends
screening for HLA-B*15:02 prior to CBZ therapy in patients of Asian ancestry due to
its high frequency in Asian populations. In this study, we aimed to a) identify a direct
and accurate method for the detection of HLA-B*15:02 as a screening tool, and b)
evaluate the prevalence of HLA-B*15:02 in a large cohort of US patients. Methods:
HLA-B references deposited with the World Health Organization Nomenclature
Committee for Factors of the HLA Systems were compared to identify a candidate
tagging single-nucleotide polymorphism (SNP) for HLA-B*15:02. Data from public
research databases as well as from approximately 30,000 patient samples were
used to test concordance of HLA-B*15:02 with the candidate tagging SNP. Ethnicity
information of patients with positive HLA-B*15:02 results were evaluated with respect
to current information about ethnicity and HLA-B*15:02 carriership. Results: We
identified a SNP in HLA-B as a candidate tagging SNP for HLA-B*15:02. Samples
from the 1,000 Genomes Project that were typed for HLA (n = 955) and sequenced
fully showed 100% concordance between HLA-B*15:02 and the minor allele of the
tagging SNP. Furthermore, in a large study cohort of US individuals (n = 29,194),
sequencing data demonstrated that the minor allele was a reliable marker for HLAB*15:02 detection with 100% sensitivity and 99.97% specificity (P = 0.0001 by twotailed Fisher’s exact-test). Notably, out of 158 positive individuals detected in this
study, 65 had physician-reported ethnicity information, including 28 (43%) Asians, 14
African Americans (22%), 11 Caucasians (17%), 2 Hispanics (3%), and 10 ‘Other’
ethnicity (15%). Conclusions: We have discovered a novel tagging SNP within
the HLA-B gene that is useful as a direct and accurate method for HLAB*15:02 screening. The SNP is strongly linked with HLA-B*15:02, with all true
positive carriers in this study also showing carriership. Our data demonstrates that it
is a good screening tool to identify potential HLA-B*15:02 carriers in the US
population. Finally, we show that screening patients for HLA-B*15:02 based on
observed or self-reported Asian ancestry may be insufficient to identify a large
portion of potential carriers of this allele in the ethnically diverse US population.
943
AMP Abstracts
G05. Spectrum of Mutations in Hbb Gene Among Thalassemia Major Patients
in a Cohort of Nepalese Population
S. Thapa, A.J. Kunwar
Kathmandu Center for Genomics and Research Laboratory, Lalitpur, Nepal.
Introduction: The thalassemias are the most common monogenic disorders with a
genetically determined reduction in the rate of one or more types of normal
haemoglobin polypeptide chain resulting in a decrease in the amount of haemoglobin
involving the affected chain. Beta thalassemia is a highly heterogeneous disorder in
its phenotype, geographical distribution and molecular mechanism. Methods: DNA
was extracted from the 26 clinically diagnosed blood sample and Amplification
Refractory Mutation System - Polymerase Chain Reaction (ARMS PCR) was used
for amplification to analyze mutations in the hbb gene and 2% gel electrophoresis
was used for visualization of PCR products. Results: Among 26 β-thalassemia
major patients, 13 (50%) had IVS 1-5 (G>C) mutation, 8 (30.76%) had 619bp
deletion, 2 (7.69%) CD 8/9 (+G), 1 (3.84%) CD 15 (G>A), 1 (3.84%) had -88 (C>T)
mutation whereas CD41/42 (-TCTT) was not detected in any of the patients. Among
the patients with 619 bp deletions 2 (25%) were homozygotes and 6 (75%) were
heterozygotes. Conclusions: This is the baseline study to assist in the regulation of
proper new health policies which will impact in the proper diagnosis and treatment.
G06. Custom NGS Panels from Optimized Gene Sets for Inherited Disease
Research
M. Andersen1, S. Roman1, A. Kothandaraman2, A. Broomer1, T. Biorac1, D.
Mandelman1, F. Hyland2, M. Manivannan2, Y. Fu1, Y. Zhu1
1Thermo Fisher Scientific, Carlsbad, CA; 2Thermo Fisher Scientific, South San
Francisco, CA.
Introduction: AmpliSeq is a next generation sequencing library preparation method
for targeted resequencing that utilizes highly multiplexed PCR to amplify regions of
interest. A key to successful AmpliSeq libraries is the primer panel used for target
amplification. Until now primers have been available as ready-to-use panels or as
custom made-to-order panels. We describe a new process for creating customized
panels from optimized PCR primers called Ion AmpliSeq On-Demand
panels. Primers for On-Demand panels are available as whole gene sets (all of the
primers needed to create libraries that cover the entire coding regions of genes) and
are selectable on the AmpliSeq.com website by either uploading gene lists or
choosing genes from the content selection engine base on disease research
areas. Predicted gene coverage metrics are also available on the website as part of
the content selection process. Methods: Ten disease research On-Demand panels
were designed on AmpliSeq.com using the content selection engine. Libraries were
prepared from a variety of sample types (purified reference genomes, fresh, frozen,
and dried blood, cheek swabs, and saliva) by both manual and automated methods
and sequenced on several different Ion Torrent sequencing systems. Sequencing
data were analyzed for overall panel uniformity (% bases in target ≥ 0.2X mean
coverage), on target (% of all HG19-mapped reads that also map to designed
targets), reproducibility (CVs of panel uniformity across library preparation methods
and sample types), and specificity and positive predictive value of variant calling for
specific reference genomes tested. Results: Coverage uniformity for all On-Demand
panels was ≥ 95% across all library preparation methods and sample
types. Sensitivity and positive predictive value of variant calling for reference
genomes for single nucleotide polymorphisms were ≥ 97%. Fifty-nine genes
recommended by the American College of Medical Genetics and Genomics for
reporting of incidental findings (ACMG59) were tested as a stand-alone panel and
together with 75-135 additional genes and their uniformity was shown to be ≥ 97% in
all contexts. Conclusions: Ion AmpliSeq On-Demand panels provide a powerful and
flexible new tool for human inherited disease research. Primer designs for genes of
interest can be mixed and matched to fit specific research interests while maintaining
high performance in sequencing applications. Inventory is expanding to ~5000
genes in the near future, covering most important genes with links to inherited
diseases.
G07. WITHDRAWN
G08. Comprehensive Carrier Testing of 9,785 Chinese Couples for Common
Severe Recessive Disorders
S. Zhao1, F. Guo1, W. Zhong1, C. Chen1, Y. Wang1, Y. Zhu1, L. Song1, M. Mao2
1Tianjin Medical Laboratory, BGI-Tianjin, BGI-Shenzhen, Tianjin, China; 2BGIShenzhen, Shenzhen, China.
Introduction: China, as a populous country, is characterized by multi-ethnicity,
complicated substructure, widely geographical distribution, as well as cultural
diversity, with long history of consanguineous marriage, leading to high carrier rates
of recessive diseases or detrimental mutations in some regions of China. With
development of society, the population mobility is strengthened, making some highly
prevalent genetic diseases in specific regions become high incidence and carrier
rate in the whole nation of China. However, a comprehensive expanded carrier
screening for Chinese population had not been performed so far. Here, a
comprehensive carrier testing based next generation sequencing was
implemented for 8 prevalent and severe recessive disorders in Chinese
population. Methods: A high-throughput method based targeted genes capture and
NGS was employed to detect variants related 8 common disorders
(Dystrophinopathies, Hemophilia, hereditary hearing loss, Galactosemia,
944
Phenylketonuria, Wilson disease, Glycogen storage disease type II and Autosomal
recessive polycystic kidney disease). The variants were interpreted by referencing
the standards and guidelines recommended by ACMG. YH genome DNA and 179
samples with known variants characterized were analyzed to evaluate mutation
detection power. Then carrier testing was performed in 9,785 couples from southern
China after an informed consent process. Results: The results of validation
samples indicted the analytical accuracy and sensitivity were all over 99%. Among
9785 couples, we analyzed all exons and selected intergenic and intronic regions of
13 target genes. 1612 (8.24%) individuals were found carrying 352 distinct
pathogenic variants. Of these distinct pathogenic variants, 67 (19.03%) loss-offunction mutations were never been reported. 16 couples and 33 females were
determined to be at high risk of having offspring with related autosomal recessive
and X-linked disorders, respectively, signifying that approximately 1.25% of
genetically affected births could be averted combined with genetic counseling and
prenatal diagnosis after the performance of carrier testing. Conclusions: Combined
with prenatal diagnosis, comprehensive carrier screening based NGS can generate
great significance for averting genetically affected births for severe genetic disorders
in multi-ethnicity and complicated substructure Chinese population.
G09. Exome Re-Analysis and Complementary Testing Identify Novel Mutations
for Rare Mendelian Disorders
C. Wei, A. Mathur, J. Denton, J. Han, S. Indugula, X. Wang, A. Kumar, L. Dyer, M.
Sun, A.C. Valencia
Cincinnati Children's Hospital, Cincinnati, OH.
Introduction: The advent of next-generation sequencing technologies, especially
the introduction of whole exome sequencing has provided an opportunity to screen a
patient’s entire exome to establish genetic diagnosis. The utility of WES to identify
novel genes and variants causative of Mendelian disorders has been clearly
demonstrated in recent years. Most excitingly, there are at least a few dozen new
genes/mutations are being discovered each year approved to be associated with
these disorders. Methods: We re-analyzed patients’ WES data with initial negative
results for new clinical phenotype and/or in novel genes that have been recently
discovered to cause Mendelian diseases. A new and lab developed data analysis
pipeline Cincinnati Clinical Exome Pipeline Analysis Suite “CCEPAS” is utilized for
the re-analyses. In addition, knowing the limitation of the WES, we were able to
utilize a few complementary testing strategies, such as, deletion-duplication analysis,
epigenetic testing and Sanger rescue sequencing to identify mutations that helped to
reach a definitive diagnosis in patients that were tested to be “negative” or
“inclusive”. Results: We were able to establish new diagnoses in a good number of
patients with rare complex Mendelian diseases. We have completed re-analyses in
19 cases. We were able to identify new findings in 4 families (21%) based on newly
discoveries of gene function and/or disease correlation. For example, we have a
patient with a suspected clinical diagnosis of CVID, but the initial WES yielded no
positive findings. Two years later, the physician requested analysis for CECR1 gene,
which was recently report in patients with recurrent infections and antibody
deficiency without vasculitis, the latter being a key feature in patients with CECR1
mutations (leading to ADA2 deficiency). Re-analysis in this patient resulted in the
finding of a nonsense mutation. Additional rescue sequencing revealed another noncoding likely pathogenic variant. Enzymatic activity of ADA2 was nearly absent in the
patient’s serum. Conclusions: Periodic re-analyses of WES data and
complementary testing helped to establish definitive diagnosis in patients with rare
Mendelian disorders.
G10. Validation of A Cystic Fibrosis 55 Mutation Screening Assay on the
QuantStudio 12K Flex Open Array System
M.M. Moradian, R. Ramjit, W. Chen, S. Liang
Kasier Permanente Southern California, Los Angeles, CA.
Introduction: The QuantStudio 12K Flex Real-Time polymerase chain reaction
(qPCR) system with Open Array block (QuantStudio; ThermoFisher Scientific,
Waltham, MA) facilitates qPCR–based applications on a nanofluidic plate, running up
to 3,072 reactions in parallel per array. In this study, an interdisciplinary review was
conducted on the State of California and Kaiser Permanente Southern California
(KPSC) patient population to determine a customized mutation panel that included
the significant variants for cystic fibrosis (CF) mutation screening. The performance
characteristics of the assay were validated. Methods: Mutation analysis review and
determination of the customized CF variants to be tested was conducted by multiple
clinical disciplines based on data from the State of California Newborn CF Screening
and the patient population characteristics of KPSC. Fifty-five variants were
selected. De-identified clinical samples were used for the validation. DNA was
extracted from EDTA whole blood samples using the QIAgen MDx and
QIAsymphony instruments (QIAGEN, Hilden, Germany) and tested in 59
independent reactions (33nL each), run in duplicate, on the QuantStudio. Results
were analyzed with the TaqMan Genotyper Software for cluster analysis and on the
Quant Studio 12K Flex expression suite for amplification analysis. Results: After
testing 338 de-identified patient samples and 54 assayed samples from Coriell and
CAP, the accuracy of testing generated 100% concordant results compared to the
reference method for 227 negative (wild type) and 165 positive (heterozygous or
homozygous mutant) samples. Analytical specificity for all reactions was 99.76% and
analytical sensitivity was 100% with a minimum requirement of 10 ng/µl of DNA. The
inter- and intra- run precision studies generated acceptable results with 100%
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
concordance. There were 4 false positive reactions from a total of 99002 reactions
(i.e., 0.004%). Using this information, we determined that the probability of having 2
false positive results in one sample was 0.00057% and for the same mutation, ran in
duplicate, it was 0.00001%. The single reaction failure rate was <0.5%, ranging from
0.1%-0.4% per run (average 0.18 reaction per sample). Conclusions: The
customized CF variant mutation screening assay performed on the QuantStudio
system generated acceptable results for clinical testing. However, it is strongly
recommended to run samples in duplicate to avoid testing repeats that may occur
due to single reaction failure (average rate of 15/5192 reaction per 88 sample
run). Although we had 4 false positive reactions, it did not affect the accuracy of the
study since the samples were run in duplicate and no samples had 2 false positive
reactions.
G11. Linked-Read Sequencing for Molecular Cytogenetics
S. Garcia, S. Williams, J. Herschleb, P. Marks, A.W. Xu , M. Schnall-Levin, D.
Stafford, D.M. Church
10x Genomics, Pleasanton, CA.
Introduction: Structural variants (SVs), including deletions, duplications, insertions,
inversions, and translocations, are important contributors to disease. However, they
remain one of the most difficult types of variation to accurately ascertain, in part
because they tend to cluster in duplicated and repetitive regions. Clinically, aCGH
and karyotype remain the most commonly used assays for genome-wide SV
detection, though there is clear potential benefit to an NGS-based assay that
accurately detects both SVs and SNVs. Recently, laboratories have demonstrated
that low coverage long-insert WGS (liWGS) can identify both CNVs and balanced
rearrangements at a cost and speed similar to aCGH/ karyotype. This introduces the
possibility of a hybrid NGS assay (liWGS+genome/exome) but liWGS protocols are
difficult and time-consuming to establish. Methods: Linked-Read sequencing is a
relatively simple, fast, and cost-effective alternative that is applicable to both genome
and targeted assays. Linked-Reads are generated by performing haplotype-level
dilution of long input DNA molecules into >1 million barcoded partitions, generating
barcoded short reads within those partitions, and then performing short read
sequencing in bulk. We performed 30x Linked-Read genome sequencing on 24
GetRm CNVPanel samples with known balanced or unbalanced SVs. Downsampling
was performed on a subset to determine the lowest sequencing depth required to
detect variation. Results: Twenty-two of 24 variants were identified, and another
known event was called as a candidate. The remaining sample had a balanced
structural variant involving 16q11.2, a region known to be difficult to assess due to
gaps in the reference assembly. With current algorithms, CNVs can be called with as
little as 1-2x sequencing depth (5-10Gb) while balanced events require on the order
of 10x coverage. We demonstrate that there is specific signal in the data for
balanced events down to ~2x coverage. Similar to recent observations, we find
complexity at breakpoints not previously detected by conventional technologies.
Conclusions: Our results indicate that low coverage LR sequencing can identify
both CNVs and balanced rearrangements. Breakpoints can be manually resolved
using LR data, uncovering both complex rearrangements and reference misassemblies/rare SVs. The ease of the LR protocol make it an attractive alternative to
liWGS.
G12. High Throughput Linked-Read Sequencing for Improved Variant
Detection
A.N. Fehr, S. Garcia, M. Prout, C. O'Keeffe, A. Price, W. Xu, S. Williams, C.
Catalanotti, P. Marks, D.M. Church
10x Genomics, Pleasanton, CA.
Introduction: Causal single nucleotide variants (SNVs) and small indels tend to
impact genes across the genome, including in regions with known paralogs. Most
current next generation sequencing (NGS) techniques depend on short reads (SR)
that often cannot be mapped confidently in degenerate regions, which necessitates
reflexing to time-consuming and expensive alternate techniques. Methods: To
address variant detection shortcomings, we developed a technology that retains
long-range information while maintaining the power, accuracy, and scalability of
short read sequencing. The 10x Genomics Chromium Genome solution utilizes
haplotype-level dilution of high molecular weight DNA molecules into >1 million
barcoded partitions to create a novel data type referred to as ‘Linked-Reads’. These
Linked-Reads (LR) enable high-resolution genome analysis with minimal DNA input
(~1 ng). Here, we describe a high throughput workflow that interfaces a fluidics
automation robot with the Chromium platform to automate the LR library preparation
solution. Using this workflow, we generated Linked-Read NGS data using Agilent
SureSelect V6 Exome baits on multiple reference samples including NA12878.
Results: For ~60x median coverage on NA12878 samples, LR sequencing yielded
73.4% of bases with >10x coverage at mapq>30 compared to 71.4% for SR. For
clinically-relevant genes in loci with known paralogs, this metric ranged between 73.1
and 84.4% for LR but only 53.6-75.4% for SR. The fraction of exons with at least
99% of bases with >10x coverage at mapq>30 in difficult loci ranged from 71.377.6% for LR and 48.6-64.8% for SR. With this additional coverage, LR detected
1,597 novel SNPs in degenerate regions with 78.6% in the coding sequence of 414
total genes, versus only 527 SNPs for SR. Single exon deletions and duplications
are not routinely assessed in exome sequencing because coverage is highly
variable. LR data allows read haplotype reconstruction, thus reducing the
signal:noise ratio for duplications and revealing deletions as a complete loss of
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
coverage on one haplotype. For NA12878 50 exon-level deletions are called with LR
with a heterozygous sensitivity of 72.7% (PPV= 72%) and a homozygous sensitivity
of 83.3% (PPV= 100%). Duplications are robustly detected when phasing is present
over the affected region. Conclusions: These data demonstrate that LR exome
sequencing can identify more clinically relevant variants than standard short read
exome sequencing in an assay that is amenable to high-throughput automation.
G13. GALC Deletion/Duplication Detection by Droplet Digital PCR for Krabbe
Disease Confirmation in a Single Dried Blood Spot Punch
R. Majumdar, N. Vidal Folch, D. Milosevic, D. Gavrilov, K. Raymond, S. Tortorelli, P.
Rinaldo, D. Matern, D. Oglesbee
Mayo Clinic, Rochester, MN.
Introduction: Krabbe disease (KD) is an autosomal recessive disorder caused by
variants in GALC resulting in globoid cell leukodystrophy from galactocerebrosidase
(GALC) deficiency. Early detection of GALC deficiency is paramount to ensure that
hematopoetic cell transplantation occurs before symptoms transpire. KD is a target
of newborn screening (NBS) programs in New York, Missouri, Illinois, and Kentucky,
or is under consideration for implementation in New Mexico, New Jersey,
Pennsylvania, Ohio, and Tennessee. NBS methods for KD include measuring GALC
DBS activity, followed by psychosine levels, 30-kb deletion of GALC exons 11-17
detection, which accounts for a significant proportion of early-onset KD cases, and
finally, GALC sequencing if NBS test results are indeterminate for KD. Unfortunately,
pseudodeficiency alleles are commonplace and other GALC deletions or duplications
occur in sufficient frequency to hinder variant detection rates by gene sequencing. In
order to improve newborn screening for KD, we developed a multiplex, droplet digital
PCR (ddPCR) method for the detection of exonic GALC deletions or duplications
from DBS DNA. Methods: Copy numbers for each GALC exon, and RPP30 as an
internal control, were simultaneously measured with a Bio-Rad AutoDG and QX200
ddPCR system. The ddPCR reaction mix was prepared with 0.15-10 ng DNA,
a GALC/RPP30 primer-probe mix consisting of FAM probes targeting GALC exons
and a HEX RPP30 probe, and ddPCR Supermix distributed into 10 wells. Each PCR
reaction was partitioned into 20,000 nanoliter-sized droplets, GALC and
RPP30 targets were amplified by end-point PCR within each droplet, and the
fluorescence of each probe by 2 channels (FAM and HEX) were measured. Copy
number of each GALC exon was derived based on normalization
with RPP30 signal. Results: Normal individuals showed 2 copies of all 17 exons as
expected. The ddPCR detected 1 or 0 copies of each exon spanning 11 to 17 in
patients (n=4) with heterozygous or in patients (n=2) with homozygous 30
kb GALC deletions, indicating 100% sensitivity and specificity for the 30kb deletion.
In addition, a DBS specimen from an individual with a heterozygous exon 8
deletion in trans to a pathogenic variant, was also identified from a single DBS
punch. Conclusions: Our ddPCR method can sensitively and specifically
measure exon copy number in GALC from a single DBS punch and is applicable to
KD NBS alongside GALC activity, psychosine levels, and Sanger sequencing.
G14. Pericentromeric Regions of Homozygosity on the X Chromosome are
Likely Benign Population Variation
E.S. Barrie1, Y. Li2, A. McKinney1, D. Lamb-Thrush3, S. Hashimoto1, T. Jacobson1, D.
Mouhlas1, R. Pyatt4, C. Astbury5, S. Reshmi1, J.M. Gastier-Foster1, R. Pfau1
1Nationwide Children's Hospital, Columbus, OH; 2Kaiser Permanente Northwest,
Portland OR; 3Ambry Genetics, Aliso Viejo, CA; 4Sanford Health, Sioux Falls,
SD; 5Cleveland Clinic, Cleveland, OH.
Introduction: Chromosomal regions of homozygosity (ROH) may provide disease
insight, but pathogenicity is unknown for many regions. ROH around the centromere
of the X chromosome (pericentromeric ROH) is regarded as benign but has not been
empirically demonstrated. We examined X chromosome pericentromeric ROH and
determined its implications. Methods: Microarray results from female patients
processed for copy number/SNP analysis containing an ROH bordering the X
centromere were further analyzed. Results: Consecutive ROH spanning Xp to Xq,
likely including the centromere, was most frequently observed for regions Xp11.23 to
Xp11.21 and Xq11.1 to Xq12, with an average total size of 16.5 Mb. X chromosome
pericentromeric ROH was unlikely related to phenotype in 41% of cases due to other
explanations: likely pathogenic deletion/duplication (17%), apparently unaffected
female (7%), other clinical explanation (7%), or consanguinity (10%). There were
gains or losses of other chromosomal regions, with unknown significance in 22%
(27). Of the remaining 44 cases with ROH as the only finding, the indication for study
was non-specific. X chromosome pericentromeric ROH observed in our laboratory
overlapped with previously reported regions. Conclusions: Pericentromeric ROH on
the X chromosome are common, found in both affected and unaffected individuals,
and are therefore likely benign.
G15. Clinical Utility of Next Generation Sequencing (NGS) studies in
Neurological Disease – Our Experience at Kokilaben Dhirubhai Ambani
Hospital, India
J.C. Vyas, P. Gadgil, M. Pusalkar, R. Kittu, A. Gomes, N. Londhe, V. Vadera
Kokilaben Dhirubhai Ambani Hospital and Medical Research Institute, Mumbai,
India.
Introduction: Genetic involvement in diagnosis and management of Neurological
diseases can never be overemphasized. But studying genetics of neurological
disease is complex and challenging as it involves analysing multiple genes. NGS
945
AMP Abstracts
technique can analyse multiple genes of multiple samples in a single attempt. But
there are several challenges in analysing the variants detected during these studies,
especially variants of unknown significance (VUS) as they create a dilemma in
counselling the patients and are not helpful in disease management. Methods: Forty
patients with various neurological conditions were studied for genetic mutations
using Next Generation Sequencing (NGS; Neurology panel 600 genes). NGS testing
was outsourced to a professional laboratory. Targeted gene sequencing of protein
coding regions was performed using NGS. Genomic DNA was used for targeted
Gene capture using custom capture kit. Library Sequencing was performed to obtain
coverage of 80X-100X on Illumina platform. Only Non Synonymous and splice site
variants were reported and silent variants were not reported. Targeted mutation
analysis was performed using Sanger sequencing for variants on the ABI 3500 Dx
genetic analyser. Results: Three of 40 cases had pathogenic mutations, 7 cases
revealed likely pathogenic mutations, 10 had no clinically significant variants and in
20 children only VUS were detected. In all, total of 32 VUS were detected. 3/7 with
likely pathogenic variants also had VUS detected and 19 patients had only VUS. In
10 cases more than 1 variant was identified. In 7 cases, where parents could be
studied by targeted mutation (both or only mother), 6 were found to be carrying the
variants. The follow up status analysis of all VUS detected in this study was
performed by referring various databases and this revealed no change in their
significance. Conclusions: Only with more and more clarity in the significance of
variants, NGS technique will become an invaluable tool in patient management. This
will be possible when extensive genetic testing is being performed and databases
are being updated by everyone diligently and accessible to all the labs for variant
interpretation. Targeted mutation analysis of the parents and other affected and nonaffected members can also help in genetic counselling and in individual patient
management.
G16. Clinical Impact of Characterizing Genomic Alterations Using WholeGenome Mate Pair Sequencing
J. Blommel, S.A. Smoley, R. Jenkins, S. Johnson, M. Webley, S. Koon, G.
Vasmatzis, L.B. Baughn, R. Rowsey, R. Ketterling, N. Hoppman
Mayo Clinic Rochester, Rochester, MN.
Introduction: Mate pair sequencing (MPseq) is used in the characterization of
rearrangements, particularly balanced rearrangements, in order to identify clinically
significant genes at/near the breakpoints, which can aid in determining pathogenicity,
diagnosis, prognosis, and in some cases therapeutic options. Due to the nature of
its unique library preparation method, MPseq has higher sensitivity than conventional
NGS for detecting structural variation in the human genome. Methods: As
previously described by Cao et al, we have validated the first clinical MPseq assay
using the Illumina mate pair library preparation in conjunction with Illumina HiSeq
Rapid (2x101bp) sequencing chemistry and BIMAv3 mapping and alignment tool for
the detection of structural variants. This is a multi-laboratory effort in which DNA is
isolated from either blood, bone marrow, or fresh/frozen tissue in the extraction core,
prepped and sequenced using MPseq in our next-generation sequencing core
laboratory, and aligned and analyzed for structural variation in our bioinformatics
core. Results are interpreted in a targeted fashion depending on the rearrangement
that is to be characterized for each particular patient, and a clinical report is issued to
the patient’s provider. All results are also confirmed by an orthogonal method such
as Sanger sequencing. In our first clinical sample, MPseq was offered for a 10 y/o
patient with relapsed B lymphoblastic leukemia. At diagnosis, conventional
chromosome studies identified a 12;15 translocation. Interphase and metaphase
fluorescence in-situ hybridization (FISH) confirmed a rearrangement involving
the ETV6 gene (at 12p13) with an unknown partner gene located on chromosome
15. MPseq was performed to determine the partner gene in hopes that prognosis or
treatment for the patient would be better defined. Results: MPseq confirmed the
presence of a t(12;15)(p13.2:q25.3) rearrangement resulting in fusion
of ETV6 and NTRK3. Specifically, this rearrangement leads to fusion of exons 1-4
of ETV6 with exons 15-20 of NTRK3. This fusion has been described previously in
several malignancies, including a pediatric patient with Ph-like B lymphoblastic
leukemia. Most importantly, this fusion is targetable using tyrosine kinase inhibitors,
and this patient’s therapeutic regimen now has the potential to be altered as a result
of this clinical assay. Conclusions: The validation and implementation of MPseq
has already, with the first patient tested, demonstrated its clinical utility and potential
impact on patient care. Because of these results, this patient will be treated with
tyrosine kinase inhibitors in addition to other chemotherapies in hopes of achieving
and sustaining remission long-term.
G17. Comparison of Specimen Collection Methods for Pharmacogenetic
Testing
H. Katzov-Eckert1, J.S. Ang1, M.N. Aloise1, D. Dawes1, M.G. Dempster2, R.
Fraser3, A. Paterson4, P. Stanley1, A. Gonzalez-Suarez1, M. Dawes1
1GenXys Health Care Systems, Inc., Vancouver, British Columbia, Canada; 2GenXys
Health Care Systems, Inc., Shoppers Drug Mart, Inc., Vancouver, British Columbia,
Canada; 3Molecular You, University of British Columbia, Vancouver, British
Columbia, Canada; 4GenXys Health Care Systems, Inc., Clinicare Pharmacists, Inc.,
Vancouver, British Columbia, Canada.
946
Introduction: A non-invasive, easy to collect sample collection method is key to the
uptake and integration of pharmacogenetic testing by clinicians. We aimed to
compare DNA isolation, whole genome amplification and genotyping quality between
2 types of buccal swabs. Methods: Thirty-one participants self-collected samples
from the ORAcollect and the Puritan PurFlock Ultra swabs. Quantity and quality of
DNA isolation was determined as well genotyping performance on a custom
pharmacogenetic panel. The panel included 30 Single Nucleotide Polymorphisms
(SNPs) on the OpenArray qPCR platform and a CYP2D6 TaqMan Copy Number
Variant (CNV) assay. Results: The DNA yield (mean±SD) and A260/280
(mean±SD) ratio was found to be 1.9±1.4µg and 1.85±0.39 for OraCollect samples
and 0.25±0.35µg and 1.72±0.78 for Puritan swabs, respectively. Whole genome
amplification (WGA) was tested for the low yield Puritan swabs with an average
increase of 0.34µg of initial DNA. However, WGA samples had a 21% SNP
genotype call rate and only 45% of samples had CNV confidence scores >95%.
Samples obtained from the ORAcollect kit had a genotyping call rate of 97%, and
100% of the samples had CNV confidence scores of >95%. Conclusions: Whole
genome amplification of DNA extracted from low yield swabs was not suitable for
genotyping on the OpenArray and may introduce bias for CNV assignment. The
ORAcollect swab is a relatively easy to collect source of genomic DNA for
downstream analysis. We found 3 main factors that can have an effect on
pharmacogenetic testing: The individual collecting the sample, the collection device,
and the assay.
G18. Using the GeneReader NGS System to Identify Mutations in BRCA 1/2,
PTEN and TP53
N. Dennison1, A. Burke1, A. Burton1, R. Dyson1, S. Hughes1, S. Jackobsen2, V.
Kapoor1, Y. Kong3, D. Lueerssen1, B. Oester2, L. Schauser2, K. Smith1, T. Rothma4
1Qiagen, Manchester, United Kingdom; 2Qiagen, Aarhus, Denmark; 3Qiagen,
Redwood City, CA; 4Qiagen, Hilden, Germany.
Introduction: Breast cancer is the most frequent cancer in women with
approximately 1 in 8 developing invasive disease during their lifetime. The key tumor
suppressor genes BRCA1/2, PTEN and TP53 are critical in DNA repair and
maintenance of chromosomal stability. Inherited or acquired mutations in these
genes allow cells to grow and divide unregulated, leading to the formation of a
tumors. Targeted Next-Generation Sequencing (NGS) has proven to be invaluable
for identification of mutations and has the potential, in the future, to dramatically
improve detection and enable earlier intervention. However, unlike some genes that
possess well characterised hotspots, any exonic mutation in BRCA1/2, PTEN and
TP53 can result in oncogenic transformation. The GeneRead QIAact BRCA
Advanced Panel in combination with the QIAGEN GeneReader NGS System
provides a solution to test for mutations, including single nucleotide variants (SNV)
and Insertion/Deletions (InDel) in the entire coding regions of the targeted genes.
Methods: The GeneRead QIAact BRCA Advanced Panel has been designed to
detect mutations throughout the BRCA1/2, PTEN and TP53 protein coding sequence
±20 bp of intronic sequence for detection of splice site mutations. A key feature of
the panel is the addition of a unique molecular index (UMIs) to tag individual
molecules prior to target enrichment by PCR. This allows duplicate reads originating
from the same molecule to be identified and used for correction of PCR and
sequencing artefacts. To demonstrate accuracy, control samples (including
reference standard Coriell samples, ATCC cell lines) and FFPE clinical samples
were used. Following target enrichment, libraries were sequenced on the
GeneReader and mutations were analysed using QIAGEN Clinical Insight (QCI)
Analyze. Results: The control samples confirmed the ability of the panel to detect a
wide range of DNA mutations in BRCA1/2, PTEN and TP53. All expected mutations
(5% allele frequency threshold appropriate for FFPE DNA) were repeatedly identified
both within and between runs with 100% concordance. Results confirmed UMI
coverage of the entire coding region of the 4 genes and testing of clinical samples
consistently detected the expected DNA mutations that had been previously
reported. Conclusions: The data presented here illustrate the reliability of the
GeneRead QIAact BRCA Advanced Assay on the GeneReader System. This study
demonstrates the ability of a single assay to detect a wide range of DNA mutations in
key tumor suppressor genes at the single nucleotide resolution for each and every
coding exon. This allows a research laboratory to efficiently and effectively detect all
possible actionable breast cancer mutations in BRCA1/2, PTEN and TP53 and not
just at mutation hotspots.
G19. Detecting Pharmacogenomic Variants Using Long- and Short-Read Next
Generation Sequencing Platforms
C.A. Schumacher, A. Wood, S. Sandhu, J. Lenhart, L. Kurihara, V. Makarov, T.
Harkins
Swift Biosciences, Ann Arbor, MI.
Introduction: The key processes underlying the field of pharmacokinetics are
absorption, distribution, metabolism, and excretion (ADME). Screening for
pharmacogenomic markers is often required to ensure safe and effective disease
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
treatment. Despite this requirement, problems including high homology with known
pseudogenes and difficult to sequence motifs arise when trying to sequence ADME
genes, like CYP2D6. Most next generation sequencing (NGS) assays utilize shortread (SR) chemistry that enables genotyping of known biomarkers in accessible
parts of the genome. Newer long-read (LR) assays provide comprehensive
sequencing of an entire gene, providing insight into copy number variations, but are
limited to fewer genes per assay. Massively parallel screening using either NGS
assay is cost effective, especially compared with traditional methods such as realtime PCR. Here we show the utility of these 2 assays to characterize ADME
biomarkers. The SR sequencing assay detected hotspots in CYP2D6 as well as 150
additional genes for germline SNP genotyping. The LR sequencing assay surveyed
the entire CYP2D6 gene for a full and detailed look at CYP2D6 and its associated
variants. Methods: Twenty-four DNA samples from the Coriell Institute with known
CYP2D6 variants were used in each assay. In the SR assay, an amplicon-based
panel targeting 364 ADME-specific targets was performed. 10ng of each DNA was
used as input for a multiplexed PCR, and the products were subsequently adapted
for Illumina sequencing. These libraries were sequenced on an Illumina MiniSeq
using a paired end read length of 151bp. In the LR assay, less than 100ng of each
DNA was used as the input for long range PCR, and the 6.5Kb PCR products were
subsequently barcoded for sample multiplexing and adapted for Pacific Biosciences
(PacBio) sequencing. These libraries were then sequenced on a PacBio RSII.
Results: Libraries from the SR assay were sequenced with >90% on-target and
coverage uniformity. Ninety-two percent of known CYP2D6 variants, among other
ADME variants, were detected using this technique. Using the LR technique, all
known CYP2D6 variants could be detected. Importantly, 100% of the CYP2D6 gene
could be covered by this assay. Conclusions: Having the right tool to address
questions concerning pharmacogenomic profiles is critical to ensuring proper,
personalized treatments. Using the short-read assay, we demonstrate an ability to
screen for a wide range of germline ADME targets. Such an assay takes advantage
of the current knowledge bank and is informative for immediate and broad use. With
the long-range assay, we show the ability to comprehensively sequence the entire
CYP2D6 gene, thereby gaining more insight into CYP2D6 variants and phenotypes
and further informing personalized treatment options.
G20. Microdeletion in SNRPN May Lead to False Positive Results for Angelman
Syndrome Using Methylation Analysis
B.M. Zhang1, A. Popa2, A. Ferreira-Gonzalez2, L.J. Jennings3, K.E. Weck4
1Stanford University Medical Center, Stanford, CA; 2Virginia Commonwealth
University, Richmond, VA; 3Northwestern University's Feinberg School of Medicine,
Chicago, IL; 4University of North Carolina, Chapel Hill, NC.Submitted on behalf of the
CAP/ACMG Biochemical and Molecular Genetics Committee.
Introduction: Prader-Willi syndrome (PWS) and Angelman syndrome (AS) are
clinically distinct developmental and neurobehavioral genetic disorders that are
attributable to abnormal imprinted-gene expression within chromosome region
15q11-q13. Currently, analysis of methylation patterns using molecular genetic
techniques is the most sensitive diagnostic approach to differentiate between
methylated (maternal) and non-methylated (paternal) alleles. Of the different
methylation assessment techniques, methylation-specific PCR (MS-PCR) targeting
the 5’ CpG island of the SNRPN locus and methylation-sensitive multiplex ligationdependent probe amplification (MS-MLPA) are the most widely used in molecular
pathology laboratories. These methods can detect more than 99% of individuals with
PWS and approximately 70% of individuals with AS. However, microdeletions in
the SNRPN gene may pose a challenge in test interpretation. We report the
investigation of an unusual CAP proficiency-testing (PT) sample and discuss some
caveats in interpreting methylation-specific molecular assays for PWS/AS.
Methods: In a recent CAP PT survey for PWS/AS, one DNA sample, derived from a
phenotypically normal male with 2 children with PWS, contained a small
microdeletion in SNRPN on the maternal allele. Neither UBE3A nor the imprinting
center was affected. A total of 87 laboratories performing PWS/AS molecular testing
returned the genotype and phenotype interpretations by their assessments.
Results: Of the 87 labs, only 5 labs using MS-MLPA methodology reported the
correct result of not consistent with PWS or AS, whereas the majority of laboratories
using either MS-PCR or MS-MLPA reported this sample as consistent with a
diagnosis of AS. This unusual survey result triggered an investigation by the
Biochemical and Molecular Genetic Committee overseeing this PT program. As a
result, it was concluded that the false positive interpretation of AS was likely due to
loss of the maternal allele in the SNRPN region where methylation-specific primers
are typically located. Although the incidence of small microdeletions in SNRPN is not
known, it is likely a rare occurrence. There have been some anecdotal reports of
inherited microdeletions in this region that are asymptomatic when transmitted
through the mother. Conclusions: Rare small microdeletions affecting SNRPN but
not UBE3A or the imprinting center may lead to false positive results of AS using
methylation-specific molecular assays, particularly MS-PCR. The more recent
techniques, such as MS-MLPA and microarray analysis, detect copy number
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
changes and therefore can more precisely identify the region of microdeletion.
G21. BRCA1 Mutation Detection Using QIAGEN GeneReader NGS System in a
Case with RET Codon 634 Mutation
B. Sarkadi1, K. Baghy2, Z. Sapi2, K. Rácz1, A. Patócs1
1Hungarian Academy of Sciences and Semmelweis University, Budapest,
Hungary; 2Semmelweis University, Budapest, Hungary.
Introduction: Next-generation sequencing is a cost-effective approach for molecular
genetic analysis. Despite its wide usage for research, it poses several issues
including sample preparation, QA and bioinformatics analysis, and that interpretation
needs to be standardized before NGS can be adopted more broadly in routine
analysis. Our aim was to evaluate QIAGEN's newly launched GeneReader NGS
System for BRCA1/2 mutation detection, via analysis of a case with a 13 base pair
deletion in BRCA1 previously confirmed by Sanger sequencing. Methods: A 16year-old female was referred to the Endocrine Unit of the 2nd Department of
Medicine, Semmelweis University due to hair loss. Two heterogeneous masses were
identified in both lobes of the thyroid gland. Fine-needle aspiration biopsy was
performed and medullary thyroid carcinoma (MTC) was diagnosed. The patient’s
young age suggested RET mutation. Indeed, mutational analysis following genetic
counseling revealed p.(Cys634Trp), a well-established pathogenic mutation
associated with autosomal dominant Multiple Endocrine Neoplasia type 2 (MEN2).
However, no other MEN2A-specific manifestation was detected. Furthermore, no
MEN2A-associated manifestation was observed in first degree relatives, despite the
same pathogenic mutation being identified in the proband’s brother and in his 2
sons. Of note, the patient's mother and maternal aunt had died of breast cancer,
suggestive of hereditary breast cancer syndrome. No samples were available from
the deceased, but the proband underwent further genetic counseling and BRCA1/2
testing. Sanger sequencing uncovered a novel frameshift mutation
(BRCA1 Ile90Serfs, NC_000017.10:g.41256905_41256917). This mutation was not
present in the brother or father, indicative of maternal inheritance, and is presumably
the culprit of the breast cancer in the aforementioned cases. To evaluate the
GeneReader System, the same DNA sample was processed, sequencing performed
and results analyzed. We observed a high coverage of this position in
the BRCA gene at 24,325, and confirmed the same frameshift mutation. 54% of all
reads harbored the mutation, consistent with a heterozygous inheritance pattern.
Conclusions: We present here a family with 2 rare monogenic autosomal dominant
tumor syndromes, resulting from 2 pathogenic germline mutations (one in RET and
the other in BRCA1). Such families require active and heightened cancer
surveillance. In addition, the QIAGEN GeneReader System successfully confirmed
the BRCA1 mutation, providing an NGS-based assay to effectively
identify BRCA mutations. A fully integrated system including sample preparation,
automation and bioinformatics analysis and interpretation, the GeneReader System
offers an easy solution for laboratories interested in genetic analysis.
G22. Colorectal Cancer Predisposition and its Genetic Characterization of
Korean Patients
K. Park1 , J. Sohn2 , J. Park1 , H. Kim2
1Samsung Advanced Institute for Health Sciences and Technology Sungkyunkwan
University, Seoul, Gangnamgu, South Korea; 2Samsung Medical Center,
Sungkyunkwan University School of Medicine, Seoul, Gangnamgu, South Korea.
Introduction: Identification of genetic alterations is important for family risk
assessment in colorectal cancers. Hereditary nonpolyposis colorectal cancer
(HNPCC; Lynch syndrome (LS)) is an autosomal dominantly inherited disorder
which is caused by a genetic mutation that disrupts the mismatch repair (MMR)
pathways, which results in microsatellite instability. APC-associated polyposis
including familial adenomatous polyposis (FAP) is defined by the presence of germline mutations in adenomatous polyposis coli gene APC or in the base excision
repair gene MutY homologue MUTYH. Peutz-Jegher syndrome (PJS) is an
autosomal dominant disorder caused by pathogenic variants of the serine threonine
kinase STK11. The aims of this study, which tested the largest number of East
Asians in a single center so far as we know, were to investigate the spectrum of
hereditary colon cancer syndromes and to evaluate the prognostic impacts of each
gene and variant type. Methods: In total, 176 individuals were enrolled for genetic
workup for FAP gene molecular testing, between Jan 2003 and May 2017 in the
Cancer Genetics Clinic of the Samsung Medical Center in Korea. Moreover, 322
consecutive individuals received for MMR gene molecular testing
(MLH1, MSH2, and MSH6 for SNV screening in the same institute from year 2004 to
May 2017. Multiplex ligation-dependent probe amplification (MLPA) done in patients
showed no mutation in the APC gene sequencing. Another 49 individuals
received STK11 gene molecular testing. Histopathologic findings and clinical
information as well as microsatellite instability data were collected. Results: A total
of 131 patients were found to have pathogenic or likely pathogenic variants in one or
more of MMR pathway genes (40.7%). We identified 76 different mutations (47
in MLH1, 24 in MSH2, and 5 in MSH6) including 24 novel variants and one founder
947
AMP Abstracts
variant (c.1024delinsAA in MSH2). APC mutations were detected in 86 unrelated
patients. Among 90 patients (51%) negative for point mutations, one of 16 patients
showed whole gene deletion by MLPA. Twelve of STK11 tested individuals showed
pathogenic/likely pathogenic variants (24.5%). Truncation mutations such as frameshift and nonsense mutations in MLH1, MSH2, and/or MSH6 were associated with
shorter tumor-free interval (TFI) compared to other mutation types (median of 3,129
days vs. 5,877 days, HR = 2.313, 95% CI = 1.101-4.858, P< 0.05).
Conclusions: This study would give valuable information about the genetic features
of Asian patients with hereditary colorectal cancer syndromes. For genetic
counseling, stratification of the surveillance programs, the elucidation of their genetic
basis would lead to the development of diagnostic algorithms that include both tumor
profiling and germ-line analysis.
G23. Tumor Mutations Can Help Classify Germline Variants: Learning from
Mismatch Repair Deficiency
B.H. Shirts1, E.Q. Konnick1, T. Walsh1, M. King1, J.M. Ranola1, A. Jacobson1, R.
Pearlman2, H. Hampel2, C.C. Pritchard1
1University of Washington, Seattle, WA; 2Ohio State University, Columbus, OH.
Introduction: Established germline variant classification guidelines do not
incorporate inferences from tumor data. Somatic mutations in mismatch repair
genes associated with microsatellite instability and loss of corresponding protein
expression by IHC have the same tumor phenotype as germline variants seen in
tumors, indicating a scenario where information may be used to classify both somatic
mutations and Lynch syndrome variants. Methods: We quantified likelihood ratios
that somatic variants in MLH1, MSH2, MSH6, and PMS2 drive microsatellite
instability and characteristic IHC staining patterns. Likelihood ratios (LR) for
pathogenicity were derived by comparing normalized variant read fraction (VRF) for
52 pathogenic mutations to those of 997 intronic passenger mutations from 78 paired
tumor-normal samples. Results: For tumors without evidence of loss of
heterozygosity, variants with normalized VRF above 0.35 had a LR for pathogenicity
of 3.4 (95% CI 2.8 to 4.2). For tumors with evidence of loss of heterozygosity,
variants with normalized VRF above 0.80 had a LR for pathogenicity of 29.8 (95% CI
14.5 to 61.0). Using these LR and corresponding LR associated with benign
passenger variants, we performed Bayesian multifactorial likelihood calculations for
missense, synonymous, and splice variants consistent with IHC staining identified in
the initial 50 of the 76 tumor-normal pairs and in 35 of the 86 subsequent tumornormal pairs. Bayesian analysis combining likelihood ratios with pretest probabilities
derived from published analyses resulted in multifactorial likelihood scores for 94
missense and splice variants, 52 of which have not been previously reported. Based
on this approach, 22 previously unclassified or uncertain variants are now classified
as pathogenic or likely pathogenic, including 5 previously observed germline
variants. Conclusions: These results are proof of principle that understanding of
specific tumor biology can permit tumor mutation data to inform germline variant
classification.
G24. Discovery of Unique Disease- and Gene-Specific Peripheral Blood DNA
Methylation Signatures Allows Molecular Diagnosis and VUS Classification in
Hereditary Genetic Syndromes
B. Sadikovic1, L. Schenkel2, C. Schwartz3, K. Boycott4, P. Ainsworth2, E. ArefEshghi2
1London Health Sciences Centre, Western University, London, Ontario,
Canada; 2Western University, London, Ontario, Canada; 3Center for Molecular
Studies, J.C. Self Research Institute of Human Genetics, Greenwood Genetic
Center, Greenwood, SC; 4Children's Hospital of Eastern Ontario Research Institute,
University of Ottawa, Ontario, Canada.
Introduction: Hereditary genetic conditions resulting from mutations in genes
involved in epigenetic machinery and chromatin remodeling present with systemic,
complex and often overlapping clinical features. Genetic diagnosis is further
complicated by a large proportion of clinically un-interpretable variants of unknown
clinical significance (VUS) in these genes. Recently, we have demonstrated
existence of unique, gene/disorder-specific DNA methylation “epi-signatures”
associated with such conditions and have begun systematic epi-signature mapping
in these disorders. Methods: Peripheral blood samples from disease-specific
cohorts of patients with Mendelian disorders of epigenomic machinery were
assessed for genome-wide methylation changes relative to controls using Illumina
Infinium 450k and EPIC genome-wide DNA methylation arrays (>1,500 patents
tested). We applied the identified epi-signatures to supervised and non-supervised
machine learning techniques to develop unique predictive models for each disorder.
Results: We identified highly sensitive and specific peripheral blood DNA
methylation epi-signatures in a number of genetic conditions including: FloatingHarbor Syndrome (SRCAP); autosomal dominant cerebellar ataxia, deafness, and
narcolepsy (DNMT1); alpha thalassemia/mental retardation X-linked syndrome
(ATRX); Kabuki syndrome (KMT2D); Sotos syndrome (NSD1); CHARGE syndrome
(CHD7); and syndromic X-linked intelectual disability, Claes-Jensen type (KDM5C).
948
We demonstrate that these unique epi-signatures, along with our large reference
database, enable near 100% sensitive and specific diagnosis of these disorders. We
also demonstrate the ability of these episignatures to accurately re-classify VUSs in
these genes as either pathogenic or benign. Finally, we show evidence of a partial
overlaps between specific episignatures in some of the above conditions including
Floating-Harbor, Kabuki, and CHARGE syndromes. Conclusions: This study
demonstrates presence of unique and highly-specific DNA methylation in peripheral
blood of patients suffering from hereditary genetic conditions involving genes that
regulate epigenetic machinery and chromatin remodeling. These epi-signatures can
be used for clinical diagnostic screening in this patient population, and enable
interpretation and clinical classification of VUSs in the associated genes.
Furthermore, these genomic DNA methylation defects provide insights in the
molecular etiologies of these disorders. Ongoing work in our research laboratory
focuses on systematic mapping of epi-signatures across other epi/genetic
syndromes, and transition of this technology into routine clinical use in our clinical
molecular laboratory.
G25. Analytical Validation of the Advanta Immuno-Oncology Gene Expression
Assay for Profiling of Immunobiology and the Development of Predictive Gene
Signatures for Response to Immunotherapies
P. Chen, T. Goralski, S. Myeres, J. Qin, L. McClean, R. Ramakrishnan
Fluidigm Corporation, South San Francisco, CA.
Introduction: The modulation of immune checkpoints offers one of the most
promising approaches toward antitumor therapeutics and drug resistance in recent
years. A variety of agents have been identified targeting inhibitory pathways for
cancer treatments. Gene expression profiling is an important tool to monitor the
human immune response to different antitumor agents for drug development and to
identify predictive signatures for response to immunotherapies. We have developed
and validated a gene expression assay panel, namely the Advanta ImmunoOncology Gene Expression Assay on the Fluidigm Biomark HD system, a
microfluidic real-time PCR system, for investigation of antitumor agent effects on
immune pathways. The assay panel consists of 170 unique genes in categories such
as T cell subset markers (e.g. CD4), cytokines and chemokines (e.g. IFNg), immune
regulation (e.g. PD-L1, PD-L2) and immune cell-fate markers (e.g. NK).
Methods: The Advanta Immuno-Oncology Gene Expression Assay kit consists of
Fluidigm 96.96 Dynamic Array IFCs, RT-PCR reagents, and assays. The assay
panel is composed of TaqMan gene expression assays and is divided into 2 subpanels, 96 assays to be run on one 96.96 IFC and 79 assays on another 96.96 IFC
with 17 unused assay inlets available for customization. Each sub-panel contains 5
housekeeping genes for assay normalization. The analytical validation was
performed by Q2/EA. Universal RNA was used to test RNA input range (2 -200 ng)
and cDNA was transcribed from a universal RNA and differentially pre-amplified to
achieve balanced amplification signals for all 170 genes. The cDNA was used to
evaluate amplification efficiency, linearity and reproducibility. Three of 20 RNA
samples extracted from archived FFPE samples using the miRNeasy (Qiagen) were
qualified based on RNA quality for analytical validation of the assays. Each sample
was tested in 3 replicates on 96.96 IFCs and in 3 independent runs. Results: The
analytical validation results show an average amplification efficiency across all
assays of 99%±4.2% and linearity (RSQ) 0.994±0.001. RNA input can be as low as
2 ng or less depending on the gene expression level of individual targets. The withinrun replicate correlation (RSQ) is 0.998 for the universal RNA and 0.978 for FFPE
samples and the between-run correlation (IFC-IFC correlation) is 0.99. Gene
expression signatures for individual FFPE samples were observed. Conclusions:
The Advanta Immuno-Oncology Gene Expression Assay on Fluidigm Biomark HD
system provides high amplification efficiency, excellent linearity and reproducibility
and reduces the cost of routine laboratory testing for both immune checkpoint
research and drug development.
G26. Genome Sequencing Reveals Variants in Non-Coding Regions Cause
Hereditary Hemorrhagic Telangiectasia
G. Akay Tayfun1, W. Wooderchak-Donahue2, P.A. Johnson2, J.A. McDonald2, P.A.
Bayrak-Toydemir3
1Zeynep Kamil Training and Research Hospital, Istanbul, Turkey; 2ARUP Institute for
Clinical and Experimental Pathology, Salt Lake City, UT; 3ARUP Institute for Clinical
and Experimental Pathology and the Department of Pathology, University of Utah,
Salt Lake City, UT.
Introduction: Hereditary hemorrhagic telangiectasia (HHT) is a genetically
heterogeneous disorder caused by mutations in one of the multiple
genes ENG, ACVRL1, and SMAD4. Determining the genetic basis of HHT
manifestations in patients who do not have a mutation in one these genes has
remained a great challenge for nearly a decade. Approximately 15% of HHT patients
do not have a mutation in one of these genes, and the genetic cause remains
unknown for some families even after exome analysis. We hypothesize that
mutations in the noncoding regions of the known genes cause HHT in these
cases. Methods: DNA from 35 individuals with HHT and 2 healthy controls from 13
families underwent genome sequencing. Approximately 100 additional unrelated
cases with HHT who tested negative were evaluated using a custom designed next
generation sequencing (NGS) panel to capture the coding and noncoding regions
of ENG, ACVRL1 and SMAD4 (referred to as the "HHT genome"). Coding regions
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
of GDF2 and RASA1 which have been implicated in the clinical differential for HHT
were also included in the NGS panel. All samples were sequenced using 2x100 PE
reads on a HiSeq2500 instrument and data were analyzed to identify novel and rare
mutations. Results: Several families had causative noncoding variants in
the ENG or ACVRL1 genes. A few families had a variant in ACVRL1 intron 9 that
disrupts splicing, including one family with an ACVRL1 intron:chromosome 3
translocation. Analysis is currently underway to identify additional HHT causative
variants in the other families and cases. Conclusions: This study shows the
importance of noncoding regions in the disease mechanism although interpretations
of the variants in these regions are very difficult. As the non-coding and regulatory
regions of these genes are better understood, these regions can also be included in
diagnostic testing to increase the overall molecular diagnosis for HHT.
G27. Genetic Testing of Noonan Syndrome Using Targeted Next-Generation
Sequencing Panel
C. Seol, B. Lee, H. Yoo, S. Choi, J. Lee, E. Seo
University of Ulsan College of Medicine and Asan Medical Center, Seoul, South
Korea.
Introduction: Noonan syndrome is a relatively common congenital genetic disorder
which is characterized by distinctive facial features, short stature, heart defects,
chest deformity, and other signs and symptoms. The pathogenesis is associated with
RAS/MAPK cell signaling pathway and the mutations of PTPN11, SOS1, RAF1,
and RIT1 have been identified. However, Noonan syndrome is genetically
heterogeneous and the diagnosis is still dependent on the clinical presentation and
the genetic diagnosis by Sanger sequencing has limitations. In this study, we
developed a targeted next-generation sequencing (NGS) panel for Noonan
syndrome and evaluated the test for the patients with suspected Noonan syndrome.
Methods: We tested peripheral blood samples from 20 patients with suspected
Noonan syndrome including 3 patients with known mutations as positive controls
(c.1391G>A (p.Gly464Glu) of BRAF, c.770C>T (p.Ser257Leu) of RAF1, and
c.188A>G (p.Tyr63Cys) of PTPN11) for the targeted NGS. The NGS panel included
14 genes such as A2ML1, BRAF, CBL, HRAS, KRAS, MAP2K1, MAP2K2, NRAS,
PTPN11, RAF1, RIT1, SHOC2, SOS1, and SPRED1, which comprised 268
amplicons and covered 27kb of genomic sequence in total. Library preparation,
template preparation, sequencing run, and data analysis were performed using Ion
Torrent Personnel Genome Machine (PGM) system (Thermo Fisher Scientific,
Waltham, Massachusetts, USA). We used Ion Reporter Software (Thermo Fisher
Scientific) for the variant analysis and annotation. The annotated variants were
interpreted and classified according to the 5-tier system by ACMG guideline. All the
variants were confirmed by Sanger sequencing. Results: The average coverage of
the panel was 1100x and >94% of amplicons showed target coverage of >20x. Nine
variants were identified from the total patients including positive controls. From the 3
positive controls, the detection of variants by the NGS panel showed the accuracy of
100% and did not showed additional pathogenic variants. Except positive controls,
the distribution of variants showed: one in BRAF, one in CBL, 2 in MAP2K1, and one
in RIT1. From the 2 variants in CBL, one was a variant of unknown significance
(c.1858C>T (p.Leu620Phe)), and the other one was a novel likely pathogenic variant
(c.1246T>C (p.Cys416Arg)). The variant of RIT1 (c.295T>C (p.Phe99Leu)) was also
a novel likely pathogenic variant. All the other variants were classified as likely
pathogenic by ACMG guideline. Conclusions: From this study the diagnostic yield
of Noonan panel was 29% (5/17) which was more efficient compared to the stepwise
Sanger sequencing. The NGS panel was also useful to identify novel variants.
Further investigation is needed for genotype-phenotype correlations for more
efficient genetic testing.
G28. Short Tandem Repeat Analysis Reveals a High Rate of Partial
Hydatidiform Moles in Triploid Conceptions Identified by Prenatal
Chromosome Microarray
X. Wu, U. Aypar, S.E. Kerr
Mayo Clinic, Rochester, MN.
Introduction: Triploid conceptions can be diandric (contribution of 2 paternal sets of
chromosomes) or digynic. The probability of diandry (partial hydatidiform mole) is
highest in first trimester specimens (85%), but the risk is thought to be much lower in
more advanced pregnancies. Identifying partial moles can be clinically important
due to the associated risk of persistent mole and choriocarcinoma. This study
characterizes a cohort of triploids, predominantly amniocentesis specimens,
identified by prenatal microarray. Methods: Of 30 triploids identified by prenatal
chromosome microarray in a 2-year period, 18 had a maternal blood specimen
submitted for maternal cell contamination studies by short tandem repeat (STR)
genotyping. Multiplex PCR amplification of 12 STR loci from different chromosomes
was performed using DNA extracted from the prenatal specimen and maternal
blood. At least 2 informative markers with 2 extra foreign alleles versus the maternal
specimen were required to call diandry. Digyny was assumed when no markers
were suggestive of diandry and allelic ratios were consistent with
digyny. Results: The 18 cases consisted of 14 uncultured amniotic fluids, 2
chorionic villous specimens, 1 fetal skin, and 1 cultured amniotic fluid. Ten digynic
triploids and 8 partial moles were identified. Triploidy was confirmed by conventional
karyotyping in all cases (9 cases 69,XXX and 9 cases 69, XXY). Our study was
limited by lack of complete data regarding gestational age and clinical follow-up for
risk of persistent gestational trophoblastic disease, but amniocentesis specimens
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
imply a gestational age of between 15 and 20 weeks for almost all cases in our
cohort. Conclusions: In this study, we found that nearly half of triploids (44%)
identified by prenatal chromosome microarray were partial moles. This distinction
has important implications for clinical management, and we conclude that STR
analysis should be carried out in all triploid cases, regardless of finding in a more
advanced pregnancy.
G29. Comparison of EUROArray HLA-DQ2/DQ8-h Direct and Olerup SSP for the
Determination of Celiac Disease Associated Risk Factors HLA-DQ2.2, -DQ2.5
and -DQ8
N. Miron1, C. Larson1, K. Axel2, A. Wilcke2, U. Steller2, F. Amirbeagi1
1Sahlgrenska University Hospital, Gothenburg, Sweden; 2EUROIMMUN
Medizinische Labordiagnostika, AG, Luebeck, Germany.
Introduction: The heterodimeric human leukocyte antigens (HLA) DQ2 (HLA-DQ2.2
and HLA-DQ2.5) and HLA-DQ8 are strongly associated with celiac disease (CD), a
chronic auto-inflammatory disease affecting mainly the small intestine. The α- and βsubunit of HLA-DQ2.2 are encoded by the HLA alleles DQA1*02:01 and
DQB1*02:02, whereas DQ2.5 is encoded by DQA1*05:01 and DQB1*02:01 (cis
conformation) or DQA1*05:05 and DQB1*02:02 (trans conformation). Patients with
DQ8 carry the alleles DQA1*03:01/02/03 and DQB1*03:02. Homozygous carriers
for DQ2.2, DQ2.5 or only the DQ2 β-subunit have a higher risk of disease
development compared to heterozygous individuals. In this study the concordance
rate between the new EUROArray HLA-DQ2/DQ8-h Direct and the Olerup
SSP tests in the detection of CD relevant HLA-DQ alleles and gene dose of DQ2 is
evaluated in 81 CD suspected individuals. Methods: In the EUROArray
test, sequence specific amplification of selected HLA-DQ gene sequences is
achieved by 2 parallel multiplex PCRs with simultaneous fluorescence labelling of
the reaction products for the subsequent hybridization. All relevant HLA-DQ
genotypes are automatically deduced by the EUROArrayScan software. In the
Olerup test systems the determination of the relevant HLA-DQ alleles is achieved by
72 parallel singleplex PCRs with a subsequent gel electrophoresis and semiautomated result output of the possible genotype interpretations. Results: The study
with 81 samples suspected of CD showed 100% concordance between the 2 test
systems regarding the determination of DQ2 and DQ8: 27.2 % DQ2.5 heterozygous,
9.9 % DQ2.5 homozygous, 3.7 % heterozygous DQ2.2, 7.4 % DQ2.2 + DQ2.5, 13.6
% DQ8, 7.4 % DQ2.5 + DQ8 and 3.7 % DQ2.2 + DQ8 and 27.2 % negative. A
DQ2.2 homozygous genotype was not found in this cohort. One of the DQ2.2 +
DQ2.5 samples contains only one copy of the DQ2 beta-subunit and therefore gene
dose for CD associated HLA-DQ2 antigens corresponds only to a heterozygous
status. Conclusions: The obtained frequencies correspond to previous studies in
similar cohorts. The concordance rate of 100% obtained for the 2 tests confirms
highest reliability of both tests for determination of patients at risk for CD, including
discrimination of CD relevant homozygous and heterozygous status. In comparison
with the Olerup test system the EUROArray test is easier and faster to perform since
it is based on 2 multiplex PCRs only and subsequent array based analysis. The data
evaluation for EUROArray is fully automated and standardized, thereby eliminating
tedious manual interpretation of 72 PCR results per sample as required for the
Olerup test.
G30. Improved Screening for Cancer Predisposition Mutations in Patients with
Advanced Solid Tumors Enabled by Tumor-Normal Sequencing
D. Mandelker, L. Zhang, Y. Kemel, Z. Stadler, V. Joseph, A. Zehir, N. Pradhan, A.G.
Arnold, M. Walsh, O. Birsoy, Y. Li, S. Jairam, A. Balakrishnan, A. Syed, M.
Prasad, M. Berger, M. Ladanyi, R. Mark
Memorial Sloan Kettering Cancer Center, New York, NY.
Introduction: A proportion of cancer patients carry a hereditary predisposition to
developing cancer, and identifying these patients is critical for the proper screening
of at-risk family members. Guidelines have been established based upon family
history and clinical and pathologic features to determine which cancer patients
should have germline testing. However, it is not known how many pathogenic
germline variants would be detected through tumor-normal analysis of all cancer
patients as compared to guideline-based selection of patients for germline testing.
Methods: Patients undergoing multi-gene panel tumor-normal testing for somatic
tumor profiling (MSK-IMPACT; 410 or 468 genes) of solid cancers were given the
option to consent to germline testing for 76 genes included in the panel that are also
associated with hereditary cancer predisposition. Germline variants were called in
the normal DNA sequenced for MSK-IMPACT tumor testing, classified based on
ACMG criteria, and pathogenic and likely pathogenic variants were included on a
clinical report. Results: Three thousand thirty-eight patients received both somatic
and germline testing in the context of tumor-normal sequencing. 16.8% of the
patients were found to carry a pathogenic germline variant that confers hereditary
susceptibility to cancer. Approximately 1/3 of these pathogenic variants were known
founder mutations from various ethnicities. Across cancer types, 12% of all cancer
patients had germline pathogenic variants in DNA repair genes, a finding with
therapeutic implications, as these patients may be more responsive to PARP
inhibitors currently in clinical trials. Moreover, performing tumor-normal sequencing
allowed us to determine loss of heterozygosity at the locus of the germline
pathogenic variant, and distinguish between germline variants and clonal
hematopoiesis. Finally, a detailed analysis of a subset of our cases showed that
limiting testing to patients meeting guideline criteria would have failed to detect
949
AMP Abstracts
approximately half of the pathogenic germline variants detected through tumornormal sequencing. Conclusions: Nearly 17% of our advanced cancer patients
undergoing tumor-normal sequencing were found to harbor a pathogenic germline
variant for an inherited cancer predisposition syndrome. These findings have led to
detection of these pathogenic variants in family members, screening and
prophylactic surgeries in the carriers, and change of treatments in the cancer
patients. Moreover, a substantial fraction of these variants would not have been
detected if genetic testing were solely performed based upon established clinical
guidelines, a finding that strongly supports germline testing for all advanced cancer
patients.
G31. Importance of Whole Exome Sequencing in Solving Complex
Phenotypes: A Case Report
R.M. Minillo1, M.C. Cervato1, M. Caraciolo2, N.H. Muto1, J.R. Pinho1, C.P.
Mangueira1, J.B. Oliveira Filho2, R. Sitnik1
1Hospital Israelita Albert Einstein, São Paulo, Brazil; 2Genomika Diagnostics, RecifePE, Brazil.
Introduction: In this study our aim is to emphasize the importance of the indication
for whole exome sequencing (WES) for investigation of cases with mixed / complex
phenotypes not described in the literature. Methods: An 11-month-old male patient
presenting polycystic kidney disease, childhood hepatobiliary fibropolystic disease,
concentric cardiomyopathy, and immunological immaturity. He was the first son of a
non-conscript couple with no family history. Given the complexity of his clinical
phenotype, WES (capture with Agilent SureSelect All Exon V4 kit and sequencing on
Illumina NexSeq 550 platform) was indicated. Results: WES identified compound
heterozygosis on PKHD1 gene, which is associated with autosomal recessive
polycystic kidney disease. The proband had 2 probably pathogenic variants,
c.10637delT and c.3367G> A; (the first one was inherited from his mother and the
other, from his father). He also had a probably pathogenic variant in heterozygosis
on MYH7 gene (c.2063delT), which is associated with several types of autosomal
dominant cardiomyopathies (also confirmed by Sanger, inherited from his father).
Conclusions: The importance of WES for identification of known and new genes in
families segregating with Mendelian diseases is well established. It is possible to
identify variants in known genes for patients exhibiting complex phenotypes through
WES (which identifies approximately 4% to 7% of these cases), since the phenotype
can be separated into its constitutive parts and causal variants can be identified to
explain specific phenotypic characteristics. This, in turn, may allow individualized
treatment with a positive impact on patient care and a better holistic understanding of
the disease.
G32. Automated Reanalysis of Genomic Data: Challenges and the Promise of
Novel Diagnoses
J. Murrell1,2, S. Baker1, E. Denenberg1, E. Dechene1, B. Krock1, A. Santani1,2
1The Children's Hospital of Philadelphia, Philadelphia, PA; 2University of
Pennsylvania, Philadelphia, PA.
Introduction: Clinical Exome Sequencing (CES) is a diagnostic test for patients with
suspected rare genetic conditions, with a reported diagnostic yield of 20% to 30%.
As new discoveries of disease causing genes and variants are made, the utility of
CES is expected to further increase. Each year, hundreds of gene-disease and
thousands of variant-disease associations are reported, highlighting the necessity for
reevaluating non-diagnostic CES cases using the most up-to-date literature. Here,
we report the systematic reanalysis of the first 300 CES samples analyzed by the
Division of Genomic Diagnostics at the Children’s Hospital of Philadelphia
(CHOP). Methods: We developed a patient database containing detailed phenotype
data, including 2,000 clinician-selected Human Phenotype Ontology terms, patient
genotype data, and annotations for 130,000 variants. Using sophisticated algorithms,
we utilize annotations that are automatically derived from PubMed abstracts, OMIM,
ClinVar, and mutation databases and significantly reduces the time required for CES
test reanalysis. Results: We demonstrate that our reanalysis strategy reveals novel
diagnostic findings in 8.8% of previously non-diagnostic cases, illustrating the
importance of reanalysis. In 12 of the 21 cases (57.1%), a diagnosis was made
possible because new disease genes were published after the initial CES analysis.
The identification of a pleiotrophic effect of gene resulted in a diagnosis for 1 patient
(4.8%). For 4 of the cases (19.0%) a diagnosis was made possible due to additional
evidence. Additionally, we identified 4 candidate diagnoses (19.0%), for which
insufficient evidence was available to make a definitive diagnosis, though existing
evidence strongly suggested a likely role for the variant in the proband’s disease. To
estimate the frequency and workload of successive reanalyses, we retrospectively
examined the number of variants with novel annotations when reanalyses were
performed monthly between January and April 2017. On average, there were 0 to
13.67 (mean=0.97) novel annotations per month/case. Importantly, there were no
novel annotations for >55% of the cases when reanalysis was performed at monthly
intervals and 38.75% of cases were not assigned novel annotations across all 3
consecutive re-analyses. Conclusions: We present an overview on challenges
faced during infrastructure development and suggest strategies that clinical
laboratories can adopt to develop an efficient, systematic and scalable re-analysis
program. Our results demonstrate that despite several logistical and reimbursement
challenges, enabling efficient and sustainable re-analysis strategies will be valuable
for patients and in advancing scientific progress.
950
G33. Chromosome Anomalies Involving the APC Gene Lead to an Increased
Risk for FAP and Developmental Delays
B.A. Hilton1,2, J.C. Palumbos1, D.H. Viskochil1, R.M. Toydemir1
1University of Utah, Salt Lake City, UT; 2ARUP Laboratories, Salt Lake City, UT.
Introduction: Interstitial deletions of the long arm of chromosome 5 that encompass
the APC (adenomatous polyposis coli) gene are rarely reported. Individuals with
deletions in this region, typically 5q15-5q22, present with varying degrees of
developmental delay and learning difficulty, in addition to familial adenomatous
polyposis (FAP). FAP results from the inactivation of APC, a tumor suppressor, and
is clinically defined by the development of colorectal adenomas beginning in
childhood and adolescence. Individuals with deletions involving APC are considered
at-risk for FAP and begin diagnostic screening by 10 to 12 years of age. On the other
hand, the gene(s) contributing to developmental delay are currently unknown.
Mutations, and rarely deletions, in the APC gene are known to result in FAP;
however, mutations and deletions contained within the APC gene have not been
reported to result in learning disabilities. In the present study, we describe 3 patients
referred to the University of Utah Genetics Clinic with a clinical presentation of
developmental delay and congenital anomalies, who have large 5q deletions
involving the APC gene. Methods: A SNP-based genomic microarray analysis
(GMA) was implemented as the first-tier clinical diagnostic test for our patients. DNA
specimens extracted from peripheral blood were run and analyzed according to the
manufacturer’s recommendations (AffyMetrix CytoScan HD). Results: The GMA
detected interstitial 5q deletions of various sizes ranging from 18.3 to 25.3 Mb. The
smallest region of overlap (SRO) between these patients’ deletions is approximately
11.9 Mb and contains approximately 31 OMIM genes. Of the 31 OMIM genes in the
SRO none have been identified as playing a role in brain development at this time,
however, at least 2 genes (CAMK4, FER) have been associated with autism
spectrum disorders. Conclusions: We present 3 patients with rare overlapping 5q
deletions. Analysis of the SRO between these 3 patients, and a review of patients
from the medical literature, suggest at least 2 genes may play a role in the
developmental delays observed in patients with 5q deletions around the APC locus.
These cases add to the current understanding of the impact chromosome 5q
deletions encompassing the APC gene have on developmental delay and
emphasizes the importance of the continued surveillance of patients with deletions in
this region.
G34. Analysis of Cell Pellets Using the Cytoscan Dx Chromosomal Microarray
C.J. Broehm1, G. Olson2, R.R. Gullapalli3
1University of New Mexico/TriCore Reference Labs, Albuquerque, NM; 2Tricore
Reference Labs, Albuquerque, NM; 3University of New Mexico, Albuquerque, NM.
Introduction: The Cytoscan Dx chromosomal microarray is validated and approved
for the postnatal diagnosis of chromosomal abnormalities utilizing DNA isolated from
peripheral blood (PB). Cell pellets (CP) may be useful when traditional PB samples
are difficult to obtain, or are unobtainable.To assess its utility in diagnosis from CP
created during routine chromosomal analysis, DNA isolated from CP was processed
and analyzed on the Cytoscan Dx Array. Methods: DNA was isolated from 16 CP
fixed in Carnoy’s fixative from patients with chromosomal copy number changes
(trisomy, n=10; monosomy, n=2), interstitial deletions and microdeletions (n=2), and
other rearrangements (translocation and isodicentric Y, n=2). Twelve CP were
analyzed 2 or more times (range 2 to 5). DNA isolated from PB from 25 patients with
both normal and abnormal genetic findings was run concurrently. Eight PB samples
were run 2 or more times (range 2 to 6). Cell pellet and peripheral blood samples
were processed according to the Cytoscan Dx instructions using manufacturer
reagents. Analysis of results, including QC metrics (MAPD and SNPQC), was
performed on the ChAS Dx Software v1.1. Results: Of 16 CP samples, 5 (31%)
failed at least one QC metric and could not be further analyzed. Three failed both
MAPD and SNPQC metrics and 2 failed only one QC metric (SNPQC in both
instances). Each failed CP sample was run at least twice (range 2 to 4) and failed on
each subsequent attempt. Of the CP samples that passed QC, 3 initially failed
SNPQC and passed MAPD, but passed both QC metrics on subsequent runs. Gel
electrophoresis performed for PCR products size distribution and fragmentation
assessment showed appropriate size distribution of products for all CP samples with
no distinct bands. Of CP samples which passed QC and were subsequently
analyzed (n=5), all showed results concordant with routine chromosomal analysis or
chromosomal microarray performed at an outside institution. Two of 25 PB samples
failed SNPQC on initial testing, but passed both MAPD and SNPQC on retesting. No
other PB samples failed. Median QC metrics for failed CP were 0.281 (MAPD) and
6.664 (SNPQC), for passed CP 0.183 (MAPD) and 17.729 (SNPQC), for passed PB
0.182 (MAPD) and 18.212 (SNPQC), and for failed PB 0.200 (MAPD) and 11.642
(SNPQC). Conclusions: Cell pellets overall perform poorly when analyzed on the
Cytoscan Dx chromosomal microarray. Failed QC metrics, particularly consistent
poor SNPQC, suggest degradation of the DNA due to processing and/or fixation
during cell pellet production, though this was not evident on gel electrophoresis.
These data suggest peripheral blood is preferable for microarray studies and cell
pellets should only be used if no alternative is available.
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
G35. Interindividual Variability of Delta-9-Tetrahydrocannabinol Metabolism by
CYP2C9 Polymorphism and Possible CYP3A Inhibitors
M. Nakano, V. Pham, T. Nappi, T. Hsieh, Z. Hsu, C.J. Sailey
Molecular Testing Labs, Vancouver, WA.
Introduction: Delta-9- tetrahydrocannabinol (Δ9-THC) is primarily responsible for
psychoactive effects like, euphoria and paranoia when smoking marijuana. Medical
use of marijuana is legalized in many states, with the purpose of controlling anxiety,
nausea, pain, and enhancing appetite. THC is known to cross the blood-brain
barrier, and the pharmacokinetics (PK) and pharmacodynamics (PD) of THC may
provide crucial data to predict the psychological adverse effects. CYP2C9 is
predominantly responsible for THC metabolism, and to a lesser extent by CYP3A4/5.
CYP2C9 is a polymorphic drug metabolizing enzyme, and CYP3A isozymes are
prone to inhibiton by several drugs. Analyzing polymorphisms of CYP2C9 and drug
interactions mediated by CYP3A is crucial to understanding the PK of THC, and
predict the possible PK alterations to prevent undesirable psychiatric events by
smoking marijuana. Methods: Published articles discussing PK, enzyme kinetic, and
in vitro data were searched with several keyword combinations: CYP2C, CYP3A,
THC, in vitro, in vivo, cannabis, polymorphisms, intravenous, oral smoking, Δ9-THC,
AUC, Vmax, Km, THC metabolism, enzyme kinetics, Dronabinol and Marinol.
Pharmacokinetic data was utilized for static PK prediction, and compared to
observed PK data. Results: A total of 9 published THC articles on PK data in
intravenous, smoking, and oral administration routes were accumulated. The
average and dose-normalized AUC were 6,950 min/mL, 136 min/mL, and 71
min/mL, and the terminal THC clearance was 147 mL/min, 8.6 µg/min, and 1.1
µg/min, respectively. The observed clearance was reduced to 50% of wild-type (WT)
by carrying 1 copy of CYP2C9*3 and less than 40% of WT by carrying 2 copies of
CYP2C9*3. The predicted hepatic clearance by CYP2C9 based on the in vitro data
was 960 L/hr, and the calculated fraction of metabolism mediated by CYP2C9 was
0.68. CYP3A5*1 genotype increased the THC clearance slightly (15% increase),
which may not be clinically important. Possible drug-drug and drug-gene interactions
were explored with the well-starred model. Conclusions: This study summarized
pharmacokinetics of Δ9-THC, a fraction of metabolism mediated by CYP2C9, and
the role of CYP2C9 THC metabolism. CYP2C9 variants may contribute to an
increase in THC levels in plasma, and result in a higher risk of psychiatric adverse
effects. Challenges of predicting gene-drug interactions by CYP2C9 and drug-drug
interactions by CYP3A isoenzymes were outlined.
G36. An Atypical Presentation of a Homozygous Delta-F508 Mutation
O. Rouhi1, T. Schneider2
1Emory University Hospital, Atlanta, GA; 2EGL Genetics, Tucker, GA.
Introduction: In approximately 10 percent of patients with cystic fibrosis, the disease
manifestations appear to be limited to one organ system; however, the variability is
usually attributed to milder CF variants. Patients homozygous CFTR delta-F508
mutation, tend to have an earlier and more severe clinical manifestation. Methods: A
15-year-old Caucasian boy presented to primary care with chief complaint of short
stature and a history of asthma. From the short stature workup, abnormal liver
enzymes were detected and further investigated. Hepatic ultrasound and subsequent
liver biopsy findings were compatible with hepatic cirrhosis. Next generation
sequencing (NGS) based genetic analysis for genetic causes of cholestasis was
ordered after genetic consultation. Two separate blood draws were obtained.
Results: The genetic analysis demonstrated homozygous CFTR
c.1521_1523delCTT (p.508del). The Sanger sequencing confirmed the alteration on
both blood samples. The possibility of a sample mix-up was investigated but after
thorough investigation no red-flags were identified. The patient was also found to be
heterozygous for SMPD1 c.785_807, pathogenic for Niemann-Pick disease type B.
Conclusions: The patient was born in Texas in 2001. The state of Texas did not
establish new born screening for cystic fibrosis until 2009. This is the first reported
case of newly diagnosed cystic fibrosis-related liver disease in a teenage boy,
without severe respiratory manifestations harboring a homozygous delta F508
mutation.
G37. Expert Review of NGS Results Removes Need for Routine Sanger
Sequencing Confirmation
D. Muzzey, K.R. Haas, L. Melroy, G.J. Hogan, S. Kash, J. Johnson, E. Olson, P.
Krenesky, S. Siddhanti, C.S. Chu, E.A. Evans, K. Ready, H.P. Kang
Counsyl, South San Francisco, CA.
Introduction: Variant calling with next-generation sequencing (NGS) incurs rare but
systematic mistakes. In hereditary cancer testing, such mistakes have enormous
consequences as they could lead to incorrect screening or surgical management
decisions. It has been suggested that routine orthogonal confirmation with Sanger
sequencing is required to verify NGS results, especially for sites where the allele
balance is between 10% and 30% (Mu et al., 2016). We sought to demonstrate that
manual review of NGS data can also identify and eliminate false calls otherwise
permitted by automated call-quality thresholds. Methods: For over 20,000 Counsyl
patient samples tested on our expanded hereditary cancer panel, we evaluated our
custom software variant-review interface, which enables licensed reviewers to
perform manual inspection of detailed quality-control criteria for batch-, sample-, and
variant-level QC metrics, including evaluation of NGS read pileups. Samples with
putative heterozygous variant calls with allele balances below 30% were subjected to
Sanger sequencing, and the respective outcomes of Sanger sequencing and manual
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
NGS review were compared. Results: There were 3,937 putative heterozygous sites
where at least one patient had a pathogenic variant or variant of uncertain
significance, and calls at all such sites underwent manual review, which focused on
the position, strand, and capture origin of the reads in the NGS pileup. Forty-seven
(1.6%) sites had samples with variant calls at <30% allele balance (2003 calls in
total), and call review revealed the source of the unexpected allele balance, allowing
>99% of calls (1991) at these sites to be classified as true positives (enriched for
long indels and homopolymers) or true negatives (often conspicuous NGS artifacts).
The remaining <1% of calls (12) appear to be mosaic. Sanger sequencing on all of
the sites verified that expert manual review correctly classified all such variants as
positive, negative, or mosaic. Conclusions: Clinical NGS variant calling requires
great care to avoid false reporting of systematic artifacts, yet this care can manifest
in different ways. If a lab chooses to forgo careful review of NGS data, routine
orthogonal methods such as Sanger sequencing should indeed be used.
Alternatively, we have demonstrated that a highly trained NGS expert operating
within an optimized and platform-specific software review interface achieves
comparable quality to routine Sanger confirmation.
G38. High Prevalence of Alpha-1 Antitrypsin Z Alleles in Formalin-Fixed
Paraffin-Embedded Liver Explant Tissue with PAS-D Globules
L. Pac, G. Cheeney, E.Q. Konnick, D.N. Greene, M. Westerhoff, C. Lockwood
University of Washington, Seattle, WA.
Introduction: Alpha-1 antitrypsin (A1AT) deficiency is an autosomal co-dominant
genetic disease with incomplete penetrance that can cause pulmonary and liver
disease. The diagnosis of A1AT deficiency is most commonly established by low
serum A1AT concentration and/or isoelectric focusing (IEF) of serum proteins.
Approximately 10% of liver explants from patients with end stage liver disease
(ESLD) contain cytoplasmic globules within hepatocytes that stain magenta with
periodic acid-Schiff-diastase (PAS-D). Since PAS-D stain nonspecifically highlights
polysaccharides, the significance of this finding is often unclear and confirmatory
testing for A1AT deficiency is infrequently performed following transplant. In order to
explore the relationship between PAS-D globules and SERPINA1 (the gene
encoding A1AT) genotype, we validated an assay for detecting the most common
pathogenic alleles (S and Z) in formalin-fixed paraffin-embedded (FFPE) liver tissue.
Methods: Genomic DNA was extracted from archived FFPE tissue from patients
who underwent liver transplantation at the University of Washington Medical Center
between 2006 and 2015 with (n=26) and without (n=8) PAS-D globules. Following
validation experiments with FFPE samples with known A1AT genotypes, real-time
PCR (RT-PCR) was used to detect the SERPINA1 S and Z
(NM_000295.4:c.863A>T, p.E288V and c.1096G>A, p.E366K, respectively) alleles
by melting curve analysis. Results: Among all causes of ESLD, 85% of cases with
PAS-D globules were homozygous (n=2) or heterozygous (n=20) for the Z allele. In
patients who underwent liver transplantation for non-alcoholic steatohepatitis
(NASH), 92% of cases with PAS-D globules were homozygous (n=1) or
heterozygous (n=11) for the Z allele. In contrast, among liver explants without PAS-D
globules (n=8), no Z alleles were detected. One specimen without PAS-D globules,
from a patient without clinical suspicion for A1AT deficiency, was heterozygous for
the S allele. Conclusions: PAS-D globules identified in liver explant specimens are
frequently associated with at least one Z allele of SERPINA1. Previous work by our
group showed that PAS-D globules occur significantly more often in explants
performed for NASH than for other causes of ESLD. Both histopathologic evaluation
and genotyping from archived FFPE tissue suggest that NASH and the Z allele may
have a synergistic effect on ESLD progression. With new therapies for Hepatitis C,
NASH is expected to become the leading cause of liver transplantation in the U.S.
Further studies are needed to identify risk factors for progression to ESLD, including
whether the presence of a SERPINA1 Z allele may identify a distinct subset of NASH
patients.
G39. Clinically Relevant Findings from Pharmacogenomic Testing in >36k
Patients Across Multiple Diagnoses
J.P. Jarvis1, J.A. Shaman2, S. Zajic1, E. Avaniss-Aghajani3
1Coriell Institute for Medical Research, Camden, NJ; 2Coriell Life Sciences,
Philadelphia, PA; 3Primex Clinical Laboratories Inc., Camden, NJ.
Introduction: Precisely how and when pharmacogenomic (PGx) testing is beneficial
in the continuum of care remains an open research question even though it is
unequivocal that some individuals respond to pharmaceutical interventions in
unexpected ways due to their personal genetic profile. Further, it remains unclear
how often PGx results will provide information directly relevant to a patient’s actual
treatment plan. Methods: Here, we present an analysis of PGx testing results for
>36k diverse individuals taking multiple medications, with and without 4 specific
diagnoses: heart disease, cancer, diabetes, and chronic pain. Using multiple
approaches, we evaluated differences in personalized, regimen-specific information
that can be provided across diagnoses and the potential impact of specific drug-gene
interactions on healthcare delivery. Results: We find that clinically relevant
information directly related to drugs prescribed to specific patients can be provided to
physicians 40.7% of the time. For heart disease patients specifically the rate was
48.0%, for cancer patients it was 44.1%, for diabetics it was 43.3%, and for those
undergoing pain management it was 42.1%. In addition, results include confirmatory
findings supporting standard therapy for most patients, an often overlooked benefit of
PGx testing. We identified 55 unique pharmacological agents that triggered delivery
951
AMP Abstracts
of patient-specific decision support information with 8 medications (Metoprolol,
Clopidogrel, Tramadol, Carisoprodol, Alprazolam, Celecoxib, Thioridazine, and
Ketoconazole) implicated in potential genetic interactions with the highest levels of
importance for clinical care. Variation in PGx utility was also noted across ethnic
groups. Conclusions: In summary, we found substantial evidence supporting the
potential benefits of genetic testing in several diagnostic contexts and with respect to
polypharmacy more generally.
G40. Analytical and Clinical Validation of Variants Identified by Exome
Sequencing through Secondary Review and Sanger Confirmation in a CLIACertified Molecular Laboratory
N.T. Strande, M. Li, J. Booker, J.P. Evans, J.S. Berg, K.E. Weck
University of North Carolina at Chapel Hill, Chapel Hill, NC.
Introduction: As clinical laboratories begin offering genome-scale sequencing as a
diagnostic test for monogenic disease, they must consider both its analytical and
clinical validity. With the large volume of information generated by this technique,
laboratories must establish a method to efficiently and accurately evaluate these
metrics. Here we discuss our tiered evaluation of exome sequencing (ES) results in
the North Carolina Clinical Genomic Evaluation by Next-generation Exome
Sequencing (NCGENES) project. Methods: During the primary review step of the
NCGENES analysis workflow a molecular analyst identifies candidate causative
variants that are consistent with the patient’s clinical presentation or are clinically
actionable for review by the “sign-out” committee. This committee, comprised of
clinical geneticists, molecular laboratory professionals, genetic counselors, and
Ph.D. level variant analysts, reviews the candidate variants and decides which to
confirm in the CLIA laboratory and ultimately return to the patient. CLIA confirmation
of the variants includes Sanger sequencing and secondary review of the variant for
evidence of clinical impact. Results: A total of 400 variants (377 unique)
representing 276/669 cases were selected for CLIA confirmation. The Sanger
confirmation rate for all 400 variants was 96.25% (385/400) and was 96.02%
(385/377) for unique variants. The majority of variants that did not Sanger confirm
(8/15) were in genes associated with cancer, that were often homologous to
pseudogenes. All but 4 variants that did not confirm were single nucleotide variants.
Only 5 of the total 400 variants (1.25%) sent for CLIA confirmation were ultimately
not returned due to lack of clinical significance upon secondary review of the
evidence. Importantly the secondary review process frequently led to modifications
of the variant's clinical interpretation. Conclusions: Both the clinical and analytical
validity of ES for NCGENES was greater than 95%, consistent with previous reports.
Variants identified by ES within genes with high sequence homology to pseudogenes
were often challenging to sequence and in some instances Sanger sequencing
demonstrated that the ES identified variant was located within a pseudogene rather
than the gene of interest. For confirmed variants, the tiered review process utilized in
NCGENES allowed the CLIA laboratory to solely focus on variants likely to be
reported, which was important to ensure accuracy regarding the clinical
interpretation. As clinical laboratories begin utilizing genome-scale sequencing for
clinical diagnosis, both clinical and analytical validity must be considered. We
present a tiered review process that may be particularly useful when considering
clinical validity of ES results.
G41. The Mother of all Confounders: Strategies to Avoid False Positives
Caused by Maternal Copy Number Variants in Noninvasive Prenatal Screening
K.E. Kaseniit, G. Hogan, K. D'Auria, C. Haverty, D. Muzzey
Counsyl, South San Francisco, CA.
Introduction: Maternal copy number variants (mCNVs) have been reported to be a
major source of false positives in noninvasive prenatal screening (NIPS). Maternal
duplications of 250kb and larger were previously predicted to increase the false
positive rate by over 40-fold, and up to one-half of trisomy 13, 18, and 21 false
positives were attributed to maternal duplications (Snyder et al. NEJM 2015, Strom
et al. NEJM 2017, Chudova et al. NEJM 2016). Here, we characterize the impact of
mCNVs on NIPS analysis approaches and outline best practices for robust fetal
aneuploidy screening via whole-genome sequencing
(WGS). Methods: Retrospective analysis of NIPS data from nearly 50,000 samples
identified duplications of 250 kb and larger in ~2% of samples on the chromosomes
of interest (13, 18, 21). In these data, 99% of maternal duplications spanned less
than ~4% of the respective chromosome; due to the rarity of large mCNVs that might
have a large impact on specificity, we systematically assessed the impact of mCNVs
on NIPS results using simulations. WGS data from non-trisomic NIPS samples with
no detectable mCNVs were used to create simulated samples harboring maternal
duplications ranging from 0.1% to nearly 100% of the chromosome. The drop in
specificity due to spurious false positives was characterized for several analysis
approaches as a function of the size of the duplication. An overall false positive rate
was estimated based on the prevalence of differently sized duplications obtained
from the WGS data of real NIPS samples. Results: An aneuploidy calling algorithm
lacking outlier-robustness features was estimated to yield 32 false positives per
100,000 samples due exclusively to maternal duplications. Employing naive outlierrobust approaches—for example, using the median rather than the mean where
appropriate and removing outliers with respect to known underlying distributions—
lowered the false positive rate to 6 in 100,000, a more than 5-fold improvement.
Finally, specific identification and removal of mCNVs during NIPS analysis reduced
the expected mCNV-attributable false positive rate to less than 1 in a
952
million. Conclusions: High specificity in NIPS can be achieved even in the presence
of mCNVs that range in size when the bioinformatics algorithm specifically accounts
for mCNVs. Due to the rarity of large mCNVs, analysis pipelines must be rigorously
characterized via simulations to ensure high performance. High specificity, which
enables higher positive predictive value, is critical to preserve clinical utility of NIPS
as adoption increases in the average-risk population.
G42. The Analysis of Oral Microbiome in CytoScan Assay Performance
D. Lizarraga, D. Lizarraga, B. Eynon, Y. Lin, K. Suyenaga, G. Mamtora, C. Chen, J.
Cuevas, A. Roter, R. Duttagupta, E. Fung
Affymetrix, Inc., as part of Thermo Fisher Scientific, Santa Clara, CA.
Introduction: The Applied Biosystems CytoScan Assay provides broad coverage
and superior performance for detecting chromosomal copy number aberrations.
While performance of CytoScan Assay has been extensively characterized in
genomic DNA (gDNA) from peripheral whole blood, saliva-derived DNA has not been
systematically evaluated on the CytoScan Assay. DNA extracted from oral samples,
can contain a large percentage of DNA from the oral microbiome. In this study, the
focus was to find the upper limit of microbial DNA contamination that has impact on
the CytoScan Assay. Methods: gDNA was spiked with DNA from 4 different
microbes known to be common in the oral cavity (Streptococcus mutans,
Staphylococcus epidermis, Lactobacillus casei and Candida albicans) at different
percentages (20%, 50% 60%, 70%, 80% and 90%) and compared with unspiked
samples in the CytoScan Assay. CytoScan assay performance was assessed by
measuring array QC metrics (MAPD, SNPQC and wSD) across all the tested
conditions (n=192) to evaluate the impact of microbial DNA on the assay. Results:
All the samples analyzed passed the CytoScan Assay in-process QC metrics criteria
across the 4 groups of microbial DNA and at all levels of microbial contamination.
Analyses of array-based QC metrics showed statistically significant differences in
MAPD and SNPQC that were correlated with increased levels of microbial
contamination. CytoScan Assay failure occurred when microbial contamination
exceeded 50% in majority of the evaluated strains. For Streptococcus mutans array
failures were observed at a concentration greater than 20%. Conclusions: In
conclusion, DNA samples containing between 20- 50% of microbial DNA content met
array-based QC metrics. This suggests that genomic DNA derived from oral samples
such as saliva and buccal swabs are compatible with the CytoScan Assay when
microbial DNA contamination is below this level.
G43. Second Specimen Testing for TP53 Variants
J. Bissonnette1,2 , M. Marshall1 , K. Barber2 , R.T. Klein2 , J. Zhao1,2 , K.S. Hruska1
1GeneDx, Gaithersburg, MD; 2BioReference Laboratories, Elmwood Park, NJ.
Introduction: As tumor sequencing (TS) is becoming integral for personalized
cancer care, many variants in TP53 will be identified as this gene is commonly
mutated across a wide variety of solid and myeloid cancers. TS is performed to
identify actionable somatic alterations, but these results may contain both somatic
and germline variants. Germline TP53 pathogenic variants cause Li-Fraumeni
syndrome (LFS), a dominantly-inherited hereditary cancer syndrome. Second
specimen testing in a blood sample differentiates between variants confined to the
tumor and germline variants that are causative for LFS. Conversely, next-generation
sequencing (NGS) for hereditary cancer syndromes from blood may identify mosaic
variants in TP53. Testing of an alternate tissue, such as cultured fibroblasts, may
determine if the variant is isolated to hematopoietic cells or is potentially
constitutional. We aim to determine the extent to which TP53 variants identified by
TS or mosaic TP53 variants identified on hereditary testing are somatically acquired
rather than a constitutional event. Methods: We retrospectively reviewed all cases
submitted for germline analysis for which prior TS identified a TP53 variant.
Information on the allele frequency of TP53 variant in the tumor specimen was
unavailable in the majority of cases. In a second cohort, we reviewed all cases for
which testing of an alternate tissue type was requested after mosaicism for one or
more TP53 variants was identified on multi-gene hereditary cancer testing in our
clinical diagnostic laboratory. Results: Fifty-three variants were identified in 47
individuals referred for testing of one or more TP53 variants initially identified in a
tumor, with 6 of these individuals having 2 TP53 variants. Upon follow up testing,
only 2 (2/53, 3.7%) of the TP53 variants were identified in blood or fibroblasts. For 10
individuals who had a mosaic TP53 variant identified by an inherited cancer panel,
testing of an alternate tissue was performed. Nine of 10 mosaic variants initially
identified in blood or oral rinse specimens were absent in fibroblasts and one was
identified in fibroblasts in the mosaic state. Conclusions: The majority
of TP53 variants identified on TS were absent in the germline suggesting that they
are confined to the tumor. In addition, the majority of TP53 variants identified in the
mosaic state on hereditary panel testing were absent in the second specimen
suggesting that they are not typically constitutional. This data indicates that low
level TP53 variants and TP53 variants identified in a tumor are unlikely to be
constitutional; however, additional studies are needed to help clinicians distinguish
when further testing for a TP53 variant identified in a tumor or in the mosaic state is
warranted.
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
G44. Runs of Homozygosity (ROH) Reveal that Segmental-UPD Occurs as a
Result of Recombination Mediated Repair of Genomic Imbalance
A.L. Penton, S. Schwartz, J. Tepperberg, P. Papenhausen
Laboratory Corporation of America, Durham, NC.
Introduction: Whole chromosome UPD results from correction of trisomy and
monosomy during embryogenesis. In contrast, segmental UPD (segUPD) is localized
to specific chromosomal regions and the etiology and risks are not well understood.
We show that segUPD occurs secondarily to recombination mediated selection
driven repair of distinct genomic imbalances including deletions, derivative
chromosomes and inverted-duplication/deletions. Although the genetic lesion may be
“repaired,” segUPD is associated with residual clinical risks. Methods: The
Affymetrix Cytoscan HD single nucleotide polymorphism (SNP) array was used to
detect runs of homozygosity (ROH) associated with UPD. Cases were collected
using criteria of positive UPD testing results, evidence of prior genetic abnormality
located at the location of the ROH, mosaicism for the ROH, or ROH associated with
a contiguous triplication. Results: Three patients displayed evidence of segUPD due
to correction of genomic imbalance during fetal development. A newborn showed an
11.1 Mb terminal ROH on chromosome 10 that replaced a terminal deletion of
10q26.13->qter observed in CVS. A 9 year old referred for developmental delay
carried a 9.4 Mb terminal ROH on chromosome 1 that was rescued from a
der(1)(t(1;17)(p36.3;q21) detected in amniocytes. NIPT studies and microarray
analysis from a placental biopsy in a woman referred for prenatal diagnosis showed
an 8.7 Mb terminal deletion and 3.4 Mb contiguous duplication replaced with a 16.3
Mb ROH on chromosome 1. Two additional patients showed evidence of partial
correction including an adult male referred for developmental delay with 2 cell lines:
10 cells had a derivative chromosome 21 (46,XY,der(21)t(12;21)(p11.22;q22.2) while
the remaining were 46,XY. Concurrent microarray analysis revealed a mosaic ROH
in addition to a mosaic 7.1 Mb 21q terminal deletion and 30.5 Mb 12p terminal
duplication. Microarray analysis from a 12-year old with growth delay showed
evidence of 3 deletion repair segUPD cell lines initiated at different mitotic
recombination sites in chromosome 15 proximal to and driven by selection against
an interstitial deletion [del(15)(q25.1q25.3)] present in 25% of cells. Conclusions:
Our data suggest that segUPD occurs secondarily to recombination mediated
selection based repair of genomic imbalance. However, presence of clinical
phenotypes indicate that the original imbalance perturbs early development, is
present in some cells or may unmask a recessive disorder. Importantly, CVS and
amniotic analysis in concert with NIPT studies suggest that the incidence of segUPD
mediated correction is underestimated, and may explain the etiology of clinical
phenotypes that are undetected my routine microarray analysis and whole exome
sequencing studies.
G45. Comparison between Different Activity Score Models for CYP2D6
Phenotype and Frequencies of Actionable Combined Genotypes of CYP2D6
and CYP2C19
M. Nakano1, M. Thai2, T. Hsieh1, C.J. Sailey1, M. Suzuki2
1Molecular Testing Labs, Vancouver, WA; 2Pacific University, Hillsboro, OR.
Introduction: CYP2D6 and CYP2C19 metabolize approximately half of the drugs in
the US market. Both CYP2D6 and CYP2C19 genes are highly polymorphic and are
important to be tested when predicting clinical outcomes of drugs that are highly
metabolized by them. The CYP2D6 phenotype, based on its activity scores, is
determined by allelic variants and gene copy numbers; however, assigning activity
scores is defined by 2 models (Models A and B). Model B assigns *9, *29, *45 and
*46 alleles 0.75, instead of 0.5 (as in model A). Thus, in this study, we retrospectively
analyzed a large genetic and ethnic data set to elucidate the clinical impact based on
2 different models of CYP2D6 activity scores, and summarized the frequencies of
actionable gene combinations of poor metabolizers (PM) and ultrarapid metabolizers
(UM) in CYP2D6 and CYP2C19. Methods: A total of 85,822 de-identified samples
with CYP2D6 and CYP2C19 genotypes, star alleles, gender, ethnicity, and
prescribed medications was queried from Molecular Testing Lab's clinical database.
Translation of genotype and phenotype were compared based on models A and
B. All analyses (student t-test, chi-square test and ANOVA) were performed in R and
GraphPad Prism. Results: The analyzed populations (n=44187) that met the
inclusion criteria consisted mainly of Caucasian (65%), Hispanic/Latino (17%) and
African-American (12%). The most prevalent star-alleles of CYP2D6 across the
whole population was *1 (f=0.39), *4 (f=0.156) and *2 (f=0.186). Interestingly,
African-American, Hawaiian and Pacific Islanders had a lower probability of PM
phenotypes compared to other ethnic background populations. The differences of
phenotype distributions by model A and B were not significantly different based on
the chi-square test. The most clinically important population, PM/PM, PM/UM,
UM/PM and UM/UM in CYP2D6/CYP2C19 were 0.1%, 0.1%, 1.6% and 0.6%,
respectively. The frequency of population carrying either PM or UM of CYP2D6 and
CYP2C19 was 35.5%. Conclusions: The comprehensive and retrospective analysis
summarized the frequency of alleles and predicted phenotypes based on 2 different
models in major populations in the US. The results yielded that no significant
differences in defined CYP2D6 phenotypes existed in either model. The frequency of
clinically important phenotypes, UM and PM, of CYP2D6 and CYP2C19 was as high
as 35.5%, which supports the importance of genetic testing of these 2 enzymes prior
to prescribing medication.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
G46. Mutation Spectrum of the KCNQ1, KCNH2, and SCN5A Genes for the
Long QT Syndrome in Korea
M. Kim1, S. Kim2, E. Bae1, C. Noh1, S. Cho1, M. Seong1, S. Park1
1Seoul National University Hospital, Seoul, South Korea; 2National Medical Center,
Seoul, South Korea.
Introduction: Long QT syndrome (LQTS) is an inherited cardiac disease with
multiple causative genes and characterized by QT interval prolongation, arrhythmia,
and sudden cardiac death. Among Koreans, both the prevalence of symptomatic
patients and apparently healthy but with abnormal electrocardiogram subjects are at
least as high as in other ethnicities; however, their genetic mutation spectrum has
not yet been elucidated comprehensively. We aimed to characterize the mutation
spectrum and characteristics in Korean LQTS patients through analyzing 3 causative
ion-channel-encoding genes. Methods: For 57 unrelated LQTS probands and their
67 family members, mutations in the genes KCNQ1, KCNH2, and SCN5A were
identified by Sanger sequencing and gene dosage analyses using multiplex ligationdependent probe amplification. Results: This study revealed that 45.6% of probands
(26 patients) harbored causative mutations in one of the LQTS
genes: KCNQ1 (53.8%, 14 probands), KCNH2 (23.1%, 6 probands), SCN5A (23.1%,
6 probands). The KCNQ1 exon deletions were frequent (28.6%
of KCNQ1 mutations, 4 probands). The KCNH2 mutations were clustered in exon 7
(66.7%). The SCN5A mutations were much frequent than in previous reports. De
novo mutation rate was 11.5%. Conclusions: This study indicated that
the KCNQ1 is the most important causative gene among Koreans, and mutations
were heterogeneous in 3 genes. Mutation was not identified in approximately half of
the probands, therefore other multi-gene panel tests may be helpful.
G47. Spectrum of MNX1 Mutations in Korean Patients with Currarino
Syndrome
S. Lee1, E. Kim2, S. Cho2, H. Park3, S. Seo4, S. Jung5, S. Lee5, K. Park5, H. Kim5, S.
Park2, M. Seong2
1Kangbuk Samsung Hospital, Seoul, South Korea; 2 Seoul National University
Hospital, Seoul South Korea; 3Gyeongsang National University Changwon Hospital,
Changwon Kyongsang, South Korea; 4Seoul National University Bundang Hospital,
Seoungnam Kyongki, South Korea; 5Seoul National University College of Medicine,
Seoul, South Korea.
Introduction: Currarino syndrome (CS) is a congenital malformation syndrome
typically characterized by sacral agenesis, anorectal malformation, and presacral
mass. Mutations in the MNX1 (motor neuron and pancreas homeobox 1) gene are
known to be the major genetic background of CS. These mutations are identified in
almost all familial cases and 30% of sporadic cases. Less commonly, a large
deletion or a complex rearrangement involving the 7q36 region is associated with
CS. In this study, we investigated the mutational spectrum of MNX1 gene in Korean
patients with CS. Methods: We enrolled 25 patients with CS, including 24 sporadic
cases and 1 familial case. Direct sequencing of all coding exons and flanking intronic
sequences of the MNX1 gene was performed. Results: We identified 6 null variants
and 1 missense variant. Six of these variants were novel. The null variants included
4 frameshift mutations (p.Gly98Alafs*124, p.Gly145Alafs*77, p.Gly151Leufs*67, and
p.Ala216Profs*5) and 2 nonsense mutations (p.Tyr186* and p.Gln212*). The
missense variant, p.Lys295Gln, was located in the well-conserved homeobox
domain and predicted to be deleterious. MNX1 gene mutations have been implicated
in 28% of all cases and 25% of sporadic cases. The clinical phenotype was variable
in mutated and non-mutated patients and no significant genotype-phenotype
correlation was observed. Conclusions: This study reveals the mutation spectrum
and phenotypic variability of MNX1 mutations in the Korean population.
G48. WITHDRAWN
G49. Genetics Insights into Hereditary Cancer Risk in the Latin American
Population
A. Leon1, A. Zimmer1, W. McFadden1, J. Rugeles2, E.S. dos Santos3, H.C.
Garcia4, F. Neffa5, F.J. Valdez6, V. Bernath7, A. Gardner1, L. Servais1, R.
O’Connor1, A.Y. Zhou1, J. Ji1
1Color Genomics, Inc., Burlingame, CA; 2Instituto Medico de Alta Tecnologia,
Monteria, Columbia; 3Sírio-Libanês Hospital, São Paulo, Brazil; 4Vida en Genoma
AC, Mexico City, Mexico; 5Laboratorio Genia, Montevideo, Uruguay; 6Genomics
Peru, Lima, Peru; 7Genda Genetica y Biologia Molecular, Buenos Aires, Argentina.
Introduction: Current hereditary cancer risk data is mostly based on genetic testing
performed in Caucasian and Ashkenazi Jewish populations. As a result, the
distribution of affected genes and their associated cancer risk in other ethnicities is
not well understood. This study aims to provide insights into genetic cancer risk
within the Hispanic population. Methods: We describe the demographics and
genetic results of 1,025 Latin American hereditary cancer high risk individuals from 6
countries (Argentina, Brazil, Colombia, Mexico, Peru, and Uruguay) who were
referred by a physician to receive the Color Test. The Color Test is a next generation
sequencing (NGS) based assessment of 30 genes associated with hereditary cancer
risk. The relatedness of individuals was not assessed in this study. Results: In this
cohort, a total of 202 pathogenic or likely pathogenic variants were identified in 189
of 1,025 individuals for a mutation carrier rate of 18.4%. The mutation carrier rate
varied by country: Brazil 21.6% (49/227), Mexico 21.3% (51/240), Uruguay 17.6%
(27/153), Colombia 17.3% (40/231), Peru 15.4% (12/78), and Argentina 10.4%
953
AMP Abstracts
(10/96). The majority of individuals in this study self-reported to be Hispanic (45.4%)
or Caucasian (24.6%). The mutation carrier rate was 21.4% in Caucasians and
19.6% in Hispanics. Almost half of this high risk cohort (46.4%) had a personal
history of breast cancer, followed by colon (6.0%), and ovarian (3.6%) cancers, while
44.6% reported no history of cancer. For individuals with a history of cancer, the
mutation carrier rate was 19.7% (breast), 37.1% (colon), and 29.7% (ovarian)
compared to 14.2% for individuals with no history of cancer. In the entire cohort,
BRCA1 and BRCA2 accounted for 51.5% of pathogenic variants identified for a
mutation carrier rate of 10.1%, similar to those previously reported for high risk
Caucasian (12.1%) and Ashkenazi Jewish cohorts (10.3%). Of tested positive
individuals, pathogenic variants were identified in BRCA1 or BRCA2 in 71.3% of
breast cancer cases, 72.7% of ovarian cancer cases, and 44.6% of individuals with
no history of cancer. Pathogenic variants in Lynch syndrome genes were found in
60.9% of positive individuals with a history of colon cancer. A total of 13 (6.9%)
individuals were found to carry 2 concurrent pathogenic variants in different genes,
indicating the importance of broader multi-gene panel testing. Conclusions: This is
one of the first studies of hereditary cancer risk genetic assessment in the Latin
American population. BRCA1 and BRCA2 accounted for the majority of genetic
alterations in individuals with a history of breast and ovarian cancer, while Lynch
syndrome genes were predominant amongst individuals with a history of colon
cancer.
G50. Comprehensive Detection of CFTR Variants Using Anchored Multiplex
PCR and Next-Generation Sequencing
M.T. Hardison1, K.E. Moore2, P.G. Roberts2, L.M. Griffin2, R.D. Walters2, B.P.
Culver2
1BabyGenes, Golden, CO; 2ArcherDX, Boulder, CO.
Introduction: Cystic Fibrosis (CF) is an autosomal recessive disease caused by
mutations in the CF transmembrane conductance regulator (CFTR) gene. Carrier
identification and newborn screening have significant implications in the overall
prognosis of CF patients. Underlying CFTR mutations were recently shown to vary
significantly across ethnic groups. However, current CFTR genotyping assays detect
mutations highly prevalent in white individuals, yet fail to detect mutations that are
more prevalent in nonwhite individuals. Furthermore, these assays also fail to detect
large deletions, such as the CFTRdele2,3(21kb), which is prevalent in Central and
Eastern European populations and confers a severe CF phenotype. Here, we
present a method based on Anchored Multiplex PCR (AMP) and next-generation
sequencing (NGS) for comprehensive, pan-ethnic detection of CFTR variants,
including common base substitutions and large deletions. Methods: AMP is a library
preparation method for NGS that uses unidirectional gene-specific primers (GSPs)
and molecular barcoded adaptors ligated to random start sites for open-ended
amplification. This enables NGS-based identification of both known and unknown
mutations across a panel of target regions. Furthermore, anchored GSPs amplify
large genomic regions from both ends independently, permitting sequencing of both
wildtype and variant alleles from the same GSPs. Results: Using a set of 150
blinded specimens from the Coriell Institute for Medical Research, we show that
AMP-based NGS detects known CFTR variants with 100% accuracy. In addition, we
detected the CFTRdele2,3(21kb) in a pre-validated DNA sample obtained from the
Coriell Institute for Medical Research. Pan-ethnic screening of 1,585 clinical samples
identified 34 unique mutations, several of which were identified in multiple
individuals. 73% (25/34) of these unique mutations and 60% (74/123) of total
mutations detected are not currently included in the ACMG-recommended 23mutation panel for CF carrier screening. Furthermore, this screen revealed ethnic
differences in clinically significant CFTR variants and a pan-ethnic carrier rate of
approximately 7%. Conclusions: We demonstrate that AMP-based NGS enables
comprehensive detection of both known and novel variants in the CFTR gene, with
the ability to detect large deletions. Identification of novel variants is critical for global
carrier and newborn screening, as CF driver mutations have not been fully
characterized across all ethnicities. As the reagents for AMP-based NGS are
lyophilized and thus do not require refrigeration, this may be a practical method for
CF screening in global communities.
Hematopathology
H01. Diffuse Large B-Cell Lymphoma Gene Expression Profiling for Cell-ofOrigin Determination (Lymph2Cx Testing) Using FFPE Tissue Sections in a
Clinical Molecular Diagnostics Laboratory
R.S. Robetorye, A.J. Wendel Spiczka, C.A. Ramsower, T.K. Yip, B.J. GlinsmannGibson, L.M. Rimsza
Mayo Clinic in Arizona, Scottsdale, AZ.
Introduction: Gene expression profiling (GEP) can classify diffuse large B-cell
lymphomas (DLBCL) into 2 groups with distinct biology and clinical behavior: the
germinal center B-cell-like (GCB) and activated B-cell-like (ABC) subtypes. Cell-oforigin (COO) assignment of DLBCL has important biological and prognostic
significance, as well as potential therapeutic implications, with the development of
selective agents for treatment of specific DLBCL subtypes in ongoing clinical trials.
Here, we describe the development of a clinical GEP assay to perform COO
assignment of DLBCL in a CAP/CLIA-certified clinical laboratory. Methods: Twenty-
954
three FFPE specimens that had previously been characterized for COO by
Affymetrix GEP on matched fresh/frozen tissue, one DLBCL cell line each from GCB
(DB) and ABC (Riva) subtypes, and 11 additional DLBCL FFPE specimens
(excisional and needle core biopsies) obtained from Mayo Clinic archives were
analyzed as a part of the assay validation. Microscopic inspection of H&E-stained
slides was performed by a pathologist to determine tumor content and area. Tumor
tissue comprising ≥60% of the surface area was macrodissected, and RNA was
extracted from 10 micron unstained FFPE sections. The resulting RNA was
quantified by spectrophotometry (NanoDrop, Thermo Scientific), and 400ng total
RNA was hybridized overnight on a thermal cycler to the 20 fluorescently-labeled
gene probes in the Lymph2Cx panel. The 20 gene panel includes 8 genes
overexpressed in ABC, 7 genes overexpressed in GCB, and 5 housekeeping genes.
Probe/RNA complexes were purified on a NanoString nCounter Prep Station and
then transferred to a NanoString nCounter Digital Analyzer for quantification via
single molecule imaging. The counts were processed using the National Cancer
Institute’s Lymphoma/Leukemia Molecular Profiling Project Lymph2Cx DLBCL COO
Classifier (patented algorithm) for quality control and subtyping. Minimum total
laboratory time required for the assay is approximately 28 hours.
Results: Performance characteristics for the NanoString nCounter system and the
Lymph2Cx COO GEP assay were established for accuracy, precision, analytical
sensitivity, analytical specificity, and specimen stability. Conclusions: We have
established the first GEP assay available in the United States for COO assignment
of DLBCLs in a CAP/CLIA-certified molecular diagnostics laboratory. FFPE tissue
samples from a variety of sources and biopsy types, including excisional and needle
core biopsies of lymph nodes and soft tissues, yield sufficient RNA for COO
assignment of DLBCLs. The Lymph2Cx COO GEP assay can be performed
relatively rapidly, is robust and reproducible, and can provide valuable biological,
prognostic, and potential therapeutic information for DLBCL patients.
H02. Performance Evaluation of a T-cell Receptor Gamma Gene
Rearrangement (TRG) Next Generation Sequencing (NGS) Assay for Clinical
Practice
V. Borodin, C. Yang, V.S. Williamson, O.C. Rafael-Rosca, A. Popa, R. Ren, F.
Sábato, A. Ferreira-Gonzalez
Virginia Commonwealth University Health System, Richmond, VA.
Introduction: TCR rearrangement is an important event in T-cell ontogeny and can
serve as the marker of hematoproliferative disease of the T-cell origin. Using
techniques such as PCR, it has been possible to characterize T-cell proliferations in
malignancy and in diseases where T-cells have been implicated in the pathogenesis.
The current gold standard PCR-based assays identify clonality on the basis of overrepresentation of amplified V-J or D-J products. T-cell clonality testing using NGS
represents a major advance compared to PCR as it can provide specific clonal
sequences that can be tracked along the course of disease. We evaluated the
performance of the LymphoTrack T-cell Receptor Gamma (TRG) targeted NGS
assay that sequences flanking fragments of the V - J region with unique sequence
and length corresponding to a single productive V - J rearrangement in each Tcell. Methods: The LymphoTrack TRG assay was performed according to
manufacturer’s recommendations on the Ion Torrent S5XL using 520 chip. Limit of
detection (LOD) and precision of the LymphoTrack TRG assay was evaluated by
testing a serial dilution of DNA isolated from FFPE Jurkat cell line in normal FFPE
DNA tonsil tissue at concentrations from 20% down to 1.25%. Intra and inter-assay
variation were evaluated by sequencing Jurkat cell line mixtures over 5 different
runs, using different barcodes within the same chip and on different chips on
different days. Assay correlation was determined by testing DNA from 40 patient
clinical samples characterized by our laboratory using Invivoscribe PCR and CE
fragment analysis (21 monoclonal, 13 polyclonal, 3 oligoclonal and 3 monoclonal
only in TRB gene region). Results: The LymphoTrack TRG showed good LOD down
to 2.5% dilution. Repeatability and reproducibility were very good with values of
100%. Correlation studies showed a sensitivity of 70.8% and a specificity of
87.5%. In order to improve these values we developed our own algorithm that
improved sensitivity to 87.5% and same specificity of 87.5%. Conclusions: Our
study of the LymphoTrack TCG NGS assay showed good performance
characteristics and it could be a good option for the clinical laboratory.
H03. Minimal Recipient Chimerism Detection by qPCR Method for the PostTransplant Patients Who Achieved Complete Donor Chimerism by STR
Method
L. Kumer, J. Tyler, C. Fisher, H. Shike
Penn State Hershey Medical, Hershey, PA.
Introduction: Insertion/deletion (ins/del) markers by quantitative real-time PCR
(qPCR) provides increased sensitivity for the post-stem cell transplant chimerism
monitoring compared to the short tandem repeat (STR) method. We utilized qPCR
method for the patients who were assigned to have complete donor chimerism by
STR method. Clinical significance of detecting minimal recipient chimerism under
STR sensitivity has been reported since increased recipient chimerism predicts risk
for relapse that can be prevented by timely immunotherapy (tapering of
immunosuppression or donor lymphocyte infusion).
Methods: Post-transplant donors and recipients were genotyped for 22 STR
markers (PowerPlex Fusion STR kit, Promega) and 29 ins/del markers (KMRType,
GenDx). Average % Donor was calculated from 5 informative STR markers or 2
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
informative qPCR markers (KMRTrack, GenDx) in post-transplant total cell fraction
and T-cell fraction. T-cell was selected either by negative selection (RosetteSep
Human T-Cell Enrichment Cocktail, Stem Cell Technologies) or positive selection
(Whole Blood CD3 MicroBeads, Miltenyi Biotec). Post-transplant DNA 0.2ng and
14ng per reaction were used for STR and qPCR, respectively. Validated sensitivity
was 5% for STR and 0.5% for qPCR. However, although not reproducibly, less than
0.5% recipient can be detected by qPCR which was assigned as residual %recipient
because recipient allele always resulted 0% in donor only DNA samples (100%
specificity). Average 95% confidence interval was 1.8% for STR (5-95% Recipient)
and 1.0% for qPCR (range: 0.5-10% Recipient). Patients who achieved stable
complete donor chimerism (>95% Donor) were tested by qPCR
method. Results: Post-transplant patients (N=14) with stable complete donor
chimerism (>95% Donor) for both total cell fraction and T-cell fraction were tested by
qPCR. Residual recipient alleles in the total and T-cell fraction were detected in 7
patients (50%) 9 patients (64%), respectively. Residual % recipient (Average ±SD)
was 0.56% ±0.77% (range: 0.04-2.02%, N=7) for Total cell fraction; 1.21% ± 1.12%
(range: 0.08-3.08%, N=9) for T-cell fraction. Conclusions: qPCR method improves
minimal recipient chimerism for the patients with stable complete donor chimerism by
STR method.
H04. Frequency and Pattern of BCR-ABL Kinase Domain Mutation in Chronic
Myeloid Leukemia-An Indian Perspective
R. Katara1, V. Kumar1, P. Chauhan1, J. Sharma1, A. Verma1, L. Kini1, S. Sharma1,
S. Verma1, J. Mukherjee3, A. Pandita2
1CORE Diagnostics, Gurgaon, Haryana, India; 2CORE Diagnostics, Palo Alto,
CA; 3University of San Francisco, San Francisco, CA.
Introduction: BCR-ABL kinase domain (KD) mutations play a significant role in
acquiring resistance to imatinib (IM) treatment and have been found to be associated
with progression and poor prognosis in chronic myeloid leukemia (CML) patients.
KD-mutated CML patients exhibit a poor response to IM therapy resulting in adverse
outcomes. The intensity of IM resistance is associated with the localization of various
mutations in kinase domain of the ABL1 gene. The objective of the current study is to
investigate the incidence and patterns of KD mutations in an Indian cohort of
previously diagnosed CML patients that relapsed on Imatinib treatment.
Methods: The study includes peripheral blood or bone marrow samples of
previously diagnosed and followed-up CML patients, for KD mutation testing
collected from different region of India. Total RNA from these samples was extracted
and analyzed to investigate the mutational status of Tyrosine Kinase domain (TKD)
using Polymerase Chain Reaction followed by Sanger sequencing based assay. A
special 2-step in-house assay was developed for this study. The first step included
the amplification of the BCR-ABL fusion transcript followed by a second PCR
reaction to investigate the ABL1 KD region for any mutations present. Results: A
total of 653 CML patient samples were screened for ABL1 KD gene mutations. 244
(37%) of the cases presented with KD gene mutation. The male to female ratio was
2:1 and the age range was 10 years to 85 years. Thirty-eight different types of
mutations were detected. T315I was the most common KD mutation observed in
13.5% cases, G255K in 8.6%, E255K in 5.7%, F317L in 5.3%, and F359V in 4.9%.
Some rare mutations, L298V and L273M were also observed in one and 2 patients
respectively. 7.7% cases exhibited more than 1 mutation. Majority of the cases that
were positive for KD mutation were in 40-60 years (45%), age group.
Conclusions: This is second largest study to report incidence of the KD mutations in
Indian population. This study also reveals a vast diversity of the KD mutation pattern
in Indian CML patients. T315I mutation was the most frequent mutation found in the
KD gene. The index study highlights the importance of KD mutation testing in
predicting treatment and outcome in CML patients.
H05. Clinical Validation of a Highly Sensitive and Highly Reproducible BCRABL1 Quantification Assay for CML Monitoring
M. Alikhan, P. Lu, Y. Wang
University of Chicago, Chicago, IL.
Introduction: Accurate and robust monitoring of BCR-ABL1 fusion transcripts (FTs)
is now an essential component in caring for patients with chronic myeloid leukemia
(CML). Clinicians rely on results of the quantitative polymerase chain reaction
(qPCR) assay to guide therapeutic decision-making in assessment of response to
tyrosine kinase inhibitors (TKIs), detection of minimal residual disease, and
monitoring for relapse. Robust drop in BCR-ABL, down to at least MR4.0 (4 logreduction, International Scale of 0.01%) is an important therapeutic milestone which
predicts good long-term survival. If there are signs of molecular relapse, patients
may be placed on alternative therapies. For this reason, an accurate, precise, and
standardized assay is required to meet the needs of CML patients and clinicians.
Herein, we outline the validation of QuantideX, an FDA-approved qPCR assay that
reliably detects BCR-ABL, even at levels corresponding to “deep molecular
response”. Methods: RNA was extracted from patient peripheral blood or non-CML
leukemic cells lines. Samples were subjected to qPCR following manufacturer’s
protocol. Results were analyzed using the QuantideX software included in the
package, which yields raw ratios between BCR-ABL1 and control gene (ABL1)
transcripts, Ct values, and the IS. Results: Six non-CML leukemic samples tested
negative for BCR-ABL1 transcripts, yielding an analytic specificity of 100%. The limit
of detection of the assay was determined using 4 contrived, pre-diluted specimens
with BCR-ABL1 values ranging from MR4.5 to MR4.9 at 0.1 intervals (IS<0.01%).
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
Four-six replicate reactions were conducted and 26 of 28 determinations yielded
positive results. Assay linearity was evaluated with a set of cell line dilutions ranging
from IS 0.01%-10% (MR4.0-MR1.0) in quadruplicate over 3 runs. We found that the
output (IS %) was linear across 4 logs of dilution between 0.01%-10% IS. Inter- and
intra-run reproducibility was evaluated using 4 samples derived from a cell line and
diluted to different IS levels. These were tested in quadruplicate (intra-run
reproducibility) over 3 independent runs (inter-run reproducibility). There were 12
distinct data points for each sample. Results showed high precision and
reproducibility within the reportable range. The assay was validated against another
clinical grade RQ-PCR assay. Concordance was acceptable, with Pearson
coefficient of about 0.9 and R2 of 0.81 (p value < 0.0001 with 95% confidence
interval between 84-94%). Conclusions: In this study, we verified that QuantideX,
an FDA-approved qPCR assay, reliably detects BCR-ABL, with high specificity,
sensitivity, linearity and intra and inter-assay reproducibility. With its high sensitivity,
the assay is amenable to measure “deep molecular response”.
H06. Comparison of Clonality Testing on B Plus Fixed Versus Formalin Fixed
Tissue
E. Castro-Echeverry, C.L. Burnes, S. Roy, M. Nikiforova, N. Aggarwal
University of Pittsburgh School of Medicine and University of Pittsburgh Medical
Center, Pittsburgh, PA.
Introduction: Zinc based fixatives provide excellent nuclear detail and antigen
preservation. However their use in molecular assays is discouraged as heavy metals
are thought to inhibit PCR. We compared B (BCR) and T-cell receptor (TCR) gene
rearrangement studies performed on B plus fixed paraffin embedded (BFPE) tissue
(a Zinc based fixative) to formalin fixed paraffin embedded tissue (FFPE).
Methods: We reviewed tissue biopsies with BCR and/or TCR gene rearrangement
studies on FFPE and performed these studies on comparable BFPE tissue block.
Tissue was microdissected from slides and DNA was extracted using the DNeasy
Blood and Tissue Kit (Qiagen). BIOMED-2 primer sets for TCRB, TCRG, IgH,
IgKappa, and Kappa deleting element (InVivoScribe) were used. PCR products were
analyzed by capillary electrophoresis using ABI-3130. Comparing FFPE and BFPE,
results were classified as fully concordant if all peaks were concordant to within
2bp. Results were classified as somewhat concordant if peaks differed by >2bp but
the overall interpretation was similar. A major discordance was defined as an overall
interpretation that is different. Overall concordance includes both fully concordant
and somewhat concordant cases. Results: Sixteen FFPE and BFPE cases were
evaluated for B and/or T-cell clonality. The purity of isolated DNA in BFPE was
similar to that obtained in FFPE, with an average 260/280 absorbance of 1.90 (1.7 to
1.98) for BFPE and 1.91 (1.84 to 1.99) for FFPE. Both DNA quality (as measured by
the size ladder) and the quantity of amplification products (as measured by
fluorescent intensity) were lower for BFPE (92% ≤300bp by size ladder, 2,301
average maximum fluorescence) compared to FFPE (75% ≥400bp by size ladder,
3,394 average maximum fluorescence, p<0.05 by paired t-test). The overall
concordance was 100% for T-cell clonality studies (10/10) and 86% (12/14) for B cell
clonality studies. Minor discordances were attributed to the lower intensity of the
peaks, with smaller peaks on FFPE being difficult to appreciate on BFPE. Two major
discordances included a negative case by BFPE that was positive by FFPE (false
negative), and a suspicious case by BFPE that was negative by FFPE (false
positive). The false negative case is attributable to a suboptimal sample of BFPE
tissue with a limited amount of DNA. The false positive case involved a diagnostic
challenge in a patient with recurrent lymphoma, raising the possibility of a persistent
B cell clone. Conclusions: FFPE tissue is still preferable over BFPE tissue due to
the better quality of DNA isolated and greater depth of experience with this fixative.
However, BFPE would be an acceptable source for molecular testing if this is the
only available tissue, provided one bears in mind the caveats of low amplification.
H07. Detection of Fusion Transcripts in Hematologic Malignancies by RNASeq
P. Szankasi1, C.N. Paxton1, W. Shen1, P.M. Rindler1, B. O'Fallon1, J.A.
Schumacher1, X. Xu2, T.W. Kelley2
1ARUP Laboratories, Salt Lake City, UT; 2University of Utah, Salt Lake City, UT.
Introduction: Detection of translocations is an important component of the workup
of hematologic malignancies. The testing provides diagnostic and prognostic
information and is critical for patient management. Translocations are typically
detected by a combination of classical cytogenetics, fluorescent in situ hybridization,
and quantitative reverse transcription (qRT)-PCR tests. Next generation sequencing
(NGS) is emerging as a method for the detection of not only point mutations but also
structural rearrangements. The latter are detected by sequencing across fusion
breakpoints. Here we describe a sensitive method for the detection of abnormal
RNA fusion transcripts resulting from translocations from both fresh and formalinfixed tissues using solution capture and next generation sequencing. Methods: Total
RNA was prepared from whole blood and bone marrow aspirate and formalin-fixed
paraffin-embedded (FFPE) tissues. Double-stranded cDNA was generated and
converted into NGS libraries using the Kapa RNA Hyper Prep Kit (Kapa
Biosystems). Solution capture of one or both fusion partner sequences was
performed using biotinylated probes (Integrated DNA Technologies) and sequenced
on the NextSeq instrument (Illumina). Translocations were detected using the
Fusion Catcher program. All samples harbored translocations previously detected by
qRT-PCR (copy number normalized to the ABL1 mRNA) or NGS of genomic DNA.
955
AMP Abstracts
Results: For an initial study, we tested a total of 17 samples with previously
detected translocations. We detected 3/3 inv(16); CBFB-MYH11 types A, D, E
(normalized copy number, NCN: 0.048 to 1.036) and 3/3 t(15;17); PML-RARA types
bcr1/2, bcr3 (NCN: 0.0082 to 0.3976). We consistently detected the t(9;22); BCRABL1 p210 in 4 samples with values of 2% to 10.6% on the international scale, IS
(NCN; 0.0175 to 0.1228) but not in 2 samples with values of 0.45% - 0.59% on the IS
(NCN: 0.0054 - 0.0067). We did not detect 2 low-level t(9;22) p190 (NCN: 0.0011 to
0.0014) and a low-level t(8;21); RUNX1-RUNX1T1 (NCN: 0.0054). We detected 2/2
rearrangements in FFPE samples, a t(11;18) MALT1-BIRC3 and a t(3;16) 5’-UTR
fusion between BCL6 and one of its recurrent non-immunoglobulin partners, CIITA,
previously detected by NGS of genomic DNA. The BCL6 and RARA fusions were
detected without capturing the partner gene. Conclusions: We demonstrate a
sensitive method for the detection of abnormal fusion transcripts in hematologic
malignancies by NGS. The method works with both fresh and archival tissues. The
ability to detect translocations by capturing just one of the partners allows the
comprehensive interrogation of genes with multiple and unknown fusion partners.
H08. Development and Validation of a Multiplex Droplet Digital PCR Assay for
the Detection and Quantification of BCR/ABL1 Fusion Transcripts
R.Y. Walder1, A.A. Stence1, M.R. Nasr2, N.V. Guseva1, D. Ma3, A.N. Snow1, A.D.
Bossler1
1University of Iowa Hospitals and Clinics, Iowa City, IA; 2University of Manitoba,
Winnipeg Manitoba, Canada; 3University of Iowa, Iowa City, IA.
Introduction: Droplet digital PCR (ddPCR) enables the detection and absolute
quantification of DNA and RNA targets. Here, we assessed the performance of a
laboratory-developed ddPCR assay for detection and quantification of BCR/ABL1
fusion transcripts in comparison with known WHO standards. Primers and probes for
the major (e13a2, e14a2) and minor (e1a2) breakpoint products were multiplexed
and tested. Methods: Total RNA samples were extracted with the QIAamp RNA
blood kit. The presence of BCR/ABL1 fusions was measured with the ddPCR system
(QX200, Bio-Rad). We validated the multiplex assay with RNA from cell lines: K562
(e14a2 fusion) and SUP-B15 (e1a2 fusion). Mutant RNAs were serially diluted into
ABL1 normal RNA from the HL60 cell line. RNA was reverse-transcribed into cDNA
with iScript Advanced reverse transcriptase (Bio-Rad). The input cDNA was
emulsified into ~20,000 droplets per well, amplified by PCR with primers and specific
FAM and HEX-labelled fluorescent hydrolysis probes and then analyzed by flow
cytometry. Data was analyzed using QuantaSoft Analysis software. Results: In this
multiplex ddPCR assay we simultaneously measured the major and minor fusion
transcripts and normal ABL1. The assay was linear between 0.005% and 50%
BCR/ABL1 to ABL1 transcript ratios. Serial dilution of mutant RNAs in HL60 RNA
showed reproducible detection down to 0.006% though variability was highest at the
low quantitative values. The false positive rate with replicate BCR/ABL1 negative
samples (ABL1 positive) was 0.0002% (0.14 of 53908.07 droplets (N=23 wells)). The
linearity of the analytical measurement range was assessed (Asuragen ARQ IS
calibrator panel) and demonstrated good correlation (R2 = 0.9955, m = 1.0275 and y
intercept = -0.0893). Conclusions: Absolute quantitation of BCR/ABL1 fusion
transcripts performed using ddPCR obviates the need for normalization of cDNAs to
a standard curve. Quantitation was linear, had a lower LOD of 0.006%, and was
highly reproducible across the range of the assay. This method is reliable for
monitoring treatment responses in clinical samples.
H09. Evaluation of the QIAGEN CALR RGQ PCR Kit for the Detection of CALR
Mutations in Suspected Myeloproliferative Neoplasms
L.J. Doyle, G.W. Procop, J.R. Cook
Cleveland Clinic, Cleveland, OH.
Introduction: CALR exon 9 mutations are found in 70-84% of essential
thrombocytosis and primary myelofibrosis lacking JAK2 V617F
and MPL mutations. Although more than 50 CALR mutations have been identified to
date, the type 1 (52 bp deletion) and type 2 (5 bp insertion) mutations account for 8090% of all CALR mutations. The CALR RGQ PCR kit (QIAGEN, Germantown,
MD) employs 7 PCR reactions designed to detect type 1 and type 2 mutations as
well as additional minor variants. In this study, we compared results of the CALR
RGQ kit to a laboratory developed CALR exon 9 fragment length PCR assay in 146
samples. Methods: One hundred forty-six samples (144 blood, 2 bone marrow)
submitted to the Cleveland Clinic Molecular Diagnostics laboratory for CALR testing
were analyzed by fragment length PCR using the ABI 3730 Genetic Analyzer
(Applied Biosystems, Foster City, CA). Aliquots from the same samples were
extracted on the QIAsymphony SP (QIAGEN). DNA concentrations were determined
with the NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Wilmington,
DE) and aliquots diluted to 10 ng/µl in TE buffer. The CALR RGQ PCR assay was
performed using the Rotor-Gene Q instrument (QIAGEN). All assays were analyzed
for validity and positivity per the manufacturer’s protocol. Results: By fragment
length PCR, CALR exon 9 mutations were found in 23 cases (14 type 1, 5 type 2,
and 4 other variants). The RGQ assay yielded positive results in 20 cases including
14/14 (100%) type 1, 4/5 (80%) type 2, and 2/4 (50%) other variants. The finding of
an apparent type 2 mutation by fragment length PCR with negative RGQ result
raised the possibility that this mutation represented an unusual variant with a similar
sized insertion rather than a true type 2 mutation. Additional testing is in progress to
definitively characterize the mutation in this case. Each of the 123 samples negative
for CALR exon 9 mutations by fragment length PCR were also negative by the CALR
956
RGQ assay. Using fragment length PCR as a gold standard, the RGQ assay
detected CALR mutations with a sensitivity of 87% and a specificity of 100%.
Conclusions: The QIAGEN CALR RGQ PCR kit allows for efficient detection
of CALR mutations with a high sensitivity and specificity, although additional
complementary methods will be required to rule out rare variants not targeted by this
assay. The use of a PCR based approach allows for potential harmonization with
allele specific PCR assays for JAK2 V617F, thereby streamlining laboratory
workflow.
H10. Comparison of FLT3-ITD Allelic Ratio by PCR Analysis and Next
Generation Sequencing
E. Castro-Echeverry, L. Kelly, C.L. Burnes, S. Roy, M. Nikiforova, N. Aggarwal
University of Pittsburgh School of Medicine and University of Pittsburgh Medical
Center, Pittsburgh, PA.
Introduction: In AML patients, a FLT3-ITD allelic ratio (AR) greater than or equal to
0.51 is known to confer a worse overall prognosis, and greater benefit from
allogeneic hematopoietic stem cell transplantation. Most publications evaluating its
clinical impact used traditional, PCR based sequence analysis to calculate the AR.
Considering the increased prevalence of Next-Generation Sequencing (NGS) assays
to detect FLT3-ITD using specialized algorithms, a formal evaluation of the feasibility
of calculating the AR by NGS is warranted. Methods: Cases of AML which tested
positive for FLT3-ITD mutations by PCR and NGS were selected. DNA for peripheral
blood/ bone marrow was extracted using an automated DNA extraction device
(MagNaPure LC). For the PCR assay, juxtamembrane regions of FLT3 were
amplified with fluorescently labeled primers and evaluated by capillary
electrophoresis using the ABI-3130 genetic analyzer and GeneMapper software (v
5.0). The AR was calculated as the ratio of mutant to wild type alleles. For the NGS
assay, samples were run using a commercially available targeted AML panel on the
Illumina Miseq platform. The AR was calculated by converting the mutant allelic
quency to an allelic ratio. Results: Seven cases of AML were positive for FLT3-ITD
mutations and were evaluated by both PCR based sequence analysis and NGS. The
median difference in ARs between the 2 methods was 0.15 (range 0.03-98). The
largest difference in AR involved a 97-99bp insert with an allelic frequency of 99% by
NGS and an AR of 1.13 by PCR. With the exception of a case involving a 110bp
insert, which showed a 40bp difference between PCR (110bp) and NGS (70bp),
there was excellent correlation in insert sizes (R² = 0.99), with a median difference in
insert size of 2bp (range 0-40bp). The magnitude of the difference in AR did not
correlate with the size of the insert. Upon stratification of AR by a cutoff of 0.51, 6/7
cases were concordant, with the single discordance involving a FLT3-ITD with 2
different insert sizes (50bp, 60bp), and an AR of 1.35 by PCR and 0.41 by
NGS. Conclusions: Due to the complexity of the duplicated/inserted sequence,
accurate quantification of FLT3-ITD mutation burden and size in AML by NGS will
require rigorous validation and standardization.
H11. Differential Mutation Patterns of the Calreticulin Gene in 14,064 Patients:
Distribution of Deletions, and Insertions in a Clinical Population
J. Sebastian1, L. Cai1, V. Bibawy1, S. Perryedegeare1, J. Riojas1, P. Chan2, L. KamMorgan1, M. Eisenberg1, A. Chenn1
1Laboratory Corporation of America, RTP, Burlington, NC; 2Labcorp, Phoenix, AZ.
Introduction: Somatic mutations are often associated with specific
myeloproliferative neoplasms (MPN), a group of clonal hematopoietic stem-cell
disorders. Mutations of Janus Kinase 2 (JAK2) are the most common, found in about
50 % of all MPN cases. In MPN patients without JAK2 mutations, mutations
of MPL are detected in about 5%. Recently, mutations of the Calreticulin (CALR)
gene have been identified in a subgroup of MPN patients, and do not co-occur
with JAK2 and MPL mutations. MPN associated mutations of CALR occur primarily
in exon 9 of the gene and include insertions or deletions ranging from 1 to 52bp. The
majority of CALR mutations involve deletion (Type 1: 52 bp deletion) or insertion
(Type 2: 5bp insertion) mutations at exon 9. To this end, we
assessed CALR mutations of a clinical population of 14,064 suspected MPN
patients. Methods: To assess mutations of the CALR gene for specimens submitted
for CALR testing in a patient population (n=14,064), genomic DNA was subjected to
a size-specific polymerase chain reaction (PCR) analysis and tagged with a
fluorescent dye. Next, the fluorescent amplification products were separated by size
using capillary electrophoresis on an ABI 3130 Genetic Analyzer instrument to
identify the presence of different variant sequences. The wild type allele was
designated by a single blue peak at the predicted product size, 283bp. Mutations
were identified by the presence of an additional peak at > or < 283bp indicating an
insertion or deletion, respectively. Peaks greater than 200 in height were used to
determine the presence of deletion or insertion. Results: This study identified a total
of 683 mutations in the CALR gene exon 9, representing 4.85% of the total
(n=14,064) sequences analyzed. The majority of these (n=555, 81.2%) were type 1
or type 2 mutational changes. Of the 683 variants identified 322 (47.1%) were (type
1), 233 (34.1%) were type 2. The remainder (n=126, 18.8%) were alternate insertion
or deletion mutations ranging from 2 bp to 43 bp. We also identified 2 individuals with
both 8bp deletion and 5bp insertion in CALR gene. Conclusions: This study
examines both deletion and insertions in exon 9 of the CALR gene in JAK-2 V617F
negative individuals suspected of having myeloproliferative disorders. We identified
683 individuals, in a population of 14,064 MPN patients, carrying indels in exon 9 of
the CALR gene. The majority of individuals had 52bp deletions (47.1%) or 5bp
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
insertions (34.1%). We also identified a series of insertions and deletions (2bp to
43bp) (18.8%) in the remaining study population, as well as 2 individuals with both
an exon 9 CALR gene insertion and deletion. Further studies are needed to assess
the impact of these mutations in treatment outcome and severity of MPNs.
H12. Impact of MYC Abnormalities, Trisomy of Chromosome 8 and Estimated
Tumor Progression Values in Plasma Cell Myeloma
R. Garcia, M. Flores, W. Chen, P.R. Koduru
University of Texas Southwestern Medical Center, Dallas, TX.
Introduction: Plasma cell myeloma (PCM) is a heterogeneous neoplasm
characterized by a distinct genetic abnormalities and patient
outcome. MYC overexpression resulting from MYC rearrangements (R) involving
immunoglobulin (IG) and non-IG genes may be related to shorter survival.
Unlike MYC R, trisomy of chromosome 8, may be an early event that results in
aggressive transformation. These MYC related abnormalities have not been fully
characterized in terms of their relationship with additional chromosome changes and
impact on clinical outcome. In this study, we explore these relationships. Methods:
Patients with a diagnosis of PCM from 2005-2016 were retrospectively analyzed
for MYC related as well as additional aberrations. Four groups were
identified: MYC R/IG, MYC R/non-IG, Trisomy 8 without MYC R and a control group
with no MYC related abnormalities. To outline evolution patterns karyotype data
were parsed using the tm library in R. The Translational Oncology package was then
used to construct tumor evolution patterns, while the Rtreemix library was used to
establish disease progression by computing a genetic progression score (GPS).
Kaplan-Meir survival curve and Log-Rank Test was plotted to assess overall survival
(OS). Results: A total of 66 PCM cases were identified. Of these, 16 had MYC/IG, 7
had MYC/non-IG,18 had trisomy 8 and 28 were controls without MYC abnormalities.
Text analysis identified 44 recurrent cytogenetic aberrations (RCAs). All 44 RCAs
were applied to an evolution pattern analysis, but only 20 were used to estimate
GPS. There were 3 major patterns: +9, 1p loss, and +8. MYC R appeared later in
tumor evolution, while +8 was an early event. Twenty sub-clones were
identified: MYC/IG group had 3 clones (+9,+15,+5,t(8;14)[2]; +9,+15,+5,t(8;22)[7];
+9,+15,+5,t(2;8)[4]); MYC/ non-IG had 1 clone (1p loss, 1q gain, t(1;8)[3]); trisomy 8
had 1 clone (+8,+20,+16[9]) and the control group had the remaining 15 clones. The
mean GPS for the groups were 1.5 for MYC/IG, 0.9 for +8, 0.7 for MYC/non-IG and
0.5 for the control group. In terms of OS, the combined MYC R group had dismal
outcome (p =.007), while +8 showed an inferior outcome (p =.02). When OS was
compared among all MYC related abnormalities, MYC/non-IG had worst outcome
followed by trisomy 8 and MYC/IG groups. In terms of the relationship between
genetic aberrations, 1p loss may herald a MYC/non-IG R, while presence of the odd
number chromosomes 5, 9 and 15 may anticipate a MYC/IG R. Conclusion: Our
analysis suggests MYC related aberrations, including trisomy 8 and increased GPS
associated with poor outcome. Moreover, unique RCAs may be useful in recognizing
early genetic events that may signal a MYC R and thus the need to monitor these
patients more closely.
H13. One Children’s Oncology Group Cytogenetics Laboratories’ Experience
With Single Nucleotide Polymorphism Chromosome Microarray Analysis of
Pediatric Acute Leukemia’s
M. Micale1,2, B. Embrey1 2, A. Knaus1
1Beaumont Health, Royal Oak, MI; 2Oakland University William Beaumont School of
Medicine, Rochester, MI.
Introduction: The 5-year survival rate for children with acute lymphoblastic
leukemia (ALL) and acute myelogenous leukemia (AML) has increased over time,
now at more than 85% for ALL and 60%-70% for AML. These improvements are
largely due to the ability to correlate morphological, immunophenotypic,
cytogenetics, and outcome data to risk stratify consolidation chemotherapy regimens
in these children, an effort led by the Children’s Oncology Group
(COG). Cytogenetic studies required of every child on ALL COG protocol AALL08B1
or AML COG protocol AAML1031 includes karyotyping, a panel of FISH probes to
identify abnormalities specific to AML or ALL, and additional reflex FISH testing if
these studies yield normal results. Current COG protocols do not require evaluation
with single nucleotide polymorphism (SNP) chromosome microarray, as the
technology is still relatively new compared with traditional karyotyping and FISH, and
is not available for oncology studies in all centers. Nevertheless, the technique holds
great promise for identifying additional important genomic
biomarkers. Methods: Newly-diagnosed ALL (5 cases) and AML (3 cases) on COG
study at Beaumont Health were analyzed per routine protocol by karyotyping and
FISH, and additionally by SNP array, and results were
compared. Results: Compared with traditional karyotyping and FISH, SNP array
identified additional abnormalities in 87.5% of cases, including: 1) an ERG gene
deletion (associated with a good prognosis) and PAX5 deletion (associated with high
relapse risk) in an ALL case that was normal by karyotype and FISH, 2) a CRLF2P2RY8 deletion (associated with lower event-free survival) in a Down syndrome child
with ALL that demonstrated no acquired abnormalities, 3) another CRLF2P2RY8 deletion along with homozygous CDKN2A/B deletion (associated with poor
outcome) in an ALL case with an abnormal karyotype demonstrating del(9p) and
i(17q) abnormalities only, 4) chromosome 15 uniparental disomy in a hyperdiploid
ALL case, 5) an RB1 gene deletion in an ALL case with ETV/RUNX1 rearrangement
by FISH, and 6) a 40 Mb telomeric region of copy neutral loss of heterozygosity on
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
chromosome 4q31.23 (containing the cancer genes NPY1R, FBXW7, VEGFC,
and FAT1 implicated in myeloid neoplasms) in an AML case with normal karyotype
and FISH. Conclusions: This study demonstrates the importance of performing
SNP array on newly-diagnosed pediatric acute leukemia’s, as such analysis provides
additional diagnostic, prognostic, and therapeutic information. We believe that
performing SNP array on every pediatric acute leukemia and incorporating such
results into COG risk-stratification algorithms will further improve the care of these
children.
H14. Use of an NGS Based Custom Myeloid Gene Panel for Sequencing of
Formalin-Fixed Paraffin Embedded Bone Marrow Clot Sections and Air-Dried
Smears in Acute Myeloid Leukemia
A.N. Huho1, J. Gale2, J. Yan3, M. Vasef3, D.R. Czuchlewski3
1University of New Mexico- TriCore Labs, Albuquerque, NM; 2TriCore Reference
Laboratories, Albuquerque, NM; 3University of New Mexico, Albuquerque, NM.
Introduction: Clinical next-generation sequencing (NGS) panels have an
expanding clinical utility in acute myeloid leukemia (AML). Fresh peripheral blood or
bone marrow specimens are optimal for clinical testing, but may not always be
available and in extramedullary disease, diagnosis may be solely tissue based.
The ability to perform myeloid gene panel sequencing on secondary specimen
types like formalin fixed paraffin embedded (FFPE) is desirable.To our knowledge,
scant literature exists on the use of FFPE-clot sections and or smears for NGS. We
investigated the validity of next-generation sequencing of DNA extracted from
archived FFPE- clot sections and smear scrapings in acute myelogenous leukemia
(AML) using a targeted custom Ampliseq based myeloid gene panel (Lehmann U. et
al., 2016) on the Ion Personal Genome Machine (PGM) with concurrent sequencing
of archived, previously fresh extracted DNA for comparison. Methods: Three AML
cases with at least 1 clinically reportable mutation found on previous sequencing on
an Illumina platform were identified. For each case, archived previously extracted
fresh DNA, unstained smears and FFPE clot sections were retrieved. H&E slides
corresponding to FFPE clot sections were reviewed to confirm the presence of
marrow 60um tissue section were trimmed from the FFPE blocks,followed by
deparaffinization and overnight lysis, the smears (6 slides) were scrapped. Both
specimens had DNA extracted using the Promega-Maxwell DNA purification Kit.
The triplicate samples (Previously extracted Freshbone marrow DNA and DNA
extracts from air-dried smear and FFPE clots) were quantified by fluorometry using
the Quantas t 2.0 followed by targeted amplification, library preparation, and
sequencing on an Ion PGM. Results: Adequate DNA was extracted from all FFPE
clot samples (average 180 ng/µl) and smear scrapings (average 132 ng/ul) with
FFPE samples yielding less DNA from the correspondingly paucicellular H&E
sections. All mutations (above a variant allele frequency of >5%) previously identified
in fresh bone marrow samples on Illumina platform were detected in the FFPEc,
scraped air-dried smears and BM extracted DNA when repeated on our Ion PGM.
Conclusions: FFPE clot sections or airdried smears with adequate cellularity may
be used for next generation sequencing for myeloid neoplasms. Using this custom
NGS panel allows for additional specimen types (FFPE clot and/or unstained airdried smears) when fresh bone marrow or peripheral blood is unavailable. This
method may also be useful when extramedullary myeloid disease is diagnosed and
only FFPE tissue sections are available. Further validation on additional samples will
allow for the clinical use of this method.
H15. A Prolonged Low Level JAK2 V617F Is Significant In Clinically Suspicious
Myeloproliferative Neoplasms (MPN)
E. Vail, M. Chevarie-Davis, C. Riley, J. Lopategui
Cedars-Sinai Medical Center, Los Angeles, CA.
Introduction: The association between JAK2 V617F mutations and MPN is well
known. Amongst the available detection methods, real time allele-specific
polymerase chain reaction (PCR) reaches the highest sensitivity, detecting mutation
levels down to 0.01%. Although it has been established that levels above 1% are
clinically significant and highly suggestive of MPN, it is less clear what the relevance
of a low level positive (LL-JAK2) may be. Previous reports have shown that LL-JAK2
mutations could be detected in a rare subset of healthy individuals, whereas others
suggested that these individuals are in fact at higher risk of eventual progression to
MPN. Our goal was to study the clinical outcome of those patients with LL-JAK2, in
order to determine their optimal clinical management. Methods: All JAK2 V617F
mutation tests with low level mutation values, defined as ranging between 0.1% and
1.0%, were retrieved from the tests performed by PCR in our lab over a 6 yearperiod. Prolonged LL-JAK2 was defined as extending over 3 months. A complete
chart review was conducted to ascertain clinical history, indication for JAK2 testing,
laboratory and clinical follow-up. Negative and positive JAK2 mutation tests were
selected at random for the purpose of comparison, and the charts were reviewed in a
similar fashion. Results: A total of 2,325 JAK2 tests were performed by PCR, of
which 27 (1.2%) fell into LL-JAK2 category. These 27 tests, corresponding to 25
patients, were comprised of 5 bone marrow samples and 22 peripheral blood
samples. Of the 23 patients with available clinical information, 9 (39%) were
diagnosed with MPN, 10 (43%) had prolonged unexplained cell count elevations
spanning over years and 4 (17%) had transient cell count elevations attributable to a
reactive inflammatory response, and resolving over weeks. In the 9 with a diagnosis
of MPN, 3 had been diagnosed previously, 3 were new diagnoses made with the
contribution of a bone marrow biopsy, and 3 were diagnosed 1-2 years later with a
957
AMP Abstracts
positive repeat JAK2 test. Conclusions: LL- JAK2 test results in the context of
persistent cell count elevations should warrant close clinical follow-up. In our
experience, 39% were associated to a MPN, and another 43% were seen in patients
with prolonged cell count elevations which may prove to be diagnosed as MPN with
time. Only 17% of LL-JAK2 was associated with transient cell count anomalies.
Thus, in the appropriate clinical context, a low level positive result should be
considered meaningful and closely monitored to detect an evolving MPN.
H16. Clinical Validation and Implementation of a Targeted Sequencing Panel
for Myeloid Neoplasms
D. Steiner1, K. Devereaux2, J. Dudley2, C. Jones3, L. Gojenola3, J. Zehnder2, C.
Kunder2
1Stanford University, Stanford, CA; 2Stanford University, Palo Alto, CA; 3Stanford
Health Care, Palo Alto, CA.
Introduction: Targeted next generation sequencing (NGS) panels have rapidly
become a useful tool in the clinical evaluation of hematologic malignancies. Our
laboratory has recently adapted, validated, and implemented the Illumina TruSight
Myeloid Sequencing Panel for clinical use. In a short time since validation, we have
profiled and reported results for over 100 patient samples and the assay has proven
effective in identifying variants with diagnostic, prognostic, and predictive
significance. Here, we present data on the validation and clinical experience with this
assay at our institution and highlight cases demonstrating the potential value and
challenges of such testing. Methods: Validation was performed on 54 clinical blood
and bone marrow samples as well as additional research specimens and cell line
controls. Results were compared to previous clinical NGS testing from other
laboratories and to validated, mutation-specific assays in our own laboratory.
Libraries were prepared using the TruSight Myeloid reagents and sequencing was
performed on Illumina MiSeq instruments. NextGene software was used for
alignment and variant calling and custom scripts were used to further filter recurrent
artifacts and polymorphisms as well as to format results for review, annotation, and
reporting. Results: Of 110 clinical samples, the most common conditions associated
with ordering of this assay were cytopenias (17%), thrombocytosis (16%), acute
myeloid leukemia (AML) (15%), myelodysplastic syndromes (MDS) (15%), and
erythrocytosis (11%). Pathogenic mutations were detected in 37 of the 54 genes on
the panel and the most frequently mutated genes were TET2, JAK2, SRSF2, and
ASXL1. The diagnoses with the most pathogenic mutations were MDS
(2.6±1.4 mutations per case) and AML (2.5±1.6 per case). Approximately 70% of
cases were peripheral blood specimens and the remainder were bone marrow
aspirates. Variant allele fractions (VAFs) of pathogenic mutations did not differ
significantly between blood and bone marrow for AML and MDS cases. The
identification and annotation of mutations helped to establish diagnoses in the setting
of abnormal cell counts and also informed risk stratification and changes in clinical
management for several patients with known hematologic disorders.
Conclusions: As the use of molecular diagnostics in hematology continues to
evolve, the shared experiences of clinicians, molecular pathologists, and laboratory
scientists will be increasingly important in order to maximize the clinical utility of
targeted NGS assays. Our experience provides useful information on the potential
value of such testing and also highlights challenges and areas for improvement in
the design and implementation of targeted sequencing panels for myeloid
neoplasms.
H17. Lack of Racial Differences in Primary Cytogenetic Abnormalities in
Multiple Myeloma
J. Richter1, E. Hansen2, W. Bartlett2, A. Lakshmanan2, S. Arunajadai2, T. Wu2, E.
Protomastro2, S. Goran2, L. Chen2, E. Jones2, K. Buttner2, R. Pe Benito2, V.
Narayanan2, K. Suryadevara2, S. Kaur2, S. Mathura2, S. Goldberg2
1John Theurer Cancer Center, Hackensack, NJ; 2COTA, New York, NY.
Introduction: Racial disparities in incidence and prevalence in multiple myeloma
(MM) are well known with black individuals having 2-3 fold increased risks. An
ECOG review (Baker, Blood 2013) and a Mayo Clinic study (Greenberg, Blood
Cancer Journal 2015) found a lower frequency of IgH translocations in black patients
(pts). However a Veterans Health Administration study (Blue, BJH 2017) did not
identify racial differences by classical karyotype procedures. Methods: The Cota
database contains diagnostic, clinical and outcome information extracted and
enriched from the electronic health records. The database is purely observational
and has been de-identified for “secondary research” purposes. Data was
aggregated from 12 cancer centers involving 77 hematologistoncologists. Cytogenetics and FISH studies were performed at various laboratories
per physician discretion. FISH data on approximately 90+ black (B) and 400+ white
(W) MM pts for genomic alterations was available for review. Results: Eight hundred
sixty-three patients with newly diagnosed MM between January 1, 2012 and May 1,
2017 were identified in the database; 162 were B and 701 were W. 48% B and 55%
W were male (p=0.11) with age at initial presentation B: 61 yrs (34-87) and W: 66 yrs
(31-90+). Subtypes included IgG kappa B: 40% W: 36%; IgG lambda B: 27%, W:
22%; IgA kappa B: 10%, W: 14%; kappa light chain B: 13%, W: 10%; IgA lambda B:
6%, W: 9%; lambda light chain B: 5%, W: 7%, other subtypes all 1% or
less. Pts had similar distribution of stage: ISS 1: B: 36%, W: 38%; ISS 2: B: 34%,
W: 33%; ISS 3: B: 30%, W: 29% (p=0.9). Similar percentages of hyper-diploid
cytogenetics (B: 26%, W: 31%; p=0.32) and hypo-diploid cytogenetics (B: 21%, W:
24%; p=0.60) were observed. Among pts undergoing FISH analysis B had similar
958
frequencies of IgH translocations: t(14:16) B: 5% W: 7% (p= 0.6); t(4:14) B: 11% W:
9% (p= 0.4); t(11:14) B: 35% W: 30% (p= 0.3); t(14:20) B: 0% W: 2% (p= 0.6); t(6:14)
B: 0% W: 2% (p=1). For all IgH translocations combined, among B pts 47 of 344
specimens were positive (14%) and for W pts 210 of 1630 specimens were positive
(13%). Del 17p was noted in B: 19%, W: 18% (p=0.8); 1q amp in B: 31%, W: 35%
(p=0.45); 1p del in B: 18%, W: 13% (p=0.4), and del 13 by cytogenetics in B: 7% W:
8% (p=1). The 2-year PFS for B stratified by ISS 1-3 were 81%, 59%, and 41%, and
for W were 76%, 79%, and 69% and the 3 year OS by ISS for B were 87%, 62%,
and 72% and for W were 84%, 72% and 65%. Conclusions: We were unable to
confirm lower frequencies of IgH translocations among B pts with MM. The number
of cases with FISH available was similar to the ECOG series, but lower than the
Mayo Clinic study. Our findings suggest that additional evaluations of racial
disparities in genomic alterations in MM are required.
H18. Distinct Patterns of PML-RARA Fusion Gene Formation in High Risk
Acute Promyelocytic Leukemia Revealed by Whole Genome Sequencing
Y. Cho1, S. Park2, M. Lee1, Y. Ju2
1University of Ulsan, College of Medicine and Asan Medical Center, Seoul, South
Korea; 2Korea Advanced Institute of Science and Technology, Daejeon, South
Korea.
Introduction: Previous data available from high-throughput sequencing have shown
that acute promyelocytic leukemia (APL) has a distinct molecular mutation profile
compared to other types of acute myeloid leukemia. However, there are still only a
few studies on the pattern of gene translocation, which is a critical leukemogenic
factor in APL. Methods: Whole genome sequencing (WGS) of 12 patients with APL
was performed to characterize the PML-RARA fusion gene formation and to identify
gene mutations. Analysis of diagnosis-complete remission matched samples was
available in 9 cases. The presence of the PML-RARA fusion gene was confirmed by
reverse transcriptase-polymerase chain reaction in all patients. Results: Six patients
were classified as high risk according to the NCCN risk stratification, and the
remaining 6 patients were classified as low risk. The numbers of bcr3 isoform,
additional cytogenetic abnormalities, and cryptic translocation were 5, 3 and 1,
respectively. WGS analysis categorized the samples into 2 groups: a relatively clean
breakpoint or a gene disruption. The former included a clean breakpoint with
microhomology (n=3) and a clean breakpoint with a small-sized genomic deletion
and/or duplication (n=6). The latter included copy-number loss in the corresponding
genes (n=2) and an insertion at the breakpoint (n=1). All 3 cases with the gene
disruption were found in the high risk patients and thus, had a frequency of 50%
(3/6) among that patient group. One of the cases with the copy-number loss
exhibited cryptic translocation in the conventional G-banding technique. The most
frequent mutation was FLT3-ITD (n=6), followed by mutations in
ARID1A (n=2), DNMT3A (n=1), RUNX1 (n=1), and WT1 (n=1). Conclusions: This
study reported on the genomic characterization of PML-RARA fusion gene formation
in patients with APL, and suggests that a gene disruption mechanism is involved in a
significant number of high risk patients.
H19. Genetic Heterogeneity and Stratification of AML Samples with NPM1
Mutation Detected by the MyAML NGS Test
S. Gramatikova
Genection, San Diego, CA.
Introduction: Abnormalities in multiple genes cause acute myeloid leukemia (AML),
a clinically heterogeneous disease. A recent study has shown that up to one-third of
AML patients would have had their treatment altered as compared to the treatment
they received with current practice recommendations if providers had used a
knowledge bank of matched genomic and clinical data. With a frequency of ~30%,
nucleophosmin (NPM1)-mutated AML is the largest genomic subgroup. The key
purpose of this study is to inform therapeutics using the characterization of the
genetic heterogeneity and stratification of NPM1-mutated AML samples. Methods:
Using the MyAML next generation sequencing (NGS) panel, we analyzed the genetic
profile of 22 AML samples with driver mutations in NPM1. The variants were
interpreted following the recommendations of the Association for Molecular
Pathology, American Society of Clinical Oncology, College of American Pathologists,
and American College of Medical Genetics and Genomics. Results: In all cases
the NPM1 mutations co-occurred with mutations in the FLT3 tyrosine
kinase, DNMT3A methyltransferase, IDH1 or IDH2 isocitrate dehydrogenase genes.
All FLT3 mutations were located in the protein kinase domain and most mutations
were internal tandem duplications (ITD). Four samples included more than one ITD.
Most of the DNMT3A mutations were detected at the mutational hotspot R882;
all IDH1 mutations were at hotspot R132; most IDH2 mutations were at hotspot
R140, and all SRSF2 mutations were at hotspot P95.
The NPM1mut/FLT3ITD/DNMT3Amut genotype, associated with poor prognosis in AML
patients, was observed in 6 samples, close to a previously reported frequency. All
samples included mutations in DNMT3A, IDH1/2, TET2 and WT1, which are involved
in DNA methylation. Conclusions: The MyAML assay allows for comprehensive
molecular profiling of AML samples, which confirms the previously reported
occurrences and associations of driver mutations. The classification “AML
with NPM1 mutation” only partially contributes to the risk stratification and did not
inform currently available treatments, which depend on the mutational profile.
Combinational therapies that target driver mutations are a promising strategy for
improving AML outcome. A combination of DNA methyltransferase
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
and FLT3 inhibitors has been shown to be effective in AML patients. Minimal residual
disease (MRD) monitoring allows a sensitive tracking of leukemia clones in order to
apply personalized medicine. NPM1-positive MRD has emerged as a sole prognostic
factor for relapse. Monitoring of pre-leukemic clone (e.g. DNMT3A, TET2 mutation)
for acquisition of driver mutations during remission might prompt initiation of
aggressive therapy, such as stem cell transplant.
H20. Impact of Molecular Sequencing Information as Related to 2008 and 2016
WHO Classification of Acute Myeloid Leukemia and Myelodysplasia
L.N. Toth, F. Blumental de Abreu, J.D. Peterson, S.F. Allen, E.Y. Loo
Dartmouth-Hitchcock Medical Center, Lebanon, NH.
Introduction: The World Health Organization classification of tumors of the
hematopoietic and lymphoid tissues underwent revisions in 2016. As such, an
uncertain proportion of myeloid neoplasms may require reclassification. Here we
compare 2016 genetic-based classification changes in acute myeloid leukemia
(AML) and myelodysplasia (MDS) from the 2008 criteria. Methods: From 2014-2017,
all bone marrow biopsies with a new suspected myeloid neoplasm were evaluated
using the Illumina TruSight Myeloid Sequencing Panel on the MiSeq platform using
at least 50 ng of genomic DNA. Base-calling and sequence alignment were
performed using MiSeq Reporter Software and analyzed with VariantStudio
v2.1. Results: Four of 43 MDS cases carried SF3B1 mutations, but no
morphologically diagnosable MDS case required reclassification. Forty-five biopsies
performed for cytopenia evaluation did not meet morphologic/cytogenetic criteria for
MDS; 17 of these carried ≥1 myeloid neoplasm associated gene variant. Variant
types recognized by the NCCN as representative of clonal hematopoiesis were
found in 7 of the 17. Three of the 17 cases carried SF3B1 mutations; one contained
at least 5% ring sideroblasts, and was reclassified as MDS with ring sideroblasts and
unilineage dysplasia. Fifty-five AML cases were evaluated; 4 primary (p) AML cases
(34 total) had no identifiable sequence variants but had recurrent translocations by
cytogenetics. All 21 secondary (s) AML cases carried ≥1 myeloid-neoplasm gene
variant. Four cases of AML NOS (14 total) showed features of dyspoiesis that were
insufficient for classification as AML with MDS-related changes, but 2 of these
carried gene variants that are associated with s-AML. Two of the remaining AML
NOS cases which lacked features of dyspoiesis carried s-AML associated variants.
Two cases of AML with dyspoiesis in >10% of a background cell line were classified
as p-AML due to detection of t(8;21) and inv(16), but both carried s-AML related
variants. Four p-AML cases carried RUNX1 variants, including 3 AML NOS and 1
AML with mutated NPM1. One p-AML NOS case and 8 of 9 s-AML cases with
RUNX1 variants showed co-mutation with s-AML related genes. Conclusions:
Sequencing interrogation of AML and MDS is a useful ancillary diagnostic
tool. Sequencing did not appear to significantly alter classification in morphologically
diagnosable MDS, but did identify clonal hematopoiesis in a significant minority of
cases with otherwise unexplained cytopenias. Detection of s-AML related gene
variants (SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, STAG2) may aid
in identifying cases of AML with MDS-related changes. Of note, most cases of AML
with RUNX1 variants were seen in s-AML or associated with s-AML associated gene
variants.
H21. Validation of a Next Generation Sequencing-Based Assay to Detect
Recurrent Translocations in Ph-Like Acute Lymphoblastic Leukemia
D. Duose, C.Y. Lan, R. Luthra, I.I. Wistuba, N. Jain, M. Konopleva, K.P. Patel
University of Texas MD Anderson Cancer Center, Houston, TX.
Introduction: Philadelphia chromosome–like (Ph-like) acute lymphoblastic leukemia
(ALL) represent a recently identified subgroup of ALL associated with poor outcomes
in both children and young adults. These cases are negative for Philadelphia
chromosome, but, carry a distinct set of chromosomal translocations involving
kinases that result in a gene expression signature similar to Ph-positive ALL, and
hence the term Ph-like. It is, therefore, important to identify recurrent translocations
in Ph-like ALL at diagnosis. However, the large numbers of translocations that
constitute this entity represent a challenge for routine clinical testing. The goal of this
study is to validate a diagnostic clinical assay for identifying Ph-like gene fusions in
acute leukemia patients. Methods: DNA from 26 de-identified patient samples
positive for various Ph-like ALL fusions were obtained from the Children’s Oncology
Group (COG) for test validation. These fusions included re-arrangements
involving ABL1, ABL2, CRLF2, CSF1R, EPOR, JAK2, NTRK3 and
PDGFRB. Libraries were prepared with an input RNA of 100ng using the
Archer FusionPlex ALL panel (ArcherDx, Boulder, CO) per manufacturer’s
instructions. Amplified libraries were quantified with Kapa kit (Roche scientific),
multiplexed (8 samples per V2 cartridge) and sequenced on Illumina’s MiSeq
platform. Data analysis was performed with Archer analysis software 4.1.1
(ArcherDx) and positive fusion calls were made for samples at least 5 breakpoint
spanning reads, minimum of 10% of the overall reads spanning the breakpoint and
minimum of 3 unique start sites. Limit of detection was performed using serially
diluted cell lines. Results: The Archer FusionPlex ALL panel detected 26/26
expected fusions present in commercially available positive control samples. These
include translocations involving ABL1 (partners NUP214, ETV6, ZMIZ1, RANBP2,
RCSD1, FOXP1, NUP153, SPTAN1, CENPC), ABL2 (CSD1, ZC3HAV1, PAG1ABL2), JAK2 (BCR, PAK5, SSBP2, ZNF274, ATF7, ETV6, TERF2), PDGFRB
(EBF1, ATF7IP, ZEB2), SSPP2-CSF1R, P2RY8-CRLF2, EPOR, and ETV6NTRK3. Assay reproducibility (R2= 0.99) was established by repeating the library
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
preparation and sequencing of 3 of the same patient samples in an independent run.
Sensitivity studies showed a limit of detection at 2.5%. Validation using in-house
patient samples for compatible sample types, recurrent additional translocations and
hotspot mutations covered on the panel is in progress. Conclusions: The
Archer FusionPlex ALL panel is able to accurately and repeatedly detect a wide
range of known Ph-like ALL fusions with an analytical sensitivity of 2.5%. This makes
it a promising diagnostic assay for stratifying treatment options for patients with Phlike ALL.
H22. Isocitrate Dehydrogenase 1 and 2 Mutations in Myeloid Neoplasms
L.N. Toth, F. Blumental de Abreu, J.D. Peterson, S.F. Allen, E.Y. Loo
Dartmouth-Hitchcock Medical Center, Lebanon, NH.
Introduction: Isocitrate dehydrogenase (IDH) 1 and 2 mutations have been
described in solid tumors and are common in acute myeloid leukemia
(AML). Missense mutations in exon 4 of either IDH gene can result in an oncogenic
gain of function with production of (R) enantiomer of 2-hydroxyglutarate ((R)-2-HG)
instead of α-ketoglutarate (α-KG). This oncometabolite inhibits α-KG dependent
enzymes, causing hypermethylation and impaired cell differentiation. Selective IDH
inhibitor drugs are in development. We reviewed all myeloid neoplasm cases
with IDH1/2 variants seen at our institution. Methods: From October 2014 through
March 2017 all bone marrow biopsies with a new suspected myeloid neoplasm were
evaluated using the Illumina TruSight Myeloid Sequencing Panel on the MiSeq
platform. Base-calling and sequence alignment were performed using MiSeq
Reporter Software and analyzed with VariantStudio v2.1. Results: Fifteen cases
harboring IDH1/2 variants were found (7 IDH1, 8 IDH2; 5/34 primary (p)-AML, 5/21
secondary (s)-AML, 2/42 MDS, 2/27 MPN, 1/112 non-neoplastic). All
identified IDH2 variants were at the R140 codon with none at R172. Five of
7 IDH1 variants involved the R132 codon; 2 other variants were at R199 and
N101. ~18% of AML and ~≤5% of the remaining cases carried IDH1/2 variants. The
distribution of IDH1 versus IDH2 was roughly equal across all groups. RUNX1 comutation was most frequent, occurring in 6 cases (2 p-AML, 3 s-AML, 1 MPN - CML),
and association with methylation pathway variants (TET2, DMNT3A, and KDM6A)
was also frequently seen (47%). Additional co-variants included MPL, NPM1,
RAD21, FLT3, GATA2, SRSF2, ASXL1, STAG2, SMC3, SF3B1, NRAS, TP53,
SETBP1, PHF-6, U2AF1, JAK2, and CUX1. In p-AML, variants were seen in normalkaryotype AML NOS (n=2) or associated with NPM1/FLT3 mutation (n=3). All treated
p-AML patients are alive (11.5 months median followup, range: 3.5-41.7 mo), 4 of 5
s-AML patients died (1.5 months median survival, range: 0.3-12.1 mo). One of the
MDS patients was thought to be in transformation to AML and underwent 7+3
induction chemotherapy; currently alive and enrolled in an IDH1 inhibitor trial. The
CML case with co-mutated RUNX1 was identified at accelerated phase; died day 72
s/p HiDAC induction. Conclusions: Variants in IDH1/2 may be seen in all myeloid
neoplasms. Frequency is higher in AML, but not generally in cases with recurrent
translocations. Variants in p-AML did not seem to portend a poor outcome, but did
appear associated with disease progression in lower-grade neoplasms and with
adverse outcomes in s-AML. Co-mutation with RUNX1 was very common, which has
not generally been described. IDH1/2 variation was extremely uncommon in our nonneoplastic cases with clonal hematopoiesis of indeterminate potential.
H23. RNA-Based Immune Repertoire Sequencing for Characterizing B-Cell
Lineage Malignancy Clonality and IGHV Mutation Status
J. Haimes1, S.J. Mishkin1, N.M. Nair1, T.D. Harrison1, L.M. Griffin1, M.L. Gulley2, N.D.
Montgomery2, B.A. Kudlow1
1ArcherDX, Boulder, CO; 2University of North Carolina School of Medicine, Chapel
Hill, NC.
Introduction: B-lymphocyte malignancies are characterized by monoclonal
expansion of cells with related, if not identical, immunoglobulin gene sequences.
Sequencing the immunoglobulin heavy chain (IGH) repertoire to define tumorassociated clonotypes has a number of potential clinical applications, including
identifying specific markers for residual disease, and determining somatic
hypermutation status for risk stratification in chronic lymphocytic leukemia
(CLL). Compared to DNA, RNA-based clonotype sequencing has advantages, as
primers can be positioned to determine isotype, and non-expressing cells do not
dilute out signals from B cells. We tested an RNA-based next-generation
sequencing (NGS) assay based on Anchored Multiplex PCR (AMP) to detect
dominant clonotypes, identify IGH isotypes, and characterize IGHV mutation status
in a cohort of patients with multiple myeloma (MM), lymphoplasmacytic lymphoma
(LPL) and chronic lymphocytic leukemia (CLL). Methods: Sequencing mRNA from
MM, LPL and CLL was performed with ArcherDX Immunoverse IGH assay. Total
RNA was extracted from formalin-fixed bone marrow collected from patients with MM
and LPL or from fresh blood collected from patients with CLL. Libraries were
sequenced on an Illumina MiSeq using v3 600-cycle chemistry, and data were
analyzed using Archer Analysis v5.1 for clonotype frequencies, IGH isotypes,
and IGHV mutational status. Results: In all IGH-expressing MM and LPL cases,
strongly dominant clonotypes representing >90% of all clones were present. In
contrast, dominant clonotypes representing ≥2% of all clones were not detectable in
IGH non-expressing MM specimens nor in blood from normal donors. The assay
accurately defined isotype, as results were 100% concordant with isotype defined by
serum immunofixation. Similar results were observed for CLL samples, although
disease-associated clonotypes were less dominant than in MM/LPL, potentially
959
AMP Abstracts
reflecting lower IGH expression in CLL. Furthermore, because reads typically
covered >90% of IGHV, somatic hypermutation status was discernable and was
highly concordant with mutation status inferred from ZAP70 and CD38 protein
expression. Conclusions: RNA-based sequencing from both formalin-fixed tissue
and fresh blood permits characterization of IGH repertoire in clinical samples. Even
in formalin-fixed marrow, the targeted NGS assay robustly identified dominant
clonotypes while providing IGH isotope and somatic hypermutation status.
Unambiguous identification of CDR3 sequences provides a marker by which to
monitor residual tumor burden after treatment.
H24. Utilization of Peripheral Blood for Diagnostic Testing for MDS/MPN
Patients: Efficacy and Benefits of a SNP Microarray Analysis
S. Schwartz1, H. Dong2, R. Thomason3, P. Papenhausen1
1Laboratory Corporation of America Holdings, Durham, NC; 2Integrated
Oncology/Laboratory Corporation of America Holdings, New York, NY; 3Integrated
Oncology/Laboratory Corporation of America Holdings, Brentwood, TN.
Introduction: Technologies utilized for the detection of cytogenetic abnormalities in
hematological malignancies has rapidly evolved over the past decade. Microarray
analysis has mainly been used to study constitutional aberrations, but more recently
has been used in oncology. Most work in oncology has involved bone marrow
analysis. The present report reviews a large laboratory’s experience with the
utilization and efficacy of microarray analysis for analysis of myelodysplasia (MDS)
and myeloproliferative disorders (MPN) from peripheral blood. Methods: This study
addresses 845 patients in which peripheral blood was studied by microarray analysis
(using an Affymetrix Cytoscan HD array) to detect aberrations in myeloid
disorders. Comparisons of the peripheral blood microarray analysis were made to
FISH and cytogenetic studies (when available) as well as microarray analysis from
bone marrow in a subset of patients. Results: Cytogenetic analysis from
unstimulated blood was only successful in 64.4% of patients, while microarray
studies were successful in all patient samples. The array analysis was concordant
with FISH studies in 99.9% of the patients; however the array analysis provided
additional information in 39% of the patients studied by both modalities. The array
detected abnormalities in 56% of the patient samples thought to have or be
suspicious for MDS/MPN by clinical presentations, flow cytometry and/or morphology
studies. Approximately 62% of the anomalies were solely copy-number changes;
whereas 22% were restricted to copy-neutral loss of heterozygosity and 16%
included both. Twenty-two concurrent peripheral blood and bone marrow samples
were studied in these patients and in all cases yielded the same result.
Conclusions: The results from the analysis of these 845 patients, have revealed
several important findings: 1) This study clearly demonstrates the improved efficacy
of using the array analysis of peripheral blood specimens assist in the diagnosis of
MDS/MPN; 2) In all concurrently studied patients the peripheral blood studies were
equivalent to bone marrow analysis; 3) Arrays on unstimulated blood was more
effective than chromosome studies, especially in samples with little or no blast
cells, providing results in all patients studied; 4) Arrays successfully revealed the
same abnormalities seen by a MDS FISH panel yielding additional prognostic
findings in more than a third of the patients (e.g. deletions of RUNX1, BCOR, TET2,
and CN-LOH 4q, 7q, 11q). 5) The findings suggest that the higher diagnostic yield of
the microarray can easily supplant the use of routine FISH that is restricted to copy
number changes in peripheral blood samples and limited by 4 probes (5/5q, 7/7q, 8
and 20q) in a routine MDS FISH panel.
H25. Clinical and Genetic Characteristics of MYC Gene Aberration in Multiple
Myeloma
S. Min, E. Seo, C. Seol, C. Park, S. Jang, Y. Cho, C. Suh, J. Hong, D. Yoon
Asan Medical Center, Seoul, South Korea.
Introduction: MYC gene rearrangement is frequently observed in non-Hodgkin
lymphoma including Burkitt lymphoma and aggressive B-cell lymphomas. However,
only a few studies have been recently conducted on MYC gene rearrangement in
multiple myeloma (MM). This study aimed to investigate the characteristics of MYC
gene aberrations and their correlation with other genetic abnormalities. Methods: A
total of 82 patients with MM were examined at the time of diagnosis. The median age
was 62 years (range, 38-87); the study included 52 males and 30 females. We
performed the Immuno-FISH using MYC breakapart probe in bone marrow samples.
Chromosomal analysis and MM FISH panel for 1q, IGH/FGFR3, IGH/CCND, 13q,
IGH/MAF, and TP53 were performed on all patients. We also analyzed laboratory
data and MM scoring results (International Staging System, Revised International
Staging System, Mayo Stratification of Myeloma and Risk-Adapted Therapy, and
IMWG Consensus on Risk Stratification in Multiple Myeloma). Results: MYC gene
aberrations were detected in 35 (42.7%) patients. Nineteen patients (54.3%) had
gain in MYC gene copy number. MYC gene rearrangements were detected in 15
patients (42.9%). One patient showed MYC gene amplification more than 10 copies.
MYC gene rearrangements showed higher LD level and higher plasma cell
proportion in bone marrow examinations (p<0.05). Patients with MYC gene
rearrangements showed tendency to high plasma cell index (p=0.074). Patients with
MYC gene rearrangements had a relation of hypodiploidy (p=0.055). MYC gene
rearrangements did not accompany IGH/FGFR3 translocation and IGH/MAF
translocation. MYC rearrangements or gains were not associated with any scoring
system (P>0.05). In patients with 1q gain, patients with MYC gene arrangements
showed higher LD level (p<0.05). However, hemoglobin levels were lower in patients
960
without MYC gene rearrangements (p<0.05). Conclusions: MYC gene
rearrangement was significantly associated with plasma cell proportion in bone
marrow, suggesting that MYC gene play a role in excessive proliferation of
neoplastic plasma cells. The relationship between MYC gene rearrangement and
poor prognostic factors such as high LD, high plasma cell proportion, and high
plasma cell index, could be an evidence of negative prognostic influence of MYC
gene rearrangement.
H26. Clinical Validation of a Molecular Barcoded Amplicon-based Next
Generation Sequencing Test for Mutation Profiling of Myeloid Neoplasms
T. Yang, A. Almradi, G. Smith, C. Hill, L. Zhang
Emory University School of Medicine, Atlanta, GA.
Introduction: Mutation profiling has become standard of care for patients with
myeloid neoplasms including myelodysplastic syndrome (MDS), myeloproliferative
neoplasms (MPN), MPN/MDS, and acute myeloid leukemias (AML). Next generation
sequencing (NGS) is widely accepted as the preferred method to obtain
comprehensive genetic abnormality information in clinical laboratories. We validated
a targeted gene NGS test with a molecular barcoded, DNA amplicon-based library
preparation for the Illumina platform. Method: Commercially available molecular
barcoded, amplicon-based library preparation kit for DNA template was customtailored to capture all or a fraction of exons of 61 targeted genes with 1278
amplicons. Extracted DNAs of previously sequenced cases were obtained from a
reference laboratory for test validation. The libraries were sequenced on an Illumina
MiSeq or Nextseq. A third-party bioinformatics pipeline was used to process the
sequence data. A customized quality control module was also built to facility the
result review by pathologists. The read depth was expressed as “super-reads” to
reflex the unique reads with different molecular barcodes. The identified variants
were compared and discrepancies were confirmed by orthogonal methods. The
analytic sensitivity was determined by diluting the DNA of known mutant allele
frequency with samples of wild type sequences. Reproducibility was tested by repeat
sequencing on a separate library prepared by a different laboratory staff.
Results: DNA extracted from 19 fresh blood or bone marrow samples and 1
formalin-fixed paraffin-embedded (FFPE) tissue, including 3 MDS, 9 MPN, 2
MPN/MDS, and 6 AML, were included in this validation. A total of 76 variants in 41
genes previously detected at ≥5% mutant allele frequency, including 57 point
mutations and 19 small deletions/insertions (indels) were detected at mutant allele
frequencies similar to that reported by the reference laboratory. Some noised was
recognized and excluded by the fact that they were present in every sample at
relatively low-frequency. Our panel detected 16 new variants not reported by the
reference laboratory. Repeat sequencing showed complete reproducibility. The
coverage in CEBPA exon is satisfactory (minimal 120, except the FFPE sample with
universal low depth). Due to the limitation of the bioinformatics pipeline, a large FLTITD and a large deletion of CALR were not detected by this NGS test.
Conclusion: The molecular barcoded, DNA amplicon-based NGS mutation profiling
easily achieves an analytic sensitivity of 5% mutation allele frequency. Continuous
validation is currently planned to investigate the lowest limit of detection.
Improvement of bioinformatics pipeline or reflex test is required to detect large
indels.
H27. Performance of ACL LDT CALR Exon 9 Assay
L.J. Mazur, B. Baltadjieva, L. Ma, O. Jedry, M. Patel, M. Cuaresma, T. Wager, L.
Sok, M. Mihalov
ACL Laboratory, Rosemont, IL.
Introduction: Myelofibrosis (MF), polycythemia vera (PV) and essential
thrombocythemia (ET) are a group of heterogenous disorders of the hematopoietic
system collectively known as Philadelphia chromosome-negative myeloproliferative
neoplasms (MPN). The prevalence of MF, ET and PV in the United Sates is
estimated to be approximately 13,000, 134,000 and 148,000, respectively. The
diagnosis and the management of patients with MPN has evolved since the
identification of mutations that activate the JAK pathway (Jak2, CALR and MPL) and
the development of targeted therapies has resulted in improvements in diseaserelated symptoms and quality of life. Since 2016 WHO diagnostic criteria include
molecular testing for JAK2, CALR and MPL mutations for MF and ET and molecular
testing for JAK2 V617F or JAK2 exon 12 mutations for PV. This study presents
performance of ACL LDT assay for CALR exon 9. Methods: Fifty-eight previously
tested clinical samples by comparator lab (14 samples positive for deletion (52 bp)
frequency range 7-49%, 6 samples positives for insertion (5 bp) frequency range 2343%, and 38 CALR wild type (WT samples). Whole blood and bone marrow
collected in lavender EDTA samples were extracted using NucliSENS EasyMAG
protocol. Method 1. Samples tested by comparator laboratory using Polymerase
Chain Reaction/ fragment analysis by capillary electrophoresis. Method 2. ACL LDT
developed dual target assay using Polymerase Chain Reaction/ fragment analysis by
capillary electrophoresis. Analytical sensitivity and LOD were determined by run
serial dilutions of ACL laboratory developed external QC reagents specific for:
deletion 52 bp and insertion 5 bp. Results: ACL LDT CALR 9 method demonstrated
complete agreement with the comparator laboratory: 20/20 clinical samples with
positive variant were detected, 38/38 wild type samples were correctly analyzed and
detected. Accuracy of the method is (100%). Analytical sensitivity was assessed 2
ways; DNA serial dilution from patient’s samples and serial dilutions of 2 QC
plasmids for deletion (52 bp) and insertion (5 bp). Dilution studies confirmed a linear
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
and quantitative assay (LOD ~2% mutant allele frequency for each variant).
Optimum DNA concentration for the assay was determined 1-4 ng/uL.
Reproducibility was assessed by testing 5 levels of dilutions and run 24 times,
24/24=100%. Precision was as follows; deletion (%cv 0.3-0.01), insertion (%cv 0.40.03). Conclusions: This study demonstrates that ACL LDT CALR 9 assay
performed very well in comparison to comparator laboratories and indicates the
assay is suitable for clinical use in MPN patient population.
H28. Haplotype Counting for Sensitive AML Relapse Detection
M. Debeljak, E. Mocci, M.C. Morrison, A. Pallavajjalla, K. Beierl, M. Amiel, M.
Noe, L.D. Wood, M. Lin, C.D. Gocke, A.P. Klein, E.J. Fuchs, R.J. Jones, J.R.
Eshleman
Johns Hopkins University, Baltimore, MD.
Introduction: Fields of forensics, paternity, and hematopoietic stem cell
transplantation (HSCT) require human identity testing. The polymorphic nature of
short tandem repeats (STRs) makes them most frequently used marker but they are
relatively insensitive. In 2014, we demonstrated that using multiple SNPs on a short
amplicon (microhaplotype) could overcome the inherently high error rate of NGS,
and can be used for ultrasensitive detection of human DNA mixes. In addition, we
identified novel loci in the human genome that can be used for
microhaplotyping. Here we present informativities of 8 loci based on 45 individuals
from three 1000 genomes populations. Furthermore, using this approach we
evaluate its ability to detect relapse at an earlier time point. Methods: We designed
primers to target 8 SNP-dense loci (TFB2M, SORCS2, HLA-A, HLA-B, CAMD1,
FARP1, MT4, and TMPRSS15). We then obtained and amplified samples from 15
individuals in 3 populations used in 1000 Genomes Study (total of 45 individuals;
CEU, JPT, YRI) at all 8 loci. Amplicons were pooled and sequenced by
NGS. Haplotypes were determined from aligned sequencing data. In addition, a
total of 15 bone marrow samples from patients who underwent allogeneic BMT were
obtained. Of the 15 samples, 8 became host STR positive and subsequently
relapsed (relapse group) while the other 7 samples remained fully engrafted
(nonrelapse group). Results: All of the 8 loci showed high diversity (mean H, 0.81;
PIC, 0.80) with FARP1 locus as the least diverse region (H, 0.73; PIC, 0.72)
and MT4 as the most diverse region (H, 0.90; PIC, 0.90). Diversity by population
showed a common trend for most of the regions where the JPT population generally
showed lowest diversity and the YRI population generally showed the highest
diversity. In addition, using a cohort of leukemic patients who underwent BMT and
remained in remission, we defined threshold of 0.26% host DNA. In patients who
relapsed, we detected increased recipient DNA above this threshold, 90 and 145
days earlier than microsatellite positivity, and 127 and 142 days before clinical
relapse in 4 of the 8 patients (50%). Conclusions: Ultrasensitive detection of human
DNA mixes can be achieved using multiple microhaplotypes. Here we provided
additional loci that are polymorphic and informative. Furthermore, using this strategy
with a cohort of leukemic patients who underwent allogeneic BMT we demonstrate
the ability to detect increased host DNA at an earlier time point than a conventional
STR-based assay.
H29. Evaluation of Fragment Analysis Assay for Detection of CALR Exon 9
Insertion and Deletion Mutation in Myeloproliferative Neoplasms
J. Cho, Y. Kim, G. Oh
Eone Laboratories, Incheon, South Korea.
Introduction: Mutation in exon 9 of Calreticulin (CALR) gene is second most
frequent somatic mutation after JAK2 in myeloproliferative neoplasm. The objective
of this study was to evaluate a sensitive fragment analysis assay to
detect CALR mutation in the clinical laboratory setting. Methods: Genomic DNA was
obtained from 49 patients with myeloproliferative neoplasm. We performed fragment
analysis and Sanger sequencing simultaneously to determine CALR mutation. The
performance of fragment analysis was evaluated in accuracy, precision and
analytical sensitivity. Results: Among the 49 patients, 8 patients were positive
for CALR mutation. Detected mutations included a 52 bp deletion (n=3), 5 bp
insertion (n=2), novel 5 bp insertion (n=1), 34 bp deletion (n=1) and co-existence of
52 bp deletion and 5 bp insertion (n=1). Fragment analysis showed 100%
concordance with Sanger sequencing in all specimens. For accuracy of the assay, 6
mutant and 4 nonmutant specimens were confirmed by second accredited
laboratory. Multiple runs of the fragment assay were performed and precision based
on the total CV of fragment size and mutant burden was <0.1% and <0.5%,
respectively. Analytical sensitivity using mutant and wild type DNA allowed the
detection of 2% mutant allele for insertion and deletion mutation.
Conclusions: Accurately identify the CALR mutation is important in patients with
myeloproliferative neoplasm for optimal diagnosis and therapeutic management.
Fragment analysis is an accurate and sensitive method for the detection
of CALR mutation in the clinical laboratories.
H30. Hematopoietic Tumor Contamination in Human Fingernail Clippings Used
as a Germline Comparator in an NGS-based Myeloid Panel
D. Olson1, Q. Wei2, M. Lovell1, A. Treece1
1University of Colorado Denver/Children's Hospital Colorado, Aurora, CO; 2Center for
Genomic Medicine/Miami Cancer Institute, Miami, FL.
Introduction: Interpretation of next-generation sequencing (NGS) somatic cancer
mutation panels can be complicated by the presence of germline variants. The most
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
reliable method of discrimination requires comparison of tumor genomic data to nontumor genomic data. Our institution uses fingernail (FN) clippings as a source of
germline DNA for our NGS myeloid malignancy panel. Advantages of FN clippings
include ease of acquisition, storage, and transport. Other easily obtained samples
are buccal swabs and mouth washings, but studies have shown white blood cell
(WBC) contamination is a significant problem. To our knowledge, there is no
literature exploring hematopoietic tumor contamination in human FN clippings used
as a germline comparator for somatic mutation testing. Methods: Cases were
identified in which FN clippings were submitted concurrently with a diagnostic bone
marrow (BM) specimen for analysis by our NGS myeloid panel. Somatic variants in
the BM were reviewed for variant allele frequency (VAF). FN clipping results were
reviewed to determine the extent of contamination from somatic variants.
Results: We analyzed 16 BM and FN clipping sample pairs. In the BM samples, we
identified a total of 43 non-synonymous, somatic variants in 18 genes. In the FN
samples, we identified 24 of the same variants in 16 genes. Tumor contamination
was present in 12/16 (75%) patients. The average contamination rate (FN VAF/BM
VAF) was 9.6%, with a median of 1.6% and a range of 0% to 48%. In the FNs, 19/43
variants were undetectable. The VAF ranged between 9% to 80% in the BMs and
0% to 20% in the FNs. The relationship between BM VAF and FN VAF was not
linear, though variants from individual patients showed similar rates.
Conclusions: These data demonstrate that tumor contamination of FN clipping
samples does occur in myeloid malignancies. Limited data are available for
comparison to other sample types, but donor DNA in buccal swabs of hematopoietic
stem cell transplant patients has been reported at an average of approximately 15%
(Chaudhary, et al., 2015). Explanations for this phenomenon include blood (and
tumor cell) contamination of the FN sample as well as tumor nucleic acid
incorporation into the epithelial cells. In all cases, the VAF in the FNs was low
enough to confidently call the variants somatic. Of note, the VAF in the BM samples
did not predict the VAF in the FN samples, suggesting specimen-dependent factors.
The specimen wash step used prior to DNA extraction may be insufficient to
completely remove patient WBCs. Limitations of this study include the absence of
data on preanalytic factors, including cleanliness of FN samples at collection.
Awareness of tumor contamination in this convenient sample type is important in the
era of precision medicine.
H31. Successful Coverage of Difficult to Sequence Genes (CALR, CEBPA, and
FLT3) Associated with Myeloid Disorders Using a Hybridisation-Based
Enrichment Approach Prior to Next-Generation Sequencing (NGS).
L. Georgieva1, E. Uddin1, J. Chan1, W. Wright2, G. Speight1
1Oxford Gene Technology, Begbroke, United Kingdom; 2Oxford Gene Technology,
Inc., Tarrytown, NY.
Introduction: The application of short read NGS for research into myeloid disorders
such as myeloproliferative neoplasms (MPNs) and acute myeloid leukaemia (AML)
has been hampered by the inability to sequence certain genes. These genes can
harbour key mutations so it is desirable to ensure suitable sequencing coverage is
obtained. These genes amongst others include: CALR exon 9 insertions and
deletions (up to 52 bp), CEBPA single nucleotide variants (SNVs) and FLT3 Internal
Tandem Duplications (ITDs) and SNVs. Each of these regions contain certain
challenging DNA sequences that can impact the quality of the data generated, e.g.
large indels and low complexity regions (CALR), high GC content (75% on average
for the whole gene with specific regions at 100%) and repetitive regions (CEBPA),
and complex repetitive elements (FLT3). Methods: We utilised a hybridisation-based
(4 hour hybridisation) enrichment approach for library preparation to determine
whether this approach would overcome the difficulties associated with these genes,
and permit the generation of high quality (sufficient de-duplicated depth) data to
allow these targets to be accurately interrogated. We utilised the SureSeq myPanel
NGS Custom AML panel and associated library preparation kit to analyse 20
research samples of varying DNA inputs that contained key variants in each of these
difficult regions. The resulting library was subsequently sequenced using a 2x150 bp
read length protocol on an Illumina MiSeq. The data was analysed using the
SureSeq Interpret Software. Results: Here we present the coverage and a broad
range of variants generated from these research samples for each of these difficult
to sequence genes. The results clearly show that this approach can reliably detect
and accurately size (including low allele frequency) both insertions and deletions
(including the 52 bp deletion in CALR (exon 9), SNVs and deletions in CEBPA with a
de-duplicated depth in excess of 2000x as well as ITDs of a size between 24 and
201 bp in FLT3. Conclusions: This approach is suitable for the analysis by NGS of
these difficult genes and therefore removes the requirements for supplementary
approaches to analyse these difficult genes, such as Sanger sequencing (CEBPA)
and fragment analysis (CALR and FLT3).
H32. Commonly Mutated Genes across Myeloid Malignancies Using a Targeted
NGS Panel: A Single Institution Experience
J. Yan1, A. Huho1, D. Czuchlewski1, M.A. Vasef2
1University of New Mexico, Albuquerque, NM; 2University of New Mexico/TriCore Ref
Laboratory, Albuquerque, NM.
Introduction: Per the Cancer Genome Atlas (TCGA) published data, there are
approximately 23 genes that are frequently and significantly mutated in acute
myeloid leukemia (AML). In this study, we analyzed a large number of myeloid
malignancies including de novo and recurrent AML, myelodysplastic syndrome
961
AMP Abstracts
(MDS) and myeloproliferative neoplasms (MPNs) using an NGS-based targeted
myeloid gene panel. Considering the unique ethnic groups residing in New Mexico,
including Native American and Hispanic populations, we evaluated the frequency of
mutated genes across myeloid malignancies and compared our results in de novo
AML cases with those published in TCGA data. Methods: A total of 42 adult AML
cases including 34 de novo and 8 recurrent AML cases were studied (mean age of
62). Additionally, we analyzed 11 MDS and 9 BCR-ABL1 negative MPN cases in
the sample cohort collected from January 2016 to May 2017. DNA was extracted
from diagnostic peripheral blood or bone marrow samples, subjected to PCR
amplification of target regions in 57 genes relevant to myeloid malignancies, followed
by library preparation and sequencing on an Illumina platform. Results: Pathogenic
mutations were identified in 27 out of 57 genes. Among de novo AML cases,
mutations in genes at frequency of more than15%, included TET2 (26.4%), FLT3
(23.5%), SRSF2 (23.5%), NPM1 (23.5%), TP53 (17.6%), NRAS (17.6%),
and DNMT3A (17.6%). AML patients in our population showed a much higher
frequency of mutations in epigenetic modifier genes compared to TCGA data. In
addition there was a significantly higher rate of SRSF2 mutations in our
cases. TET2 was the most frequently mutated gene (27%) in MDS, followed by
DNMT3A and NRAS (each 18%). JAK2 was the most frequently mutated gene
(78%) in MPN, followed by DNMT3A and ASXL1 (each 22%).
Conclusions: Compared to TCGA published data on de novo AML cases, our
patients showed more frequent mutations in epigenetic modifier genes particularly
in TET2 that showed a much higher frequency at 26.4% compared to 8.5% in the
TCGA data. We also noted significant difference in the frequency of the spliceosome
gene SRSF2 (23.5% compared to 0.5% in TCGA) and tumor suppressor
gene TP53 (17.6% compared to 8% in TCGA) in our AML cases. These findings are
significant and likely bear clinical impact on risk stratification and clinical
management of our unique patient population.
962
a gene list is recommended for implementing medically informed genomic profiling of
HM. This is a resource intensive exercise that currently lacks supportive references.
Described are implementation considerations and methods to develop a targeted
genomic assay for HM. Methods: A literature search for descriptive implementation
references was performed. VRAs were reviewed for gene targets interrogated
including alteration type. Concurrently, a custom list of targets with potential clinical
relevance was compiled using: clinical guidelines, a literature search, and assay
content from clinical labs performing NGS testing for HM. Each target on the custom
list was reviewed for: clinical utility (diagnostic, prognostic, therapeutic), relevant
alteration type, and coordinates requiring coverage. The list was cross-checked
across 5 VRAs to determine if any were adequately supportive or if a custom build
would be required. The list was further vetted by the Technical Director to assess
design feasibility and assess viable technical solutions. The master list was sorted
into disease types and compared to similar genomic panels offered by other
labs. Key stakeholders (hematopathology, molecular pathology, oncology, and
payers) independently reviewed the content and design proposal and feedback was
incorporated. Results: There is a lack of supportive references for implementing
clinical NGS for HM. We have compiled our process and resources deemed useful in
creating a preliminary custom list of 153 genes with potential clinical relevance. We
describe logistical clinical considerations for panel design and define 6 HM
categories: leukemia, myeloproliferative, myelodysplastic,
myeloproliferative/myelodysplastic, B-cell lymphomas, and T-cell lymphomas.
Rigorous transdisciplinary vetting resulted in a custom panel of 104 targets: 75%
supported by clinical guidelines and 25% supported by medical literature showing
clinical relevance in patients. Conclusions: Developing a genomic panel for HM is a
resource intensive effort that is not well-supported in the literature. We describe
implementation considerations and present a medically-vetted custom gene list that
is supported by clinical guidelines, up-to-date medical literature, and key
stakeholders as described herein with the goal of simplifying this task for the many
laboratories entering into the clinical NGS space for HM.
H33. Targeted Sequencing of Recurrently Mutated Genes in Myeloid
Neoplasms Using the Raindance Thunderstorm-Illumina Miseq Platform: My
Heme (Myeloid Hematologic Malignancy) Panel
S. Cheng1, K. Singh1, Y. Liu2, M.J. Kluk1, D.C. Hassane1, W. Tam1
1Weill Cornell Medicine, New York, NY; 2University of Pittsburgh Medical Center,
Pittsburgh, PA.
Introduction: The mutation status of many recurrently mutated genes is becoming
increasingly important for the classification, prognostication and initiation of targeted
therapy in patients with myeloid neoplasms. For example, mutations in TET2,
SRSF2, JAK2, CALR and MPL have been used for diagnosis of CMML and MPN.
Precision therapies with JAK2 inhibitors, hypomethylating agents and midostaurin in
patients with MPN, MDS and AML rely on mutation statuses identified in JAK2,
TET2/DNMT3A/ASXL1 and FLT3, respectively. A practical and robust targeting
sequencing panel assay is needed to meet this clinical demand. Methods: Targeted
enrichment of 45 genes recurrently mutated in myeloid malignancies was performed
using the Thunderstorm system with a custom primer panel. The primers target
coding exons of the genes, leading to a total of 726 amplicons. Libraries were
prepared by microdroplet-based PCR target enrichment method from DNA, followed
by sequencing using the Illumina MiSeq yielding 260-bp paired end reads.
Sequencing data were analyzed and reported with a customized analytical pipeline.
Besides 6 established cell lines, a cohort of 72 fresh or frozen bone marrow and
peripheral blood samples with available orthogonal mutation data were used for
assessing overall assay performance, including its sensitivity, reproducibility,
specificity and accuracy. Results: In 72 individual samples, a total of 29 unique
SNVs and 22 INDELs variants were identified with a broad range of VAFs (1.3%96.8%). QC metric showed that the minimum and median Q30 of mapped reads are
94.17% and 96.43% respectively, and that median coverage of the sequencing
targets was 1309X with 96.75% of median coverage uniformity (Bases 20% Mean
Depth). Comparison between our panel and the orthogonal assays demonstrates
100% concordance in all the variants detected. The sensitivity of the assay is ~ 2 to
5% for SNVs, and ~1% for INDELs. Of note, the current assay can detect FLT3 ITD
up to at least 120 bp long, and also captured 8 additional and clinically significant
INDELs or SNVs variants undetected by the orthogonal methods, some of which can
be used to direct precision therapies for AML patients, further confirming its superior
accuracy and sensitivity and broader clinical utility. Moreover, the low intra-run (SE:
0.24 to 1.08%) and inter-run (SE: 0.25 to 2.39%) variability in the VAFs near the limit
of detection was observed, indicating excellent reproducibility of the
test. Conclusions: This target sequencing assay demonstrates excellent
performance characteristics, and has recently been approved for clinical use by the
New York State. It should serve as an invaluable tool to improve clinical
management of patients with myeloid neoplasms for hematologists and
hematopathologists.
H35. Development of a Targeted Next Generation Sequencing Panel for
Multiple Myeloma
M. Mai1, L. Frederick1, M. Hubbard1, K. Bessonen1, M. Timm1, D. Jevremovic1, A.K.
Stewart2, P.L. Bergsagel2, E. Braggio2, N. Mehta1, R. He1, D. Viswanatha1
1Mayo Clinic, Rochester, MN; 2Mayo Clinic, Scottsdale, AZ.
Introduction: Multiple myeloma (MM) is cytogenetically classified by hyperdiploid
and translocation subgroups. Next generation sequencing (NGS) studies have
recently revealed recurrent single nucleotide or insertion/deletion (indel) genetic
changes with potential prognostic and/or therapeutic significance in MM. We
describe a 61 gene targeted NGS assay for use in newly diagnosed or treatmentrefractory MM patients. Methods: FACS-sorted bone marrow plasma cells (5 X105)
from MM patients were obtained and DNA extracted. 200 ng of sheared adapterligated DNA was used for targeted hybrid capture library preparation (Agilent
SureSelect, Santa Clara CA). NGS was performed on Illumina instruments (Illumina,
San Diego CA) and sequence data processed using a custom analysis pipeline
(Mayo NGS Workbench). Final variant classification and reporting was completed by
a dedicated NGS analytical team. Variant calls were confirmed using an orthogonal
NGS method (Ion Proton, ThermoFisher, Waltham, MA). Results: A minimum 250X
depth of coverage (DOC) was obtained except for some GC-rich or low complexity
regions that were masked from analysis. Accuracy samples included NA12878 and
NA19240 cell lines and 51 previously sequenced (Ion Torrent PGM) MM samples
from Mayo Clinic, AZ (MCA). The capture-based/Illumina platform detected all
variants (100% concordance) with similar clonal variant allele fractions (VAF) in
these samples; however, additional variants were identified in 13/51 patient samples
by the capture-based/Illumina protocol (confirmed by orthogonal Ion Proton NGS),
mainly resulting from better DOC, enhanced bioinformatics detection of indel events,
and reduced sequence artifact (e.g. homopolymers). Reproducible analytical
sensitivity was 5% VAF from mixing experiments of DNA from MM patient samples
with NA12878 cell line. Dilute long oligonucleotide controls with known variants
spanning all regions of interest were used for ongoing assay performance
monitoring. Conclusions: We describe a robust and sensitive NGS application to
evaluate new onset or relapsed/refractory MM for gene mutations of prognostic
value, as well as targets associated with sensitivity to immunomodulatory imide
drugs (IMiDs) or proteasome inhibitors (PIs). Enhanced prognostic and therapeutic
evaluation of MM is desirable beyond the current cytogenetic classification in order
to identify patients at higher risk of progression, or those requiring a change in
therapeutic management. This panel is comprehensive yet readily applicable for
efficiently screening multiple genetic targets in a relatively rapid and cost-effective
platform. High purity cell sorting eliminates non-tumor cells and can be combined
with copy number analysis to determine true VAF and clonal hierarchies in MM.
H34. Implementation Considerations: Designing and Medically Vetting a
Targeted Gene Panel for Hematologic Malignancies
N. Sidiropoulos, M.L. Grant
University of Vermont Medical Center, Burlington, VT.
Introduction: Genomic analysis is increasingly included in the clinical care of
hematologic malignancies (HM). While vendor-ready assays (VRA) are relatively
efficient to implement, they often contain targets that are not clinically relevant and
relevant targets not adequately covered. Therefore, designing and medically vetting
H36. Comparison of a MALDI-TOF-based SNP Panel with STR Analysis for
Chimerism Testing
Y. Linnik1, S.F. Allen1, L.J. Cook1, J.L. Kilbourn1, C.M. Studwell1, A.A.
Nakorchevsky2, E.Y. Loo1, G. Tsongalis1, J.A. Lefferts1
1Dartmouth-Hitchcock Medical Center, Lebanon, NH; 2Agena Bioscience, San Diego,
CA.
Introduction: Chimerism testing in hematopoietic stem cell transplant (HSCT)
patients has become a standard of care to measure the degree of engraftment at
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
various time points post-transplant. This is accomplished by quantitatively
genotyping panels of polymorphic DNA loci, generally short tandem repeats (STR),
in which the donor and recipient have different genotypes. SNP loci can also be
targeted for chimerism testing but a larger number of SNP loci are required to ensure
an adequate number of informative loci. The needed degree of multiplexing makes
chimerism testing with SNPs impractical with most SNP genotyping technologies but
the MALDI-TOF-based MassARRAY System with iPLEX technology (Agena
Bioscience) is capable of performing highly multiplexed SNP genotyping in a highly
quantitative manner. To evaluate the potential use of MassARRAY SNP
genotyping for clinical chimerism testing in HSCT patients we compare the
performance of the MassARRAY 92 SNP ChimericID Panel to standard STR
chimerism testing currently used in our laboratory. Methods: Contrived mixtures of
residual DNA samples from 2 related individuals (siblings) and 2 unrelated
individuals were made to mimic post-transplant samples representing 19 different
concentrations between 0% “donor” and 100% “donor.” “Engraftment” (% donor
DNA) was measured using the MassARRAY 92 SNP ChimericID Panel and standard
STR chimerism testing methods previously validated in our laboratory. Additionally,
14 residual DNA samples representing a wide range of results from previous STR
analysis were tested using the 92 SNP ChimericID Panel. Results: When compared
to the expected results for the contrived sets of mixed DNA samples, both
MassARRAY and STR techniques demonstrated excellent performance in terms of
linearity. Following linear regression analysis, R2 values exceeding 0.99 were
obtained for MassARRAY and STR with both related and unrelated sets of contrived
DNA mixtures. The results for the 14 clinical samples were also highly concordant
between the 2 methods with results within 7% of each other for all samples.
Conclusions: Chimerism testing using a highly multiplex MassARRAY SNP panel
produced results that were highly concordant with standard STR chimerism testing.
Although a more complete validation study would be required, these results suggest
that SNP-based chimerism testing using the MassARRAY System would be a
suitable alternative to STR analysis.
analyzing each PCR reaction. The results are independent of a calibration curve and
can increase the precision of quantification. The t(9;22)(BCR-ABL1) translocation is
associated with chronic myelogenous leukemia (CML) and a subset of acute B
lymphoblastic leukemias (B-ALL). Several studies have demonstrated the clinical
utility of quantifying BCR-ABL1 fusion transcript levels for detecting minimal residual
disease (MRD). Quantitative RT-PCR is a well-established method for MRD
monitoring but has inherent limitations with regard to lower limit of detection (LOD)
and limit of quantification (LOQ). Here we report the results of a pilot project to
evaluate the potential advantages of ddPCR in detecting and monitoring BCRABL1 e1a2 transcripts in patients with B-ALL. Methods: ddPCR assays for BCRABL1 e1a2 and ABL1 were developed and optimized using 2ug input RNA. 8 wells
of BCR-ABL1 e1a2 and 2 wells of ABL1 were analyzed in parallel per sample.
Droplets were obtained with a QX100 droplet generator, PCR products were run on
a QX100 droplet reader and analyzed with QuantaSoft software. ddPCR droplets
were designated positive or negative based on a fluorescence amplitude threshold.
Quantification of the target concentration was determined as BCR-ABL1 copies per
microliter and BCR-ABL1/ABL1 ratio. Results: Analytic performance was assessed
using the Invivoscribe BCR-ABL1 e1a2 RNA dilution set. By combining the counts of
8 wells, the ddPCR assay consistently demonstrated a LOD of 10-6 dilution and the
LOQ was highly linear down to 10-5 dilution. The assay reliably detected <10
copies/sample and a BCR-ABL1/ABL1 ratio of <0.005. Using the SUPB15 cell line
we also established a minimum amount of input RNA that can be reliably quantified
as 0.001ng. A cohort of 10 BCR-ABL1 positive B-ALL patients is currently being
evaluated using ddPCR at time points day 0 (diagnostic sample), and days 33 and
78 post-induction for MRD determination. The performance of ddPCR will be
compared with standard qPCR along with clinicopathologic correlation.
Conclusions: In this pilot project we developed and optimized a ddPCR test for
monitoring MRD in BCR-ABL1 positive B-ALL patients. This ddPCR assay
demonstrates robust performance characteristics and may offer improved detection
of low-level disease and lead to earlier detection of relapse or persistence of MRD.
H37. Performance Evaluation of a Novel, Rapid, Multiplexed, One-Step RT-PCR
Assay for Simultaneous Detection of Common Leukemia-Associated
Translocations
S.S. Talwalkar, J. Barry
CPA Lab / Norton Healthcare, Louisville, KY.
Introduction: Rapid and effective determination of common translocations
associated with acute myeloid (AML) & B-lymphoid (B-ALL) leukemias is critical in
order to guide treatment. This is especially true for acute promyelocytic leukemia
(APL) where specific treatment can be initiated upon comfirmation of PMLRARA translocation. Enterogen Inc. has recently launched multiplex RT-PCR assay
for detection of most common translocations seen in AML & B-ALL. This assay
detects 11 variants in 6 common translocations: AML1-ETO, CBFB-MYH11, MLLAF4, TEL-AML1, E2A-PBX1, and PML-RARA. This assay also includes probes for
the most common BCR-ABL1 translocations: b2a2, b3a2 and e1a2. Objective of this
study was to test the performance characteristics of this assay in a mid-sized clinical
reference lab. Methods: Total RNA extracted from blood (N=17) or bone marrow
(N=11) in 28 patient samples (AML=15, B-ALL=5, CML=6, APL=2) and 9 synthetic
fusion gene controls were tested. Total RNA input per reaction varied from 254640ng. 9 synthetic fusion gene controls of known allelic frequency were diluted with
HL60-derived total RNA for limit of detection studies, up to 1.1%. This assay
comprises of 2 wells, each multiplexing 4 reactions, with ABL1 serving as an
amplification control. Positive samples from each well were selected and tested in 10
separate reactions to evaluate well to well variability of both the control and target
probes. Run to run variability was also evaluated by selecting a positive and negative
sample to be tested on separate dates and validation runs. The results were
compared with the Asuragen multiplex assay (discontinued by the manufacturer).
Results: There was 100% correlation between the 2 assays with 16 of the 28
samples showing a positive result (AML1-ETO=3, CBFB-MYH11=3, MLL-AF4=1,
TEL-AML1=2, E2A-PBX1=1, PML-RARA=2, BCR-ABL b3a2=3, BCR-ABLe1a2=1).
All 9 synthetic controls were detectable at 1.1% allelic frequency. All samples
showed 100% correlation in terms of reproducibility and run to run variability. There
was minimal variation of cycle threshold (CT) between runs with optimum
amplification achieved at inputs around 500ng per reaction. The total run time for this
assay was 65 minutes with average hands-on time of 15 minutes (excluding RNA
extraction). Conclusions: The multiplexed RT-PCR assay is a rapid and costeffective method compared to FISH panels to screen for the most common
translocations seen in acute leukemias. Being a one-step RT-PCR format, it reduces
hands-on time and has capability to provide fast turn-around time especially in cases
suspicious for PML-RARA translocations. It also provides results for BCR-ABL1
fusions with specific breakpoint information thereby helping in triaging for specific
quantitative assay(s).
H39. Multi-Year Review of Cytogenetic Abnormalities in Patients with Multiple
Myeloma from a Single Institution and a Proposed Testing Algorithm
P. Paulraj1,4, B. Hilton1,4, J. Herriges2, M. Serrano3, B. Hong1, E.F. Andersen1, A.N.
Lamb1, X. Xu1, M. Salama1, R.M. Toydemir1
1University of Utah, Salt Lake City, UT; 2Children's Mercy Hospital, Overland Park,
KS; 3Lineagen, Salt Lake City, UT; 4ARUP Laboratories, Salt Lake City, UT.
Introduction: Multiple myeloma (MM) is characterized by heterogeneous
cytogenetic abnormalities of variable prognostic significance. Recurrent findings
such as hyperdiploidy and t(11;14) confer a better prognosis while gain of CKS1B,
t(4;14), t(14;16), and TP53 deletion predict poor outcomes. Chromosome analysis is
recommended per NCCN guidelines, however obtaining metaphase preparations
from plasma cells is challenging. Therefore FISH analysis on CD138+ cells is often
used for detection of recurrent abnormalities in MM, including cytogenetically cryptic
rearrangements. Genomic microarray analysis on CD138+ samples is also gaining
popularity as a diagnostic tool, as it does not require cell culture. Methods: In this
retrospective study, we reviewed cytogenetic, morphologic, and flow cytometry data
for diagnostic and follow-up samples to evaluate the clinical utility of various
cytogenetic techniques in the workup of patients with plasma cell disorders. Our
cohort included 2706 encounters of 921 patients from a single institution over a
period of 8 years. For patients with multiple encounters, we also evaluated new
abnormalities detected on follow-up specimens. Results: Of the 2706 specimens
analyzed, a clonal abnormality was seen in 359 (13.3%) cases by chromosome
analysis. Hyperdiploidy was most common (16.7%), followed by 1q gain (15.6%).
When performed on unsorted cells, FISH was positive in only 4.5% of cases,
compared to a 67.5% abnormality detection rate from CD138+ cells. Hyperdiploidy
involved chromosome 9 in 46% of the cases, chromosome 11 in 32%, and
chromosome 15 in 43%. Rearrangements of the IGH locus at 14q32 were seen in
9.1% of the specimens by chromosome analysis, whereas IGH rearrangements were
seen in 41.7% of cases by FISH. The t(11;14) was the most frequent structural
abnormality, seen in 32.4% of cases. Ten samples were further analyzed on a SNPbased genomic microarray platform. Microarray data were concordant with results
obtained using conventional cytogenetic techniques in all cases, and included
detection of hyperdiploidy, 17p deletion and 1q duplication. Additional microarray
findings included copy number changes not targeted by FISH, as well as loss-ofheterozygosity in 6/10 samples. Conclusions: Cytogenetic analysis is a critical
component of the diagnostic and follow-up studies of MM. Based on our findings we
propose a diagnostic algorithm which includes genomic microarray analysis
complemented by FISH analysis on CD138+ cells with probes targeting IGH
rearrangements, 1q gain, and 17p loss. In particular, we provide evidence supporting
the clinical utility of genomic microarray analysis in follow-up specimens with low
plasma cell percentages.
H38. Development of a Droplet Digital PCR Assay for Detection and
Quantification of BCR-ABL1 e1a2 Fusion Transcripts in Acute B
Lymphoblastic Leukemia
P. Mroz1, Q. Kang1, N.G. Bailey2, N.A. Brown1, M. Tewari1, B.L. Betz1
1University of Michigan, Ann Arbor, MI; 2University of Pittsburgh, Pittsburgh, PA.
Introduction: Droplet digital PCR (ddPCR) is a novel PCR platform that provides
absolute quantification of amplification targets by partitioning and separately
H40. ALL-ICP, a Simple and Comprehensive Method to Detect Chromosome
Abnormalities in Acute Lymphocytic Leukemia
R. Babu1, D. Van Dyke2, E. Fuentes1, S. Fuentes1, S. Papa1, S. Kopuri1, C.
Williamson2, P. Koduru3
1InteGen LLC, Orlando, FL; 2Mayo Clinic, Rochester, MN; 3University of Texas
Southwestern Medical Center, Dallas, TX.
Introduction: B- and T-cell acute lymphocytic leukemia (ALL) affects children and
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
963
AMP Abstracts
adults. Cytogenetic and recent molecular genetic findings at diagnosis constitute
important, independent prognostic factors in both age groups. Due to the typically
poor morphology of the chromosome preparations, conventional cytogenetic analysis
can miss one or more of the recurrent abnormalities and most ALL FISH panels only
target select abnormalities. Hence, a simple, fast, and comprehensive improvement
to karyotype analysis would greatly enhance the capabilities of diagnosis and
research for ALL. Methods: We recently developed and validated a novel molecular
technology - Interphase Chromosome Profiling (ICP) (Babu et al., in press). We
tested a variation (ALL- ICP) of our original probe set, for interrogating interphase
cells and metaphase chromosomes in ALL. The design consisted of a multiplex
approach with analysis of 6 chromosomes per hybridization site. Each telomere and
pericentromeric region is targeted and the resulting fluorescent signals are spectrally
distinct and easily recognizable from each other. After overnight hybridization, up to
20 metaphase spreads are analyzed for each chromosome. When 20 metaphases
are unavailable, interphase nuclei are used to complete the 20-cell analysis. Thirtytwo ALL samples were studied blindly by the laboratory that developed the
technology. Two institutions provided cell pellets with a variety of known cytogenetics
and FISH results. Results: ALL-ICP detected all numerical and structural
abnormalities except interstitial deletions and duplications. More importantly, it
identified balanced and unbalanced translocations missed by cytogenetics and
standard ALL FISH probes including characterization of derivative chromosomes.
These included t(12;21), t(5;10), variant t(9;22), der(19)t(1;19), der(3)t(3;13),
idic(17q), idic(10q), der(14)t(14;21), der(4)t(4;6), der(6)t(5;6), and der(15)t(3;15).
Terminal deletions (TD) identified only by ALL-ICP included 10p, 15q, 16p, 16q, 18p,
21q, 22q (proximal), and a novel recurrent terminal 14q deletion. A 12p TD was
detected by ALL-ICP and FISH, but not by conventional cytogenetics. Work is in
progress to add probes for 3 recurrent deletions (6q21, 9p21, 13q14) and one
duplication/amplification (21q22), and CRLF2 rearrangement to the ALL-ICP probe
set. The average analysis time was less than 40 minutes per case when sufficient
metaphases were available. Conclusions: ALL-ICP is a technically simple method
that saves valuable technologist’s time, and with the added probes will detect all
clinically relevant recurrent B- and T-cell chromosome abnormalities. These
attributes supported by the study results, make ALL-ICP ideal for the genetic
diagnostic workup of ALL.
H41. Multi-Platform-Based Comprehensive Molecular Analysis of
Hematological Malignancies for Somatic Mutations, Copy Number Changes
and Translocations for Routine Clinical Testing
R. Kanagal-Shamanna, R. Luthra, M.J. Routbort, R. Singh, K. Floyd, R. Abraham, S.
Hai, C. Yin, Z. Zuo, C.Y. Ok, S. Loghavi, L. Medeiros, K.P. Patel
M.D. Anderson Cancer Center, Houston, TX.
Introduction: Assessment for somatic mutations, copy number (CN) changes and
recurrent translocations is crucial for diagnosis, risk-stratification and targeted
therapeutics. While targeted NGS-based mutation analysis has the advantage of
higher sensitivity, it is suboptimal to assess genome-wide CN changes. Similarly,
identification of translocations by NGS requires additional RNA-based analysis. We
present our multi-platform approach for comprehensive molecular analysis for workup of hematologic malignancies (HM). Methods: Somatic mutations were assessed
using a targeted NGS 81-gene panel. Sequencing libraries were prepared using
hybridization capture-based enrichment of genomic regions of interest (Agilent
Haloplex HS) followed by bi-directional paired-end sequencing (Illumina MiSeq) for
single nucleotide variants and insertions/deletions (≤52 base-pairs) [analytical
sensitivity (AS), 5%]. Standard procedures were used for data analysis and variant
interpretations. CN changes were assessed using array-based comparative genomic
hybridization (aCGH; Agilent, 4x180K) [AS, 20%]. In patients with ≥6% blasts, a
qualitative nanofluidics-based multiparametric reverse-transcriptase PCR (leukemia
translocation panel, LTP) interrogated 14 recurrent translocations in 9 WHO-defined
leukemia subtypes. Cases also underwent conventional karyotyping (CK), FISH
studies and capillary electrophoresis for FLT3 ITDs. Results: A preliminary cohort of
23 patients with suspected HM underwent comprehensive testing. NGS identified 28
somatic mutations in 16 patients; 7 were potentially targetable involving NRAS,
KRAS, IDH1 and FLT3 genes. aCGH identified 21 CN changes (3 not detected by
CK) in 9 patients. In 1 patient with suspected MDS with no growth by CK, aCGH
showed absence of CN changes. Molecular results facilitated accurate diagnosis and
sub-categorization using 2016 WHO: 10 AML [3 with PML-RARA, 1 with t(8;21), 1
with inv(16), 1 with mutated NPM1, 1 with double CEBPA mutations, 1 therapyrelated and 2 AML NOS], 3 B-lymphoblastic leukemia/lymphomas [2 with t(9;22)], 4
myelodysplastic syndrome and 1 chronic myelomonocytic leukemia. In 5 patients
with suspected HM but no morphologic support, absence of molecular alterations
helped clinical management. In 1 AML patient, inv(16) transcript was incidentally
detected by LTP (qPCR, transcript:ABL1 ratio>100), while karyotype and FISH were
negative. A 139 kb loss in 16p13.1 region involving MYH11 gene was noted by
aCGH. Conclusions: Routine clinical testing of HMs using multiple modalities for
somatic mutations, CN changes and translocations facilitates timely and appropriate
clinical management. Efforts to consolidate into a single platform to improve
efficiency and synchronize results are underway.
964
H42. Correlation between Calreticulin (CALR) Mutations as Detected by PCR
and CAL2 Antigen Expression by Immunohistochemistry
E. Wolak1, K. Buehler2, J. Gale2, T. George3, M.A. Vasef3
1University of New Mexico, Albuquerque, NM; 2Tricore Ref Laboratory, Albuquerque,
NM; 3University of New Mexico/TriCore Ref Laboratory, Albuquerque, NM.
Introduction: Frameshift mutations of the Calreticulin (CALR) generate a novel
peptide at the C-terminus region, leading to hyperactivity of JAK/STAT signaling
pathway. Up to 80% of JAK2 and MPL unmutated myeloproliferative neoplasms
(MPNs) have CALR mutations. Therefore detection of CALR mutations is useful to
confirm the diagnosis of these MPNs. A recently developed CAL2 antibody reacts to
this novel epitope, recognizing the most common CALR mutations: 52 base pair
deletion and 5 base pair insertion. However, no study has yet documented how the
CAL2 antibody reacts with other CALR mutations that result in frameshifts other than
1+ bp. Methods: A total of 18 bone marrow biopsies of previously characterized
non-CML MPNs were selected for Immunohistochemistry using the CAL2 antibody
(DIA-CAL-250). A known CALR mutated specimen served as an internal positive
control and a 50-tissue microarray construct of non-MPN cases served as negative
control. For molecular analysis of CALR exon 9 mutations, the extracted DNA was
subjected to PCR amplification of the exon 9 of CALR gene followed by capillary
electrophoresis for sizing of PCR products. Results: CAL2 expression was detected
in 12 out of 12 specimens with molecularly confirmed CALR mutations including 52
base pair deletion and 5 base pair insertion. CAL2 was negative in 50 out of 50
negative control specimens from patients with either no disease state or
with CALR unmutated myeloproliferative neoplasms. There was no difference in
staining between clot and decalcified biopsies in the same patients, and no apparent
difference between formalin fixation/Immunocal decalcification and AZF fixation/rapid
decalcification specimens. However, CAL2 was negative in 2 out of 2 specimens
from a known CALR mutated patient with a 23 base pair deletion. Further Sanger
sequencing confirmed a 23 bp deletion with 2+ bp frameshift mutation in these
specimens. Conclusions: The CAL2 antibody shows reproducible positive staining
in CALR mutated specimens with 1+ bp frameshift mutations. However, CAL2
antibody will fail to detect CALR indels that lead to 2+ bp frameshift including cases
with 23 or 8 bp deletions. Although CAL2 immunohistochemistry is useful for
detecting common CALR mutations that lead to 1+ base pair frameshift, it misses a
subset of mutations that lead to 2+ bp frameshift. Therefore, laboratories that
screen CALR mutations by this antibody should reflex all negative cases to
molecular analysis for CALR variant mutations that will be missed by CAL2
immunostaining.
H43. Validation of a Low-input, Amplicon-Based Dual-Strand Assay to Detect
DNA Variants in Lymphomas by Next-Generation Sequencing (NGS)
A. Oran, A. Bigdeli, J. Morrissette, R.T. Sussman
University of Pennsylvania, Philadelphia, PA.
Introduction: The use of Next-Generation Sequencing (NGS) for the detection of
somatic DNA variants with either prognostic or therapeutic significance in tumors has
become a critical step in developing a treatment plan for both hematological and
solid tumors. However, these targeted-sequencing panels typically miss genes with
prognostic or therapeutic value for lymphomas. To provide an assay that meets
these needs, we sought to validate a targeted sequencing panel that could be used
with low-input amounts of DNA extracted from Formalin-Fixed Paraffin-embedded
tissue (FFPE), blood or fresh tissue. Methods: DNA from 28 lymphomas (23 blood, 6
FFPE, 1 fresh tissue) that had previously been sequenced using a 68-gene true-seq
custom amplicon heme-oncology panel from Illumina were made into libraries using
a dual-strand 40-gene, low-input true-seq custom amplicon lymphoma-focused panel
from Illumina. Fourteen genes were covered on both panels and used to validate the
lymphoma panel. For the dual-stranded assay, 25ng of DNA from each sample was
prepared into 2 libraries (strand A and strand B) and sequenced as separate
samples on the same run and analyzed separately. Variant Call Files (VCFs) were
merged and only variants present in both strands with an allele frequency above 3%
were considered true variants. Results: Of the 28 samples tested, 24 samples
passed quality control (QC, 20 blood, 3 FFPE, 1 fresh tissue). Within the 24
samples, there were 60 known variants in the 14 genes that were covered by both
the heme and the lymphoma focused panels. From these, there was a 98%
concordance in the detection of variants between the 2 panels (one was previously
called at 4.5% and was present in the BAM files from the lymphoma panel, but fell
below 3% in one of the 2 strands) and a 97% concordance in variant allele frequency
between the 2 panels. Conclusions: The lymphoma community is currently
underserved by clinical NGS oncology services. Here we present data supporting the
use of a low input, dual-strand assay to detect variants in lymphoma samples that is
agnostic to tissue type. We demonstrate that this assay is able to detect variants that
were previously detected by an orthogonal assay that required 10x the amount of
DNA. Together, these data aim to provide lymphoma clinicians with a new assay to
aid in the therapeutic decisions made for the care of their patients.
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
H44. HDAC6 Regulates MicroRNA-27b that Suppresses Proliferation, Promotes
Apoptosis and Target C-MET in Diffuse Large B-Cell Lymphoma
Y. Jia, Z. Liu, X. Zhou
Fudan University Shanghai Cancer Center, Shanghai, China.
Introduction: Diffuse large B-cell lymphoma (DLBCL) is the most common type of
non-Hodgkin lymphoma. HDAC6 is frequently altered in DLBCL and inhibition of
HDAC6 has potent anti-tumor effects in vitro and in vivo. Methods: We profiled
miRNAs that altered in the HDAC6 knockdown DLBCL cells with NanoString
nCounter assay and identified miR-27b as the most significantly increased
miRNA. Results: We validated decreased expression of miR-27b in DLBCL tissues,
and we found that low expression of miR-27b was associated with poor overall
survival of patients with DLBCL. In addition, forced expression of miR-27b
suppressed DLBCL cell viability and proliferation in vitro, and inhibited tumor growth
in vivo. Mechanistically, Rel A/p65 is found to negatively regulate miR-27b
expression, and its acetylation and block of nuclear translocalization caused by
HDAC6 inhibition significantly elevates miR-27b expression. Furthermore, miR-27b
targets c-met and thus represses the c-met/PI3K/Akt pathway. These findings
highlight an important role of miR-27b in the development of DLBCL and uncover a
HDAC6-Rel A/p65-miR-27b-MET signaling pathway. Conclusions: Elevating miR27b through HDAC6 inhibition would be a promising strategy for DLBCL treatment.
The resulting libraries were sequenced on an Illumina HiSeq 4000 or MiniSeq
instrument using 2x150bp paired end reads to mean depths of 992x (MiniSeq) to
6,785x (HiSeq) across the FLT3 gene. Sequence data were trimmed to remove
adapter sequences, aligned to the b37 reference genome using BWA mem (v0.7.15)
and analyzed for the presence of FLT3 ITD mutations using a de novo assembly
approach, Pindel (v0.2.4w), and the Platypus variant caller (v8.1). Results: All 8
cases of AML with known FLT3 ITD mutations were correctly identified using the de
novo assembly approach and pindel. Platypus identified an insertion in one case with
an ITD of 21 bp; larger ITD alleles were not detected. Both assembly and pindel
failed to detect multiple ITD alleles in 2 cases where multiple insertions were
identified by PCR and capillary electrophoresis (CE), and ITD allele sizes from NGS
were discordant with CE analysis in 2 of 3 samples where CE data were available.
ITD mutations were not detected in any of the 10 non-FLT3 ITD controls by any of
the 3 methods. Conclusions: Using a custom amplicon-based assay incorporating
UMIs we identified the FLT3 ITDs in 8 of 8 (100%) AML samples with 100%
specificity in 10 negative controls. ITD allele size and number appear difficult to
resolve using the methods tested. However, our approach reliably detected the
presence of ITD mutations, while also identifying other AML variants in a single
assay. These results demonstrate the ability to detect FLT3 ITDs using errorcorrected amplicon-based sequencing panels.
H45. Clonality Detection Using Next-Generation Sequencing and Capillary
Electrophoresis Methods in Suspect Lymphoproliferative Samples
Y. Huang1, K. Hutt1, J. Panganiban1, A. Jacobsen1, N. Wong1, P. Shah1, S.
Zheng1, Z. Xie1, R. Bob2, T. Stenzel1, J.E. Miller1
1Invivoscribe Technologies, San Diego, CA; 2Institute for Pathodiagnostik, Berlin,
Germany.
Introduction: Detection of clonal rearrangements within the immunoglobulin (Ig) and
T-cell receptor (TCR) genes in clinical specimens is a finding used to assist in
diagnosis of lymphoproliferative disease (LPD). Capillary electrophoresis (CE)
methods have been used to detect clonality; however, next-generation sequencing
(NGS)-based assays can be more powerful, as they both identify unique clonal
Ig/TCR rearrangements and identify the tumor-specific V-J DNA sequences
necessary to track clonal sequence in follow-up testing. Here we report test results
from 59 clinical samples using LymphoTrack Dx (IGH (FR1/2/3), IGK, TRG and TRB)
assays on MiSeq, CE-based assays, and samples tested on the Ion PGM platform
(IGH and IGK). Methods: LymphoTrack Dx Assays targeting
the IGH, IGK, TRG and TRB loci have been developed for the MiSeq and Ion PGM
platforms. Consensus primers targeting V and J gene segments include both
platform-specific adapter sequences and individual barcodes, so the PCR products
from multiple independent samples and multiple clonality assays can be combined
and sequenced together in a single NGS run. Individual master mixes were
manufactured with 24 indices (MiSeq) or 12 indices (Ion PGM). Single step PCR
amplification of 50 ng DNA input was followed by pooling of equimolar amounts of
purified amplicons, and then loaded onto the sequencing machine. The sequencing
data was analyzed using LymphoTrack Dx Software, to generate frequency
distributions, determine V-J usage, identify specific sequences for top sequencing
reads, and determine the somatic hypermutation rate of IGH FR1 amplicons.
Results: All LymphoTrack Dx assays were validated for analytical performance
using dilutions of contrived samples with known V-J rearrangements. The assays
demonstrated excellent linearity (R2>0.90), sensitivity to detect 2.5% clonality, and
reproducibility (<20% CV). Concordance between the LymphoTrack Dx MiSeq or Ion
PGM assays and the CE assays demonstrated greater than 90% concordance for all
targets. Concordance between the LymphoTrack Dx Ion PGM and MiSeq assays
was greater than 95% for IGH and IGK targets. Conclusions: This pilot study has
demonstrated that the comprehensive NGS LymphoTrack Dx Assays have the
potential to be utilized for routine Ig/TCR clonality detection in LPD diagnosis.
Furthermore, the NGS assays can identify clonal V-J rearrangements and provide
the clonal DNA sequences of the tumor-specific clonotypes required to perform
follow up testing to detect residual disease. Combining Ig/TCR assays within one
NGS run can improve the overall clonality detection rate and reduce the cost.
H47. Clinical Utility of Semiconductor Based Next Generation Sequencing for
Evaluation of IgVH Somatic Hypermutation Status in Chronic Lymphocytic
Leukemia / Small Lymphocytic Lymphoma (CLL/SLL)
B. Tandon, C. Porter, A. Layton
Molecular Pathology Laboratory Network, Inc., Maryville, TN.
Introduction CLL/SLL is the most common leukemia diagnosed among adult
patients. IgVH Somatic hypermutation (SHM) status is one of only 5 recommended
prognostic indicators included in the recently defined CLL International Prognostic
Index for baseline risk stratification (Lancet Oncol. 2016 Jun;17(6):779-90). Given
the highly polymorphic nature of somatically recombined heavy chain VDJ segments
at IGH, and given that semiconductor based NGS has been associated with
relatively higher error rates in base calling compared to other methodologies, we
sought to investigate the utility of semiconductor based NGS testing for routine
clinical laboratory evaluation of IgVH SHM status. Methods: Twenty peripheral
blood/marrow aspirate samples from newly diagnosed CLL/SLL patients were
studied. Samples were subjected to semiconductor based NGS testing on the Ion
Torrent Personal Genome (PGM) system, Illumina based sequencing, and rt-PCR /
Sanger sequencing. NGS and Sanger sequencing utilized DNA and RNA,
respectively. PGM sequencing involved pooling and quantification of IGH framework
1 V-J amplicon libraries utilizing Invivoscribe panel primers and PGM
adapters. Generated FASTQ files were analyzed by LymphoTrack IGH_SHM_PGM
Software (Version 1.0). VJ-utilization profiles and SHM results were compared
among the 3 assays. Results: PGM sequencing demonstrated 100% concordance
with Sanger Sequencing and Illumina MiSeq data for VJ-utilization profiles and rates
of SHM. In our hands, a limit of detection of <3% of total reads was established for
PGM sequencing, and intra- and inter-assay reproducibility was 100% across 3x
replicate testing of 2 patient samples. Conclusions: Semiconductor based NGS
testing demonstrates sufficient accuracy and precision for routine clinical laboratory
evaluation of IgVH SHM status in newly diagnosed CLL/SLL. No significant
differences in SHM rates or VJ utilization profiles were noted among samples tested
by the 3 methodologies included in this study. NGS, whether semiconductor or
Illumina based, confers significant advantages over the current standard Sanger
sequencing methodology. Specifically, utilization of DNA in NGS assays eliminates
many of the practical challenges associated with RNA based testing, such as
specimen lability and necessity of rapid specimen receipt by the clinical
laboratory. Furthermore, NGS assays facilitate batch multiplex testing of patient
samples, streamlining laboratory workflow and facilitating reductions in labor, costs,
and overall turnaround times for clinical reporting. The NGS assay limit of detection
is also significantly lower than that of Sanger sequencing, potentially allowing for
generation of technically robust results in patient samples with less than 15-20%
tumor cellularity.
H46. NGS Based Identification of FLT3 ITD Mutations Using Unique Molecular
Indexes
B.A. Parikh, E.J. Duncavage, D.H. Spencer
Washington University School of Medicine, St. Louis, MO.
Introduction: Acute myeloid leukemia (AML) is defined by a core group of recurrent
gene mutations, many of which are associated with differences in prognosis. One of
the most common clinically relevant mutations in AML is the internal tandem
duplication (ITD) in FLT3 gene. The FLT3 ITD alleles range in size from 15bp to
300bp, and are notoriously difficult to detect in NGS data. While FLT3 ITDs can be
detected hybrid capture based NGS data, is unclear whether similar performance
can be achieved using amplicon-based enrichment approaches that incorporate
unique molecular identifiers (UMIs) for improved sequencing error correction. Here
we describe the use of the UMI-based Agilent HaloplexHS to detected FLT3 ITDs.
Methods: We selected a series of 8 AMLs with known FLT3 ITD mutations ranging
from 21bp to 108bp and 10 control cases without FLT3 ITD mutations. DNA was
isolated from either whole blood or bone marrow aspirates. Haloplex HS probes
incorporating 10bp degenerate oligonucleotide UMIs were designed to target both
DNA strands of all FLT3 exons along with 24 other recurrently mutated AML genes.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
H48. Unusual Cases of MDS/MPN-RS-T Without Overt Anemia Share Molecular
Signatures Classic for MDS and MPN Overlap Syndromes
P. Li1, N.M. Shahmarvand2, T. George3, D. Arber4, R. Ohgami2
1Baylor College of Medicine, Houston, TX; 2Stanford University, Stanford,
CA; 3University of New Mexico, Albuquerque, NM; 4The University of Chicago,
Chicago, IL.
Introduction: Myelodysplastic and myeloproliferative neoplasms with ring
sideroblasts and thrombocytosis (MDS/MPN-RS-T) is a rare independent disorder
and refractory anemia is essentially necessary for the diagnosis in the revised 2016
WHO classification. Here we reported unusual MDS/MPN-RS-T cases without
apparent anemia (MDS/MPN-RS-T-NA) with molecular signatures classic for typical
MDS/MPN-RS-T cases, extending the spectrum of this disorder beyond overt
refractory anemia with ring sideroblasts and thrombocytosis. Methods: We reviewed
documented clinical details, pathologic and cytogenetic features, and reported JAK2
V617E mutations in these 4 somewhat unusual MDS/MPN-RS-T cases. Further
Sanger sequencing targeting the hotspot mutations of SF3B1 exon 12 to 16 was
performed following PCR amplification. The clinicopathologic features and molecular
965
AMP Abstracts
signatures were compared to those in near 500 typical MDS/MPN-RS-T cases by
review of literatures. Results: Patients without apparent anemia manifested at a
younger age (55.5 year in MDS/MPN-RS-T-NA versus 71-75 years in MDS/MPNRS-T) and persistent thrombocytosis was the only abnormal laboratory finding. As in
typical MDS/MPN-RS-T-NA, morphologic features characteristic of both
myeloproliferative neoplasms and myelodysplasia were present. Further, 3 of 4
(75%) patients with MDS/MPN-RST-NA carried a somatic mutation of JAK2 V617E,
whereas at least one SF3B1 mutation was present in all patients, and the molecular
signatures were similar to those reported in typical MDS/MPN-RS. Moreover, almost
all SF3B1 mutations in MDS/MPN-RST regardless anemia were missense
substitutions in coding exons 12 to 16 and the mutation hotspots were located at the
same 22 non-identical tandem HEAT repeat domains of C-terminal region.
Conclusions: MDS/MPN-RS-T-NA represents the most “benign” end of
the spectrum of MDS/MPN-RS-T and anemia might not be required for the diagnosis
of MDS/MPN-RS-T. These unusual cases might interest pathologists and clinicians
due to the rarity of this disorder, the recently revised diagnostic criteria, and the
existence of few cases without overt anemia. Precise classification and further risk
stratification by clinical and molecular features might provide more rigorous data to
reach a consensus on optimal clinical treatment.
H49. Sensitive CXCR4 Sequencing Using Bridged Nucleic Acid (BNA) PCR
Clamp Technology
K.E. Halverson, M. Mai, D.S. Viswanatha, R. He
Mayo Clinic, Rochester, MN.
Introduction: CXCR4 mutations are identified in approximately 30% to 40%
of Lymphoplasmacytic Lymphoma/Waldenstrom Macroglobulinemia
cases commonly in association with the MYD88 L265P mutation. CXCR4 mutations
in the context of MYD88 L265P, particularly the hotspot mutations c.1013 C>G/A,
are clinically associated with higher IgM level, bone marrow involvement, lower blood
counts, and lower Ibrutinib response rate. We developed a sensitive qualitative test
for CXCR4 mutation detection to complement our MYD88 clinical
assay. Methods: Sanger sequencing was targeted to c.870 to c.1059 of
the CXCR4 gene. To achieve high sensitivity for c.1013 C>G/A mutations, BNA
Clamp technology was used to block PCR amplification of the wild-type (WT) DNA in
this region. A custom 13 base-pair BNA clamp spanning c.1007 to c.1019 was
designed (Bio-Synthesis, Lewisville, TX). The BNA clamp was partnered with Sanger
sequencing to allow high sensitivity PCR amplification and sequencing of the c.1013
C>G/A mutant DNA. Other mutations in the c.870 to c.1059 region were detected by
routine Sanger sequencing without BNA clamping. Results: For the BNA-clamped
Sanger sequencing targeting c.1013 C>G/A, analytical sensitivity studies were
conducted by spiking of the mutant c.1013 C>G/A gBlock oligos (Integrated DNA
Technologies, Coralville, IA) into lab negative control DNA and established a 1%
analytical sensitivity with a minimum DNA input of 20ng. Analytic specificity was
evaluated on the WT oligo, lab negative control DNA and 20 normal donor samples,
demonstrating a 100% specificity. Accuracy evaluated by oligo spiking studies into
lab negative control DNA showed 100% accuracy at mutational levels above 1%.
Accuracy was also assessed on patient samples that tested positive and negative for
c.1013 C>G/A mutations in Sanger sequencing without BNA clamping, confirming all
positive routine Sanger calls and identifying additional positive cases in the routine
Sanger-negative samples, explained by the analytic sensitivity differences introduced
by BNA clamping. Reproducibility and precision of the BNA-clamped Sanger
sequencing were determined at 100%. Conclusions: Traditional Sanger sequencing
covers broad regions with universally recognized 10% to 20% sensitivity. BNA
Clamp technology can be used to block the WT DNA template allowing improved
sensitivity of mutant detection. We successfully combined Sanger sequencing with
BNA clamping to develop a higher sensitivity assay for CXCR4 c.1013 C>G/A
mutation detection.
H50. Del(7)(q22) Resulting From an Unbalanced der(7)t(3;7)(q26;q21)
Generating a CDK6-MECOM Fusion and FLT3 Alterations in Pediatric Acute
Myeloid Leukemia with Myelodysplasia-Related Changes
E.M. Azzato, D. Hedges, S. Foy, S. Newman, S.C. Raimondi, J. Nakitandwe, S.A.
Shurtleff, T. O'Neill, J.M. Klco, R.C. Ribeiro, K.E. Nichols, D.W. Ellison, J. Zhang
St. Jude Children's Research Hospital, Memphis, TN.
Introduction: CDK6-MECOM fusions, t(3;7)(q26;q21), have been reported in acute
myeloid leukemia (AML) and myelodysplastic/myeloproliferative disorders and are
associated with a poor prognosis. This rare alteration can present in a balanced or
unbalanced form; when unbalanced it is interpreted as a del(7)(q22), with the
der(7)t(3;7) presenting as a cryptic lesion. In such cases, the CDK6-MECOM fusion
is often not suspected. Here we report 2 pediatric cases of AML with a CDK6MECOM associated with a del(7)(q22), detected by 3-platform next generation
sequencing (NGS). Methods: As part of a prospective clinical research study,
Genomes for Kids, we are conducting a feasibility study of the clinical impact of
NGS, utilizing an integrative whole-genome (WGS), whole-exome (WES) and
transcriptome (RNA-Seq) analysis on fresh/frozen tumor samples from children with
cancer. Eligible patients also undergo conventional cytogenetic and molecular
testing. Matched tumor/normal samples undergo WGS at 45x, WES at 100x
coverage and RNA-Seq at 20-30x coverage; the assay pipeline integrates genetic
lesions detected by all 3 NGS platforms to cross-validate and characterize single
nucleotide variations (SNVs), short insertions and deletions (indels), structural
966
variations (SVs) including fusions, digital karyotypes, focal copy number variants
(CNVs), and copy neutral loss of heterozygosity (CN-LOH). Results: Over 20
months, among 30 newly diagnosed pediatric AML cases, 2 were classified as AML
with myelodysplasia-related changes (AML-MRC) and an isolated
del(7)(q22). Comprehensive sequencing revealed that both cases had
t(3;7)(q26.2;q21.2), which repositioned MECOM
near CDK6. Increased MECOM expression was observed by transcriptome
analysis. Both cases also had recurrent FLT3 alterations (FLT3 A680V
and FLT3 Y591delinsLRSD). Conclusions: The CDK6-MECOM fusion is rare and
associated with a poor prognosis in AML. Del(7)(q22) in pediatric AML-MRC
represents a heterogeneous group and has not been considered an unfavorable
cytogenetic marker. However, it may actually represent the unbalanced, cryptic form
of CDK6-MECOM. To improve prognostication in these cases, the presence of
del(7)(q22) in AML-MRC warrants reflex testing for CDK6-MECOM fusion
and FLT3 alterations.
H51. Going Beyond MMR to the Analysis of Deep Molecular Response
K. Drafahl, J. Toplin, D. Smith, C. Galderisi
MolecularMD, Portland, OR.
Introduction: Quantification of BCR-ABL transcript levels in blood samples is
routinely used to assess CML patient response to targeted kinase inhibitor (TKI)
therapy. With the approval of front and second line drugs for CML treatment, BCRABL measurements such as major molecular response (MR3) and deep molecular
response (MR4.5) play an increasingly critical role in patient management decisions.
Moreover, timing and depth of these responses have been shown to be predictive of
long term outcomes. In this study, we show the validation and comparison of 2 highly
sensitive methods, droplet digital PCR (ddPCR) and RT-qPCR (MRDx), for the
analysis of MR4.5 using RNA extracted from EDTA and PAXgene Blood RNA tubes.
Methods: Both assays are designed to detect BCR-ABL (e13a2/b2a2 or
e14a2/b3a2) and ABL transcript levels. Total RNA extracted from blood is used as
the template in one-step thermal cycling protocols with BCR-ABL and ABL targetspecific primers in separate wells. The Bio-Rad QX200 and ABI 7500 Fast Dx
systems are used for the ddPCR and MRDx assays, respectively. International Scale
(IS) conversion factors based on the 1st WHO International Genetic Reference
Panel were established for both assays. Limit of blank (LoB), limit of detection
(LoD), precision, linear range, and specificity studies were conducted using diluted
CML patient samples or cell lines. A method comparison of the 2 assays was
performed using 100 CML patient samples for both blood collection tube types.
Results: Both assays met acceptance criteria for all validation studies. The
sequence of the BCR-ABL and ABL PCR amplicons are 100% homologous to the
reference sequences. The LoD of the ddPCR assay is ≤MR5.0 for EDTA tubes and
between MR4.5 and MR4.8 for PAXgene tubes. Based on probit analysis, the LoD of
the MRDx assay is ≤MR5.5 for EDTA tubes and MR4.8 for PAXgene tubes.. The
linear range of the ddPCR assay is MR0.8 - 5.0 for EDTA and MR0.9 - 4.5 for
PAXgene tubes. The linear range of the MRDx assay is MR1.0 - 4.5 for EDTA and
PAXgene tubes. In precision studies, the SD log10 was below 0.25 in the established
linear range for all samples across lot, operator and day for both assays. As
expected, the level of variability was proportional to MR levels. No clinically
significant bias at the clinical decision points was observed when the molecular
response values were compared between the 2 assays for both tube
types. Conclusions: In this study, BCR-ABL quantification assays based on RTqPCR (MRDx) and ddPCR methods were validated for detecting MR4.5 using RNA
extracted from EDTA or PAXgene Blood RNA tubes. These assays provide
important diagnostic tools for monitoring deep therapeutic responses in CML patients
treated with TKI therapies.
H52. Comprehensive Assessment for Structural Rearrangements Using a
Customized Anchored Multiplex PCR-Based Next-Generation Sequencing
Assay Targeting 199 Genes
A.K. Dupuy, P. Sukhadia, K. Mullaney, K. Rios, J. Yao, C. Ho, R. Benayed, K.
Nafa, M.E. Arcila
Memorial Sloan Kettering Cancer Center, New York, NY.
Introduction: Structural chromosome rearrangements are increasingly recognized
as driver genetic alterations in a wide range of hematologic malignancies. The
detection of specific gene fusions has proved essential for personalizing patient
treatment, diagnosis, risk assessment, and monitoring. Fusions can be highly
heterogeneous in their structure, as genes often recombine with multiple partners
and involve variable breakpoint regions. Given this heterogeneity and the increase in
number of clinically relevant fusions being discovered, common methods of detection
such as FISH and qRT-PCR are becoming unsuitable for comprehensive
assessment as they are laborious, have low throughput, and lack scalability. Here
we describe the clinical validation of a targeted RNAseq assay for fusion detection in
hematologic malignancies, which allows for the simultaneous assessment of 199
genes. Methods: Clinical cases submitted for routine targeted fusion assessment by
standard assays (FISH, qRT-PCR, or MSK Archer Heme) were selected based on
presence or absence of gene fusions. Amplification of cDNA fragments was
performed using Archer Anchored Multiplex PCR technology v3, with a custom panel
targeting 199 genes recurrently involved in fusions, nucleotide variants, and
expression level differences in hematologic malignancies. The products were
sequenced on the Illumina MiSeq platform and analyzed with the Archer analysis
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
pipeline v5.0.4. Results were assessed for accuracy of fusion calls and for NGS
clinical passing criteria. Results: A total of 42 samples were analyzed (33 positive, 9
negative). The custom MSK Archer Pan-Heme assay produced expected results for
samples positive for gene fusions including: BCR-ABL1 (e1a2, e13a2, e13a3,
e14a2), KMT2A-AFF1, ETV6-RUNX1, RUNX1-RUNX1T1, CBFB-MYH11, and PMLRARA. However, only 8/9 samples were concordant for negative results. The PanHeme assay revealed one of the negative samples to be positive for a TFGGPR128 fusion. The average percentage of unique reads among all samples was
66%. All positive samples passed the minimum number of gene fusion reads of n=5
(mean=672) and unique start sites of n=3 (mean=170). Conclusions: The custom
MSK Archer Pan-Heme assay produces more comprehensive results for gene
fusions compared to standard methods by concurrently screening for genetic
alterations in 199 genes. Even among a small set of 9 negative patient samples, the
Pan-Heme assay’s targets revealed a fusion that was previously undetected. This
assay provides several advantages, including high multiplexing capability and
scalability, detection of rare fusions, and characterization of fusion genes involving
heterogeneous partners, which may provide insight into heme biology.
H53. Validation of the LeukoStrat CDx FLT3 Mutation Assay to Detect Internal
Tandem Duplication (ITD) and Tyrosine Kinase Domain (TKD) Mutations in
1058 Patients with AML and Response to Midostaurin
A. Osgood1, E. Shakeri1, V. Atkinson1, R. Navarrete1, K. Garcia1, D. Caguioa2, S.
Ferguson2, N. Kha2, C. Simon3, J. Thornes2, T. Stenzel1, J. Miller4
1Invivoscribe Technologies Inc., San Diego, CA; 2Laboratory for Personalized
Molecular Medicine LLC, San Diego, CA; 3Laboratory for Personalized Molecular
Medicine, Kawasaki Kanagawa, Japan; 4Invivoscribe Technologies, Inc. and
Laboratory for Personalized Molecular Medicine, San Diego, CA.
Introduction: Acute myeloid leukemia (AML) in general has a poor prognosis
with FLT3 mutations being the most important prognostic indicator. The Invivoscribe
LeukoStrat CDx FLT3 Mutation Assay (CDx), recently approved by the FDA, targets
regions of the FLT3 gene to identify internal tandem duplication (ITD) and tyrosine
kinase domain (TKD) mutations to predict the response to midostaurin. Both ITD and
TKD mutations result in constitutive autophosphorylation and activation
of FLT3. Methods: Mononuclear cells were isolated from peripheral blood and bone
marrow aspirate by a proprietary method. DNA was extracted and amplified by PCR.
TKD amplicon was enzymatically digested and along with the ITD amplicon were
analyzed by capillary electrophoresis. The assay measured the ratio of signal from
mutation to that of wild type. A FLT3 mutation was detected (and reported as
positive) if the mutant:wild type signal ratio (SR) ≥ 0.05, the cut-off. Proprietary
software calculates the SR for both ITD and TKD and reports positive or negative. A
validated Roche 454 Next Generation DNA Sequencing method (454 Sequencing)
was used for comparison in which DNA was amplified, and the amplicons were
inspected, purified, quantified, normalized, pooled and sequenced. Specimens were
collected from AML patients through the RATIFY drug trial and from LabPMM San
Diego. Results: The clinical performance of the CDx was evaluated using 1058 AML
patient samples from the RATIFY trial who had previously been enrolled using a
CTA. The CTA/CDx concordance for FLT3 status was, 97.2% for positive percent
agreement (PPA), 97.3% negative percent agreement (NPA) and 97.3% for overall
percent agreement (OPA). Further, the hazard ratio for the CTA+/CDx+ population
was 0.67 versus 0.77 for the CTA+ population alone. CDx agreement to 454
Sequencing (PPA, NPA and OPA) was 94.0%, 97.5% and 95.1%, respectively,
based on 764 patient results. Peripheral blood and bone marrow concordance (PPA
and NPA) was 98.4%, indicating both specimen types can be used for patient
diagnosis (N=184 pairs). Additionally, 20 clinical samples tested at 3 independent
clinical LabPMM laboratory sites on 3 continents showed 100% positive and
negative percent agreement between all sites and showed high concordance for
signal ratios. Conclusions: Overall, there was high concordance between the CDx,
the RATIFY CTA, and 454 sequencing with respect to FLT3 ITD and TKD gene
mutation detection. In addition, CDx+/CTA+ patients demonstrated superior
prediction for response to midostaurin relative to the CTA alone. The high
reproducibility between the 3 LabPMM clinical laboratories provides evidence that
the CDx is a safe and effective, internationally harmonized test and useful for patient
enrollment on multiple continents.
H54. Clinical Validation of the Lymph2Cx Assay to Determine the Cell of Origin
of DLBCL
D. Abdel Azim, A. Crowley, X. Zhang, K. Fu, T.C. Greiner, A. Cushman-Vokoun
University of Nebraska Medical Center, Omaha, NE.
Introduction: The stratification of Diffuse Large B-Cell Lymphoma (DLBCL)
according to cell-of-origin (COO) into GCB (Germinal Center B-cell), ABC (Activated
B-Cell) and unclassifiable subtype is important for determining prognosis and
potential therapeutic modalities and is required by the WHO, 2016. Several panels of
immunohistochemical (IHC) stains have been clinically used since 2008 to
subclassify DLBCL into GCB and ABC subtypes with moderate concordance rates
with Affymetrix-based gene expression profiling (GEP). In 2016, the WHO proposed
the use of RNA quantification methods, as they showed higher concordance rates to
GEP than IHC. The Lymph2CX assay employs the NanoString platform (Seattle,
WA) to determine ABC or GCB gene expression from DLBCL FFPE tissues using
mRNA transcripts from 20 genes. The purpose of this study is to evaluate the
Lymph2CX Assay for clinical use. Methods: FFPE blocks from 22 DLBCL cases,
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
with ≥60% malignant cells, were subjected to RNA extraction using the QIAcube
Instrument and the RNAeasy FFPE kit (QIAGEN, Germany). In addition, 4 cell lines
(2 ABC and 2 GCB) were evaluated (2 FFPE and 2 frozen, extracted by Trizol
method). Lymph2Cx expression profiling was performed using NanoString’s
nCounter Elements technology with processing on the nCounter Prep Station. The
nCounter Digital Analyzer was used for data collection, with analysis through
NanoString’s nSolver Analysis Software and a proprietary NIH algorithm. Results:
For the 4 cell lines tested, the Lymph2Cx Assay correctly assigned all cases (100%).
Ten cases classified by GEP (5 GCB, 3 ABC and 2 unclassifiable cases) were 90%
concordant with Lymph2CX –based classification. Twelve cases classified according
to a Choi IHC algorithm (4 GCB and 8 ABC) were 83% concordant with Lymph2Cx–
based classification. Among the 3 discrepant cases, 2 cases (1 ABC by GEP and 1
ABC case by IHC) were assigned as indeterminate/unclassifiable by Lymph2Cx. The
other case, previously categorized as ABC by IHC, was designated as GCB by
Lymph2CX. Sensitivity assessment was performed on LY3 RNA, which showed a
cut-off of 60% LY3 RNA necessary for assigning ABC versus GCB subtypes.
Conclusions: We demonstrate that the Lymph 2Cx DLBCL - COO predicting assay
has high rates of correct subtype assignment when compared to GEP and IHC
methods (90% and 83% respectively), similar to published studies. Tumor cells
should represent ≥60% of cell population assessed. Limit of detection of ABC versus
GCB prediction requires at least 60% of tumor cell RNA. The NanoString Lymph
2CX assay can be adapted to the clinical setting for DLBCL stratification.
H55. BCR-ABL1 Minor Breakpoint (e1a2) Monitoring Using an Analytically
Validated Multiplex Assay
M. Dodge, I. Beldorth, S. Filipovic-Sadic, K. Masson, M. Fahey, J. Hedges, J. Brown
Asuragen, Austin, TX.
Introduction: Detection of BCR-ABL1 e1a2 fusion transcript (minor breakpoint, mBCR) of t(9;22) assesses tumor burden in chronic myeloid leukemia (CML) and
Philadelphia-chromosome-positive precursor B-cell acute lymphoblastic leukemia
(Ph+ B-ALL). While the International Scale standardized reporting against a common
baseline for Major breakpoints e13a2 and e14a2, such a scale does not exist for
e1a2. Regardless, as newer TKI therapies create deeper responses, analytical
sensitivity has become a critical topic in investigations into TKI efficacy for minor
breakpoint and even TKI discontinuation. Researchers require an assay for minor
breakpoint that confidently calls molecular responses of ≥4 logs of reduction. We
describe analytical validation and method comparison of a multiplex system reporting
continuous BCR-ABL1:ABL1 %ratio values via automated
analysis. Methods: Armored RNA Quant (ARQ) molecules form a blend of nucleaseresistant BCR-ABL1 and ABL1 transcripts to calibrate and control the system.
Multiplexed 4-point curves using ARQ blends provide BCR-ABL1 and ABL1 copy
values and account for the relative batch run-specific efficiency of the RT step.
Controls (high, low, negative) were also developed using ARQ technology. cDNA
generation and qPCR were optimized, including allowance of high mass of nucleic
acid without inhibition. Cell line RNA positive for e1a2 was diluted into non-leukemic
human RNA specimens to create challenge panels for precision, LOB, LOD, LOQ,
linearity, shelf life, and RNA input studies. Cell line RNAs negative for e1a2 were
used to test specificity. Residual clinical RNAs were tested in comparison to the
BCR/ABL1 Quant Test (RUO) as a predicate. Software was developed that includes
a logic algorithm that flags any specimen requiring further
review. Results: Preliminary performance characteristics were established.
Specifically, LOB was estimated for both BCR-ABL1 copy number (LOB of 1.1
copies/qPCR) and %ratio of 0.0011%. The LOD (classical parametric) and LOQ
were estimated to be 0.0032% and 0.0039%, respectively. Despite deep analytical
sensitivity, this system maintains analytical specificity (below LOD for BCR-ABL1negative samples); however, it does demonstrate the expected low-level crosstalk
from contrived clonal specimens of extremely high BCR-ABL1 Major breakpoint
transcripts. Linearity was estimated to be at least 0.01 to 10%. Conclusions: The
BCR-ABL1 minor test quantifies deep molecular responses and improves workflow
with its streamlined reagent formulation, multiplex assay format, and automated
software analysis.
H56. Droplet Digital PCR Method for Absolute BCR-ABL1 Major and Minor
Transcript Quantification
C.A. Schandl, Y. Wang, J. Woolworth-Hirschhorn
Medical University of South Carolina, Charleston, SC.
Introduction: Relative quantitation of the BCR-ABL1 transcript is the standard of
care for monitoring chronic myeloid leukemia (CML) and Philadelphia chromosome
positive acute lymphoblastic leukemia (Ph+-ALL). Current methodology employs an
International Scale for purposes of decreasing inter-laboratory result variability.
Using single molecule / picoliter droplet digital PCR (pddPCR), absolute
quantification with normalization to a constitutively active gene not under specific
interrogation (GUSB) may allow for improved sensitivity and reproducibility across
platforms. In addition, pddPCR allows for definition and quantification of major (p210;
e13/e14) and minor (p190; e1) transcripts in the same assay without a Poisson
correction or commercial calibrators. Methods: One hundred patient RNA samples
characterized by traditional methodology were identified for re-analysis.
Fluorescence gates (FAM and TET) were optimized using plasmids containing the
BCR-ABL1 minor and major transcripts and annealing temperatures were further
optimized using patient samples that demonstrated both the minor and major
967
AMP Abstracts
transcripts. Using a one-step reverse transcription and real-time PCR assay
including primers specific for the minor and major BCR-ABL1 transcripts, samples
were dropletized with less than 10% droplet occupancy as determined by DNA
concentration. PCR, droplet fluorescence interrogation, and data analysis were then
performed. The experimental pddPCR data were then normalized (GUSB count)
and compared to the historical data to determine correlation. Results: With the
optimized annealing temperature (56oC), adequate separation was obtained between
the major and minor BCR-ABL1 transcripts using this methodology and absolute
nucleic acid molecule counts were possible. No false negative results were
obtained. In one case, based upon historical patient data, the pddPCR method
identified CML disease progression as evidenced by appearance of a minor
transcript in one who previously demonstrated only the major transcript
approximately 6 months earlier than current methods. Conclusions: Correlation with
pddPCR method to RT-PCR method is obtainable without the use of calibrators.
Widespread implementation of pddPCR would significantly decrease inter-laboratory
variability. This laboratory developed assay is thus far robust, accurate, and sensitive
and may better identify individuals at risk for relapse earlier in disease than RTPCR. When separation of the major transcripts becomes more clinically relevant,
the assay can be easily adapted to provide separate quantification by adjustment of
the fluorophores.
H57. Method Based Validation of 94 Genes Next Generation Sequencing (NGS)Based Hematologic Malignancy Panel and Confirmation of Variants Using
Sanger Sequencing
C.S. Sears, K. Nandakumar, B. Mineo, E.P. Murray-Carmichael, J.J. StocksCandelaria, S. Arumugam, C. Lutz, C.D. Jennings, R. Arora
University of Kentucky, Lexington, KY.
Introduction: The introduction of high-throughput, massively parallel sequencing
technologies and its application to hematologic malignancies in recent years has
provided novel insight into disease pathogenesis, risk stratification and therapeutic
intervention. This is reflected in the revised WHO 2016 classification which includes
integration of genetic data into clinicopathologic classification of hematologic
malignancies. We performed the analytic validation of a next-generation sequencing
(NGS) based panel of 94 genes using Illumina platform with different types of
hematologic malignancies. Methods: A total of 37 hematological malignancy and
solid tumor samples that were previously tested at ARUP Laboratories & Foundation
Medicine were used in a method based validation using an Agilent SureSelect
custom probe targeting coding exons for 94 genes in hematologic malignancies and
198 genes in solid tumors. The DNA was extracted from blood, bone marrow and
formalin- fixed paraffin embedded tissue (FFPE). All targets were captured and
sequenced using Illumina HiSeq 2500 with Rapid V2 chemistry (Illumina, Inc., CA).
The method based validation for the implementation of the Heme panel required 3
steps: 1. Comparison between known cancer samples sequenced in our lab & other
reference laboratories 2. Using reference cell line materials (HapMap samples
NA12878 and NA19240) to identify the limit of detection 3. Method based validation
of Sanger sequencing and confirmation of NGS variants. Results: 1. A total of 46
mutations were detected in a total of 12 hematologic malignancy samples used in
the validation. The concordance between known hematologically sequenced
samples that were sequenced in our lab and simultaneously sent out to reference
laboratories was found to be 96%. 2. The limit of detection using the HapMap
samples was determined to be 5% allele frequency, at which our assay has
sensitivity of 98% and 94% respectively to detect single nucleotide variants (SNV)
and indels. 3. Twenty-two out of these variants were in 15 actionable genes which
comprise 16% genes (15/94) on our panel. These 22 variants (22/46) which
comprise 48% of the total variants detected in the hematologic malignancies on this
method based validation were confirmed using Sanger sequencing and
PCR/Capillary Electrophoresis (for FLT3-ITD & FLT3-TKD) assays and showed
100% concordance. Conclusions: This study summarizes the design, validation and
implementation of a 94-gene Hematologic malignancy NGS panel using Illumina
HiSeq platform and in-house bioinformatics pipelines with variant calls at 5% allele
frequency at an average read depth of 1500 amplicons. This is a comprehensive
panel that includes several recently discovered genes that are frequently mutated in
hematologic malignancies.
H58. Multiple Highly Concordant Assays Facilitate Clinical Analyses of
Samples at Different Scales and Sensitivities
L.M. Chamberlain1, V. McClain1, A.R. Carson1, O. Kiya1, W. Huang1, C. Chander1, S.
Zheng1, D. Hubbard2, D. Caguioa2, Z. Xie1, J. Thornes2, B. Patay1, T. Stenzel1, J.E.
Miller1
1Invivoscribe, San Diego, CA; 2LabPMM, San Diego, CA.
Introduction: A variety of assays at different scales and sensitivities, from capillary
electrophoresis (CE) to next generation sequencing (NGS) panels, can be used to
detect clinically relevant mutations. This range of assays provides flexibility to
choose the appropriate tool for different objectives. Large panels are useful in the
early investigational stages of drug and treatment while highly targeted assays are
effective in later development stages where single biomarkers can be tracked to
monitor patient progress. At this scale, assays are available with cost and sensitivity
tradeoffs. While flexibility in assay availability is vital to development, it is equally
important that results are reliable and concordant. Here we show the ability to
reliably detect clinically actionable FLT3 and NPM1 mutations at varying frequencies
968
in clinical samples using a range of CE and NGS assays. Methods: Eight clinical
AML samples were tested using multiple Invivoscribe developed diagnostic assays,
including single gene analyses by CE and NGS, small scale (<20 genes) NGS
panels, and large scale (20+ genes) NGS panels. Of these 8 samples, 2 are positive
for a single nucleotide mutation in FLT3’s TKD domain (FLT3-TKD), 5 are positive
for an internal tandem duplication (ITD) in FLT3 (FLT3-ITD), and 5 are positive for a
4bp insertion in NPM1. One sample is negative for all 3 of these mutations. In
addition, the clinical samples were diluted as low as 1 in 1000 in order to test the
ability to detect mutations at frequencies down to MRD (10-4) levels. Results: All
assays exhibit perfect concordance for both positive and negative mutations in the
tested samples. In addition, the measured allele frequencies are highly similar
between the assays, with R2 values ranging from 0.77 to 0.99. The lowest R2 values
(0.77 to 0.78) are observed when comparing results in amplicon-based versus
hybridization-based assays. Following dilution of the clinical samples, low frequency
mutations are no longer observable in the less sensitive CE assays. However, the
mutations continue to be observable at frequencies as low as 10-4 using the more
sensitive NGS assays. Conclusions: Multiple assays at different scales and
sensitivities can be used to detect clinically actionable mutations in AML samples.
Excellent detection and frequency concordance is observed within this small clinical
sample set, indicating that reliable mutation analysis that is comparable between
assays. This data indicates that multiple assays at different scales and sensitivities
can be used in drug development and molecular diagnostics, thus validating the
flexibility available in assay choice for the evaluation and improvement of patient
outcomes and personalized medicine.
H59. Detection of Rare Variant NPM1 Transcripts Using an Allele Specific RealTime qPCR Assay Targeting Mutation Types A, B, and D
J.A. Schumacher1, P. Szankasi1, T.W. Kelley2, J.L. Patel2
1ARUP Laboratories, Salt Lake City, UT; 2University of Utah, Salt Lake City, UT.
Introduction: Nucleophosmin (NPM1) mutation is a common genetic abnormality in
acute myeloid leukemia (AML). NPM1 mutations prototypically occur in exon 11 and
result in a net 4bp insertion. The most common variants, type A (ins TCTG), B (ins
CATG), and D (ins CCTG) constitute >90% of all NPM1 mutations. However,
numerous rare variants have been described. Mutated NPM1 represents a desirable
molecular target for minimal residual disease (MRD) monitoring because it is a
stable, disease-specific marker. Allele-specific (AS) PCR is a standard technique for
quantitation of mutant allele burden. In addition, monitoring of RNA allows for a
higher sensitivity compared to genomic DNA. Most NPM1 AS-PCR assays target
type A, B, and D but the ability to detect rare variants is unknown. A strategy
employing patient-specific primers may be desirable but is laborious. In this study,
we describe detection and quantification of rare NPM1 variant transcripts using a
multiplex assay targeting type A, B, and D. Methods: RNA was isolated from 6
whole blood or bone marrow specimens and reverse transcribed into cDNA using
standard methods. cDNA or control plasmids were amplified in duplicate using
singleplex or multiplex AS-primers targeting NPM1 types A, B, D, and ABL1 in the
presence of a common hydrolysis probe. Real-time PCR was performed on the
ABI7500 Fast Dx instrument and data analyzed with 7500 Fast System
Software. Quantification was performed by normalizing mean Cp values between
replicates to a plasmid calibrator harboring a NPM1 type A and ABL1 insert,
mimicking a 1:1 ratio of NPM1 Type A and ABL1 cDNAs. Data was expressed as
normalized copy number (NCN). Results: Specimens from 4 patients harboring a
non-type A, B, or D mutant (NPM1 ins CCGG, TAGG, CTTG or CAGA, respectively)
tested positive for NPM1 mutations using multiplex AS-RT-qPCR, with NCN values
ranging from 1.598 – 0.023. Singleplex reactions showed specific amplification by 1
or 2 of the 3 AS-primer sets. A putative post-induction chemotherapy sample
collected from one of these patients 32 days after the initial specimen tested
negative. Conclusions: This study describes detection of rare NPM1 variants using
AS-RT-qPCR targeting types A, B, and D. We detected 100% (4/4) of unique rare
variants available for testing. We postulate that allele-specific primer cross-reactivity
accounts for the ability to detect the rare (non-targeted) variants. These data suggest
that a multiplex AS-RT-qPCR reaction targeting type A, B, and D NPM1 mutant
transcripts could be employed for disease monitoring in AML patients with less
common NPM1 mutations. However, further studies are needed to evaluate the
sensitivity of rare NPM1 variant quantitation using non-matched type A, B, and D ASprimer sets.
H60. Comparison of Clinical Digital Karyotyping by Comprehensive Next
Generation Sequencing with Standard Cytogenetic Analysis in Pediatric
Leukemia
E.M. Azzato, S. Raimondi, J. Zeiler, D. Hedges, Z. Zhang, X. Chen, J.
Nakitandwe, S. Shurtleff, S. Newman, S. Foy, M. Rusch, T. O'Neill, J. Downing, K.
Nichols, J. Zhang, D. Ellison
St. Jude Children's Research Hospital, Memphis, TN.
Introduction: Genome-wide copy number information obtained by next generation
sequencing (NGS) provides several advantages to conventional cytogenetics,
including increased resolution, improved detection of cryptic or subtle translocations,
and unequivocal detection of masked hypodiploidy by identification of copy-neutral
loss of heterozygosity (cnLOH). Despite these advantages, there is limited
experience with this technology in the clinical realm. In order to evaluate the clinical
validity of this technique, we performed a prospective comparison of cytogenetics
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
with digital karyotype generated by comprehensive sequencing on fresh/frozen
pediatric leukemia samples. Methods: As part of a clinical research study, Genomes
for Kids, patients have standard cytogenetic and molecular evaluations. Matched
tumor/normal samples undergo whole genome (WGS) at 45x, whole exome (WES)
at 100x and transcriptome at 20-30x coverage; the assay pipeline integrates genetic
lesions detected by all 3 NGS platforms to cross-validate and characterize single
nucleotide variations (SNVs), short insertions and deletions, structural variations
including fusions, digital karyotypes, and cnLOH. FISH for iAMP21, ETV6-RUNX1,
BCR-ABL1, IGH-CRLF2 and KMT2A (MLL) was performed by Cytogenetics on
appropriate cases. Cytogenetics chromosome count was based on the dominant
clone. For NGS, chromosomes were considered gained or lost only if a centromere
was involved. Chromosome numbers were binned into clinically known categories.
Results: Over 14 months, 84 pediatric leukemia patients (57 ALLs, 22 AMLs, 5
biphenotypic) underwent cytogenetics evaluation and comprehensive NGS. By
cytogenetics analysis, 11 samples were diploid, 30 hyperdiploid [12 (47-50), 4 (5154), 14 (55+)], 7 hypodiploid (45), 4 near tetraploid, and 32 pseudodiploid. When
NGS and modal numbers were compared, 5 samples were discordant: 51-54 vs.
55+; 46 vs. 47 and 45 vs. 46 in samples with complex rearrangements). 39 samples
had recurrent translocations detected by both techniques; 12 additional undetected
cryptic/subtle translocations were identified by NGS (PAX5-NOL4L, PAX5-AUTS2,
DUX4-IGH, P2RY8-CRLF2, IGH-EPOR, MEF2D-BCL9, CBFA2T3-GLIS2, BCL11BTLX3, BCL11B-NKX2). iAMP21 was identified in 4 samples by NGS; 2 were initially
missed by cytogenetics but confirmed by FISH subsequently. Ten samples had armlevel or full chromosome cnLOH detected by NGS. Conclusions: Comprehensive
sequencing successfully identifies relevant karyotypic abnormalities for leukemia,
while also detecting fusions not easily identified by cytogenetics. With a 28 day turnaround-time and the capability to provide additional relevant variant detection (SNVs,
cnLOH), comprehensive sequencing can replace traditional testing in many
scenarios.
H61. Sequential NGS-Based Multi-Gene Mutational Analysis in de novo Acute
Myeloid Leukemia with RUNX1 Mutation
R. Luthra, K.P. Patel, M.J. Routbort, L. Medeiros, A.E. Quesada, R. KanagalShamanna
University of Texas M.D. Anderson Cancer Center, Houston, TX.
Introduction: Acute myeloid leukemia with RUNX1 mutation (AML RUNX1mut) is a
provisional entity in the 2016 revision to the World Health Organization classification.
Studies have shown that patients with AML RUNX1mut are often resistant to induction
chemotherapy and have inferior event-free survival. These patients also often have
concomitant mutations in other genes in addition to RUNX1. In this study, we
performed sequential multi-gene NGS-based mutation analysis in patients with
AML RUNX1mut analyzed at the time of initial diagnosis and subsequently, either at
relapse with persistent disease, allowing a comparison of the gene mutation profiles.
Methods: From our database of 47 patients with de novo AML with RUNX1
mutation, 10 patients had a repeat bone marrow (BM) work-up either at the time of
persistent AML (n=4) or at the time of relapse (n=6). Using genomic DNA extracted
from BM, we performed NGS-based mutation analysis using the Illumina Miseq
platform and a 28-gene panel designed to target the entire coding regions. NGSmutation analysis was performed both at the time of diagnosis and subsequently,
either at relapse or in patients with persistent disease. Results: Thirteen
RUNX1 mutations were identified in 10 patients and included: 5 point mutations, 3
deletions, 2 duplications, 2 insertions, and one splice mutation. Concomitant
mutations at the time of initial diagnosis included: ASXL1 (n=5), IDH1 (n=3),
DNMT3A (n=4), TET2 (n=3), IDH2 (n=2), EZH2 (n=2), NRAS (n=1), TP53 (n=1),
and FLT3 (n=1). At time of relapse or persistent disease the same RUNX1 mutations
were retained in 8 (80%) patients; this patient subgroup included 2 patients who had
2 different RUNX1 mutations. In 2 (20%) patients with relapsed AML (blasts 15%
and 13%), the subsequent NGS-panel did not detect any RUNX1 mutation. NGS
analysis at the time of relapse or persistent disease showed additional mutations in
the following genes: NRAS (n=2), IDH1 (n=1), and TP53 (n=1). Three patients
received specific drugs targeting mutations in NRAS (n=1), IDH1 (n=1),
and FLT3 (n=1) with subsequent disappearance of IDH1 and FLT3 and reduction in
the number of mutations of NRAS. Conclusions: In most
AML RUNX1mut patients, RUNX1 mutations are stable, but RUNX1 mutation can
disappear in a subset of patients at the time of relapse. Therefore, the mutational
landscape in AML RUNX1mut is fluid in a subset of patients and sequential NGSbased mutation analysis can provide additional information including the uncovering
of potential therapeutic drug targets.
H62. A Case Report of Donor-Derived Clonal Hematopoiesis After Allogeneic
Stem Cell Transplantation
J. Smith, G. Eickelberg, R. Cook, R. Press
Oregon Health & Science University, Portland, OR.
Introduction: Over the past decade, advances in allogeneic hematopoietic stem cell
transplantation (HSCT) have allowed older people to serve as donors for their
siblings. In approximately 5-10% of otherwise healthy individuals over 60 years of
age, somatic mutations in genes encoding epigenetic regulators are present at
relatively low frequencies in peripheral blood cells, known as clonal hematopoiesis of
indeterminate potential (CHIP). CHIP is associated with an increased future risk of
hematologic malignancies and all-cause mortality. We have identified a HSCT
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
patient with CHIP, whose clone was derived from her donor. Methods: We searched
the Oregon Health & Science University (OHSU) HSCT registry for patients who
received molecular disease monitoring, both pre- and post-HSCT, by using our 42or 76- gene targeted next-generation sequencing (NGS) panel. Sequencing was
performed using Ampliseq libraries and an Ion-torrent PGM sequencer. To determine
whether new post-HSCT mutations were of donor or recipient origin, we examined
pre-HSCT time points for the same mutations. Results: In the HSCT registry, 41/427
donors were older than 60 years. Among recipients receiving stem cells from donors
older than 60 years, there was one allogeneic HSCT recipient who had a new
DNMT3A p.P777S mutation (CCDS33157.1) in her bone marrow 33 days postHSCT, at 5% mutant allele frequency. NGS of the stored peripheral blood donor
DNA from her 63 year old brother (drawn 20 days prior to HSCT) revealed this same
DNMT3A mutation at 5%. This donor had mild thrombocytopenia (and no other CBC
abnormalities) at the time of his sister’s HSCT. The sister had been diagnosed with
acute myeloid leukemia 63 days before HSCT, and had achieved complete
remission after 7+3 induction chemotherapy. During 3.5 years of post-HSCT followup, she showed no evidence of relapse, but she had persistent mild
thrombocytopenia. The DNMT3A mutation doubled to ~10% at ~0.5 year after
HSCT, and remained stable thereafter until 2 years post-HSCT (with 2 additional
NGS time points). No new gene mutations were detected in the serial post-HSCT
bone marrows. Conclusions: We identified an HSCT donor with CHIP, marked by a
mutation in a canonical genetic driver of myeloid malignancy (DNMT3A). The same
clone was transmitted to the HSCT recipient. This aberrant clone may have been the
underlying cause for the persistent mild thrombocytopenia in both the donor and the
recipient. Our findings confirm the possibility of transmitting a CHIP clone to HSCT
recipients, with the consequent increased risk for the development of abnormal
hematopoietic function. As the clinical significance of donor-derived CHIP is
elucidated, donor selection strategies for HSCT may need to be revised.
H63. Acute Promyelocytic Leukemia with Atypical Karyotype and FLT3 ITD
Mutation is Associated with Inferior Clinical Outcome
A. Idrees1, R. Sharma2, L. Nguyen1, M. Menes2, P. Papenhausen3, L. Zhang1
1Moffitt Cancer Center, Tampa, FL; 2University of Florida College of Medicine
Jacksonville, Jacksonville, FL; 3Laboratory Corporation of America, Research
Triangle Park, NC.
Introduction: Acute promyelocytic leukemia (APL) is an aggressive leukemia
involving PML-RARA gene fusions, which usually respond to all-trans-retinoic acid
(ATRA) plus idarubicin or arsenic trioxide (ATO) with 90-100% complete remission
and overall survival of 86-97%. A minority of APL clones reveal additional
cytogenetic abnormalities beyond t(15;17)(q22;21) or variants of RARA gene
rearrangements. FLT3-ITD mutations are seen in a subset of newly diagnosed APL
(15-25%) cases and are associated with early relapse, lower complete remission
rate, and higher mortality during induction therapy. However, whether
FLT3 mutations are more frequently associated with APL with an atypical karyotype
is unclear. Methods: Retrospective analysis (Institutional Review Board approved) of
APL cases seen at Moffitt Cancer Center between 1/2009-5/2017 was performed.
The patients with atypical karyotypes other than the standard reciprocal translocation
and additional cytogenetic aberrations were retrieved. Clinical and laboratory
investigation including bone marrow biopsy, flow cytometry, fluorescent in situ
hybridization (FISH), karyotyping, molecular studies, and clinical outcomes were
analyzed. Results: Eighty-six patients diagnosed with APL (median age 51 years,
range 22-70 years; male to female ratio of 5:9) and treated at Moffitt Cancer Center
(81) and University of Florida at Jacksonville (5) are identified and confirmed by
FISH and/or polymerase chain reaction (PCR).Conventional karyotyping was
performed in 80% (69/86) of the APL patients. 14 of 64 exhibited atypical karyotype
including t(7;17;15)(1), t(4;17;15)(1), t(6;17;15)(1), t(14;15;17)(1), t(1;17;15)(1),
tetraploidy (92, XXYY, t(15;17))(1), ins(15;17)(q22;q21.1q21.3)(1), t(11;17)(1),
additional cytogenetic aberrations including del(9q), del(7q), and t(4;21)(3) , and
complex cytogenetic abnormalities(3). FLT3 analysis was carried out in 6 of 14
patients. 3 of the 6 patients showed FLT3-ITD mutation, one of which had an
additional c-Kit mutation. A subpopulation of patients demonstrated CD34 positivity
(28.5%,4/14) by flow cytometry, not frequently seen in typical APL. All 14 patients
received ATRA and arsenic or iadrubicin-based chemotherapy. Two patients died
during initial induction and one died due to relapse. Among the 3 patients, 1
demonstrated t(11;17)(ZBTB16-RARA fusion) and 2 harbored FLT3-ITD mutations.
With median follow-up of 16.5 months (1-74 months), overall survival is 79%, which
is lower than reported APL cases. Conclusions: The study shows that APL patients
with an atypical karyotype have inferior clinical outcome, particularly those with FLT3 ITD mutations and t(11;17). It is recommended that karyotyping with FLT3 mutation analysis be routinely performed when a diagnosis of APL has been made.
H64. High Frequency of MYD88 L265P Mutation in Ocular Adnexal Marginal
Zone Lymphomas and Its Clinical Correlates
A. Behdad1, X.Y. Zhou2, D. Dittman3, C. Qi1, J. Gao1, B. Betz4, Y. Chen1
1Northwestern University, Chicago, IL; 2Bascom Palmer Eye Institute, Miami,
FL; 3Northwestern Memorial Hospital, Chicago, IL; 4University of Michigan, Ann
Arbor, MI.
Introduction: Ocular adnexa are structures that surround the eye and include orbital
soft tissue, lacrimal glands, the conjunctiva, and the eyelids. Primary ocular adnexal
extra-nodal marginal zone lymphoma or mucosa-associated lymphoid tissue (MALT)
969
AMP Abstracts
lymphoma (POAML) is the most common orbital tumor, comprising more than half of
all orbital tumors. The etiology and underlying molecular pathogenesis of POAML
has not been well characterized. Association with Chlamydia psittaci infection
appears to be seen in some parts of the world but is uncommon in North
America. Unlike other types of MALT lymphoma, POAML is usually not associated
with chromosomal translocation. Two recent studies investigated the mutational
landscape of POAML and found different results: A Korean study found frequent
alterations in TNFAIP3, TBL1XR1 andCREBBP; and a study from the United States
found frequent MYD88 mutations. MYD88 mutations are generally uncommon in
marginal zone/MALT lymphomas. The goal of this study was to investigate frequency
of MYD88 L265P mutation in our series of POAML patient and correlate the results
with the chlamydia infection and clinical findings. Methods: A total of 28 cases with
the diagnosis of marginal zone lymphoma in ocular adnexa, diagnosed between
2003 and 2017 in our institution were identified. The archival material was retrieved
and relevant clinical information was obtained. Presence of MYD88 L265P mutation
was evaluated using an allele specific real-time PCR assay. PCR testing for the
detection of Chlamydia species was performed using primers that targeted the 23S
ribosomal RNA gene. Results: Our cohort included 24 cases of primary ocular
MALT lymphoma and 4 cases with history of MALT lymphoma in other organs, also
involving ocular adnexa. The male to female ratio was 11:17, with a median age of
64 years old (range: 30-86). Median follow up of 94 months was available for the
patients. 8 of 22 (36%) POAML cases harbored MYD88 L265P mutation. Age, tumor
size, stage and event free survival was similar in patient with or without mutation. In
2 cases with suspected POAML, IGH gene rearrangement was negative, but
detection of MYD88L265P mutation confirmed the diagnosis. None of the MALT
lymphomas with secondary involvement of ocular adnexa showed MYD88 L265P
mutation. No amplification using Chlamydia-specific PCR primers was observed in
any of the samples. Conclusions: MYD88 L265P mutation was seen with high
prevalence in our POAML patients, which may be helpful in establishing the
diagnosis in some cases. High prevalence of mutation was not associated with
chlamydia infection and the clinical manifestation of the cases with mutation was
similar to the cases without this mutation.
H65. A Balanced Formulation of Dimethyl Sulfoxide and Bovine Serum
Albumin Provides Highly Uniform Coverage of CEBPA in a Droplet PCR-Based
NGS Panel
S. Mallampati, D. Duose, M. Mehrotra, C. Lan, J.P. Windham, I.I. Wistuba, S.
Verstovsek, R. Luthra, K.P. Patel
University of Texas MD Anderson Cancer Center, Houston, TX.
Introduction: Recurrent mutations are very common in myeloid neoplasms.
Targeted next-generation sequencing (NGS) technologies offer an unparalleled
advantage for identifying these recurrent mutations across multiple genes in a single
assay. However, achieving a uniformly high sequencing coverage of high-GC
content genes such as CEBPA can be extremely challenging with conventional
NGS-based workflows. Therefore, low-throughput and less sensitive approaches
such as Sanger sequencing may still provide a better alternative. In the current
study, we aimed to improve the coverage of CEBPA and obtain highly uniform
coverage of a targeted NGS-based multi-gene myeloid panel. Methods: For assay
analytical validation purposes, we used a myeloid origin cell line, OCI-AML3, and
myeloid neoplasm patient specimens (n = 10) and prepared the sequencing libraries
using a panel that covers 49 genes which are frequently mutated in myeloid
neoplasms (Thunderbolts Myeloid Panel, Rain Dance Technologies). Amplification of
input DNA material was carried out in the droplets created on the RainDrop source
instrument prior to first-stage polymerase chain reaction (PCR). After first-stage PCR
amplification, each library was indexed in second-stage PCR, and pooled samples
were subsequently sequenced on MiSeq. NGS data were analyzed using NextGene
software (Softgenetics). We modified the vendor’s workflow during first-stage PCR
and evaluated the effect of the addition of dimethyl sulfoxide (DMSO) and/or bovine
serum albumin (BSA) on the coverage of amplicons across the panel. Results: We
found that DMSO, a DNA secondary structure modifying agent, could improve the
coverage of CEBPA more than 5-fold. However, this improvement was associated
with a modest but significant decrease in the coverage of other amplicons across the
panel. We reasoned that this could be due to inhibitory effects of DMSO on the
enzymatic function of Taq polymerase used in first-stage PCR. Therefore, we tested
whether enzyme stabilizers such as BSA could improve the DMSO-mediated
systemic reduction in coverage of non–GC-rich amplicons in the myeloid panel. The
combination of DMSO and BSA improved the coverage of high–GC-rich genes such
as CEBPA and yielded highly uniform coverage across all genes in the myeloid
panel. Conclusions: Our findings suggest that the combination of DMSO and BSA
could uniquely improve coverage of GC-rich genes without adversely affecting the
coverage of other amplicons in the current panel. This simple modification provides a
tool to improve the coverage of CEBPA and other GC-rich regions in targeted NGS
panels.
H66. Next Generation Sequencing-Based Heme Panel Testing for Myeloid
Neoplasms at a Tertiary Care Hospital and Cancer Center
K. Shah, J.J. Stocks-Candelaria, S. Wei, R. Arora
University of Kentucky, Lexington, KY.
Introduction: The diverse group of myeloid neoplasms is further stratified into
subgroups including AML, MDS, MPN and MDS/MPN. There has been a massive
970
expansion in the mutational landscape of different myeloid malignancies in recent
years which have been reflected in the revised WHO 2016 classification. It integrates
genetic data with clinicopathologic classification of myeloid malignancies. These
mutations help in diagnosis, risk stratification and therapeutic intervention of AML
and are critical in accurate diagnosis of other myeloid neoplasms in patients with
cytopenias or dysplasias of undetermined significance and normal karytotype/FISH
findings. Methods: A custom probe panel designed using SureSelect (Agilent
technologies) targeting coding exons in 94 genes involved in hematologic neoplasms
was used to capture all targets on extracted DNA from blood or bone marrow
samples. Samples were sequenced using Illumina HiSeq 2500 using the Rapid V2
chemistry (Illumina, CA). Sequence data alignment and variant calling were
performed using an in-house bioinformatics pipeline that incorporates several tools
including Novoalign, GATK haplotype caller and VarScan. Human genome build
19(hg19) reference sequence was used for alignment. All new AMLs were tested
using PCR capillary electrophoresis for FLT3 and Sanger sequencing for NPM1 and
CEBPA, the final results were compared with the NGS data. Results: Bone marrow
or blood specimens (n=100) from patients with suspected, treated or confirmed
myeloid neoplasm over the last 6 months were analyzed quarterly as part of the
quality control data. 46 of these specimens showed multiple heterogenous driver
mutations; 52% and 43% of these showed normal Karyotype and FISH findings,
respectively. The highest prevalent actionable mutations were in the gene category
epigenetic modifier (34%). The other categories include transcription factors (15%),
receptor kinase (14%), RAS pathway (8%), splice site (6%), cohesion-complex (5%),
adaptor signalling (3%), tumor suppressor (3%), phosphatase and others (12%).The
most common mutations were seen in ASXL1, TET2, FLT3, NPM1 and RUNX1
genes. Conclusions: There is limited genomic data on myeloid neoplasms in cancer
patients from Kentucky. This data will provide a baseline for future comparison and
trend analysis. The use of multi-gene NGS panels specific for myeloid malignancies
has several benefits that include differentiating benign mimic conditions from
myelodysplastic and myeloproliferative neoplasms, prognostic risk stratification,
defining targeted therapy, minimal residual disease monitoring for persistence of
AML-associated mutations and monitoring of evolving low allele frequency mutation
clones in otherwise healthy cytopenic individuals.
H67. Subclonal CEBPA Mutations Identified by Deep Sequencing Using a
Clinically Validated Deep Sequencing Assay in Acute Myeloid Leukemia
S. Png, B. Kosmo, P. Lee, L. Chiu, W. Chng, B. Yan, C. Ng
National University Hospital, Singapore.
Introduction: Mutations in the gene encoding the transcription factor
CCAAT/enhancer binding protein alpha (CEBPA) are detected in about 5% to 15%
of acute myeloid leukemia (AML) patients, with double mutations in the gene
conferring a relatively favorable prognostic outcome. Although the mutational
spectrum and clonal architecture of AML is generally well-characterized, the
phenomenon and clinical impact of subclonal CEBPA mutations in AML have not
been widely elucidated. Herein, as part of the development of a next-generation
sequencing (NGS) assay for CEBPA mutational analysis, we report the identification
of subclonal CEBPA gene mutations that were not detected by conventional capillary
sequencing. Methods: Mutational analyses of the CEBPA gene of 137 AML bone
marrow or peripheral blood retrospective specimens were performed by the
amplification of the CEBPA gene using the Expand Long Range dNTPack. The
1340-bp PCR products were concurrently processed by capillary sequencing and
NGS. CEBPA-specific libraries were constructed using the Nextera XT v2 kit. All
FASTQ files were then processed with MiSeq Reporter v 2.6.2.3. Using the PCR
Amplicon workflow via the customized CEBPA-specific manifest file. The VCF files
were analyzed using the Illumina Variant Studio v2.2. Results: Deep sequencing of
the CEBPA gene generated a minimum read depth per base ranging from 3,136x to
28,184x across 137 samples. 16 of 137 specimens (11.6%) harbored mutations
in CEBPA; 4 cases (25.0%) harbored a single mutation, 11 (68.8%) harbored double
mutations and 1 (6.3%) had triple mutations. The variant allele frequencies (VAFs)
ranged from 5.6% to 58.2%.Of a total of 30 mutations detected using NGS, 3
mutations were not detected by capillary sequencing; these had VAFs of 5.6%
(p.Ser190Pro), 9.4% (p.Gln312Lys) and 11.7% (p.Asp80GlyfsTer28). Conclusions:
The identification of subclonal CEBPA mutations by NGS has clinical implications
with regards to the diagnosis of AML with biallelic mutations of CEBPA, as this AML
category has conventionally been defined by capillary sequencing, which is known to
have a limit of detection of approximately 15-20%. In addition, the presence of such
subclonal CEBPA mutations may have biological implications. Further studies will be
required to elucidate the clinical impact of subclonal CEBPA mutations.
H68. Diagnostic Yield of Somatic Mutation Detection in Hematologic
Malignancies Does Not Increase with Additional Mutation Analysis, and
Supports More Focused Disease-Specific Testing Models
S. Szelinger, S.L. Kang, T.D. Lee, R.R. Xian
University of California, Department of Pathology, Los Angeles, CA.
Introduction: Clinical next-generation sequencing become widely adopted for
hematologic malignancies in the diagnosis, prognosis, and treatment of patients with
these diseases. UCLA performs this test in a focused, customized fashion depending
on clinical indication, disease type, disease status, and clinical utility. Preset diseasespecific expanded panels (5 or more genes) and limited panels (1-4 genes) designed
as screens can be performed. Each clinical request is assessed for these
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
parameters, and the most appropriate gene(s) are selected to ensure return of
clinically meaningful data. We hypothesize that analysis of all available genes in
every case will not yield additional results that would alter patient care. Methods:
Retrospective review of hematologic malignancy mutation panels was performed to
identify 258 cases without reported mutations collected between March 2016 and
Jan 2017. Negative cases were grouped based on size of original tested panel
(expanded vs. limited). All 44 genes were then re-analyzed using the same
bioinformatics filtering as the original clinical cases. Criteria to report include
consequence of the change where only amino acid coding, splice altering, nonsense
changes, or insertion-deletion mutations were included, variant allele frequency
greater than 5%, total allelic depth of greater than 100X, and an ExAC minor allele
frequency of <0.01%. Results: Of all 539 cases performed during this period, 61.6%
were expanded and 38.4% were limited panels. The expanded panels comprised
MDS/MPN 29.9%, myeloid neoplasm 22.9%, AML 19.4%, lymphoma, 10.9%,
comprehensive 7.6%, ALL 7.6%, disease monitoring in remission 1.2%, and acute
Leukemia NOS 0.6%. Limited panels comprised CALR/JAK2/MPL 31.7%, disease
monitoring in remission 30.3%, custom 26.9%, MYD88 8.7%, ABL1 1.4%, and BRAF
1%. Excluding disease monitoring cases, expanded panels had a positivity rate,
defined as 1+ mutations, of 72.1%, and the limited panels had a positivity rate of
17.5% (p=2.6x10-14). Re-analysis of negative cases yielded an additional 9 positive
cases from expanded panels, and 12 positive cases from limited panels, which
brings overall positivity rates up to 74.7% and 23.3% respectively. Review of clinical
charts in these positive cases show that few would alter clinical management.
Conclusions: Our study shows that expanding to a more comprehensive list of
genes did not result in a statistically significant increase in diagnostic yield indicating
that the positivity rate of expanded panel versus limited panel testing is likely a
reflection of the underlying disease, or lack thereof in screening scenarios, as
opposed to the sheer number of genes tested. Our results support more focused
sequencing analysis in a disease specific and clinically relevant manner.
H69. The Utility of SNP-Array Analysis in the Detection of 1p36 Abnormality in
t(14;18)-Negative Follicular Lymphoma
L. Wang, P. Khattar, R. Aryeequaye, R. Cimera, M. Rao, J. Wang, L. Cao, A.
Fattah, C. Shen, J. Pichardo, Y. Zhang, A. Dogan
Memorial Sloan Kettering Cancer Center, New York, NY.
Introduction: A distinct subtype of BCL2-IGH/t(14;18)-negative nodal lymphoma,
characterized by predominant diffuse growth pattern, frequent inguinal involvement,
CD23 expression and deletion in chromosome region 1p36, was illustrated by
Katzenberger et al. in 2009. High frequency of TNFRSF14 mutation was recently
reported in this subtype of lymphoma. The 1p36 deletion/loss of heterozygosity
(LOH) was also identified in a subset of IRF4 translocation negative pediatric
follicular lymphomas and in primary cutaneous follicle center lymphomas. In this
study, we evaluate the utility of SNP-array analysis in the detection of 1p36
abnormalities in selective t(14;18)-negative follicular lymphoma samples. Methods:
Formalin-fixed paraffin-embedded (FFPE) follicular lymphoma tissue samples were
analyzed by BCL2-rearrangment FISH assay (Abbott molecular). Sixteen t(14;18)negative cases with morphologic and immunophenotypic features that are
suggestive of follicular lymphoma with 1p36 deletion were selected for SNP-array
analysis. There were 9 males and 7 females. Median age was 57 years (range of
34~85 years old). Tumors were located in groin (10), pelvis (1), scalp (2),
supraclavicular (1), maxillary sinus (1) and spleen (1). Genomic DNAs extracted from
FFPE tumor material were used for copy number and allelic imbalance analysis by
SNP-array (OncoScan, Affymetrix). Results: Abnormalities at 1p36 were observed in
13/16 cases, 10 of which presented as copy-neutral loss of heterozygosity (CN-LOH)
and the other 3 showed deletion. Of note, all except one case showed a focal CNLOH or deletion in the terminal portion of the short arm of chromosome 1 (common
affected region: 1p35.2-pter), and TNFRSF14 gene (1p36.32) was always involved.
The remaining one showed a CN-LOH of the entire arm of 1p. In addition, other
recurrent abnormalities revealed by SNP-array analysis in the 13 cases with 1p36
deletion/LOH included CN-LOH of 16p (common affected region: 16p13.1-pter) in
7/13 cases; and loss of PTEN (10q23) in 5/13 cases, of which 2 cases harbored CNLOH of 16p as well. In contrast, no 16p abnormality was observed in any of the 3
cases negative for 1p36 deletion/LOH; however, PTEN deletion was seen in one
(1/3) case. Interestingly, a TNFRSF14 gene mutation was identified in this case.
Conclusion: We demonstrated that CN-LOH of 1p36 is more common than 1p36
deletion in the subtype of t(14;18)-negative follicular lymphoma associated with 1p36
abnormalities. Therefore, SNP-array analysis is superior to 1p36 FISH assay in the
genetic evaluation of this subtype of lymphoma. Our study also showed that CN-LOH
of 16p and PTEN deletion are recurrent abnormalities frequently seen in 1p36
deletion/LOH-associated follicular lymphomas.
H70. Next Generation Sequencing Targeting IGH Demonstrates Clinical Utility
in Detection of B-Cell Clonality in Non-Hodgkin Lymphomas
B. Tandon, C. Porter, A. Layton
Molecular Pathology Laboratory Network, Inc., Maryville, TN.
Introduction Molecular detection of clonal IGH gene rearrangement is critical for
diagnosis of B-cell neoplasms, particularly when morphology or ancillary testing
results are indeterminate for malignancy. PCR is the current gold standard approach
in this setting. Commercial NGS assays targeting IGH have only recently become
available, and optimal testing approaches incorporating NGS are not yet well
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
established or standardized. We therefore sought to assess the potential clinical
utility of a commercially available NGS assay for B-cell clonality detection. Methods:
Twenty-four blood, marrow, or archival FFPE patient samples previously tested by
PCR were included for study. A variety of mature B-cell non-Hodgkin lymphomas, as
well samples from histologically benign lymphoid infiltrates were tested. PCR
primers employed in this study have been utilized on a clinical basis in our laboratory
for the past 10+ years and were designed according to the specifications of Meier et
al. (Am J Pathol. 2001 Dec;159 (6):2031-43). Briefly, PCR forward primers
targeting IGH V region frameworks (FR) 1 and 3, and a reverse consensus J
primer were utilized, covering >90% of expected functional IGH rearrangements.
Fragment analysis was performed on an ABI 310 Genetic Analyzer. Clonality was
defined by a relative peak height ratio >3.0. NGS testing utilized MiSeq adapters and
Invivoscribe LymphoTrack IGH FR1 primers designed in accordance with BIOMED-2
consensus recommendations. Pooled IGH FR1 V-J amplicon libraries were
quantified and sequenced on an Illumina MiSeq. Generated FASTQ files were
analyzed by LymphoTrack IGH_SHM_MiSeq Software (Version 1.0). ≥ 4-fold
difference between 1st and 3rd most abundant IGH sequences defined clonal NGS
reads. Results: PCR and NGS results were concordant among 22 of 24 samples
tested (12 polyclonal, 10 clonal). Two discordant samples showed only clonal FR3
peaks by PCR; NGS targeting IGH FR1 was negative for clonality in these cases
(flow cytometry confirmed the presence of light chain restricted B-cell
subsets). Results of ongoing NGS testing utilizing FR2 and FR3 primers are still in
progress, subsequent results to follow. Conclusions NGS is feasible for routine
clinical testing, however the optimal combinations of IGH framework primers requires
individual laboratory validation. Interpretation of PCR results for lymphoid clonality
detection is known to be subjective to some extent. Given the nature of NGS read
data, this subjectivity may be minimized. NGS delineates index
clonal IGH sequences with single nuclotide level resolution and specificity. The utility
of NGS to assess clonal relatedness among lesions at disparate sites, or evaluate
for minimal residual disease at subsequent time points, warrants further
investigation.
H71. JAK2-Negative Refractory Anemia with Ring Sideroblasts Associated with
Marked Thrombocytosis (RARS-T) Occurs More Commonly in Women
M. Hussaini, J. Song, L. Zhang
Moffitt Cancer Center, Tampa, FL.
Introduction: Refractory anemia with ring sideroblasts associated with marked
thrombocytosis (RARS-T) is a rare provisional entity that falls under the rubric of
myelodysplastic/myeloproliferative neoplasm overlap syndromes. It is characterized
by features of refractory anemia with ring sideroblasts (RARS) as well as
thrombocytosis (platelet count >450 x 109/L) with atypical megakaryocytic
hyperplasia. RARS-T is often associated with JAK2(V617F) mutation. In this study,
we interrogated cases of RARS-T by next-generation sequencing to determine
clinicopathological differences between cases that are JAK2-positive versus JAK2negative. Methods: Departmental molecular database was queried for cases
all JAK2+ hematologic malignancies with associated NGS data (Genoptix 5-gene
panel, Genoptix 21-gene panel, Illumina Myeloid 32 or 54 gene panels). Cases were
interrogated to identify cases of RARS-T. Statistical analysis was performed using ttest and Fisher's exact test. Results: Twenty cases of RARS-T with next-generation
sequencing data were identified. Nine cases were JAK2-positive and 11 cases
lacked JAK2 mutation. Overall cases, SF3B1 mutation was detected in 17/20 (85%)
of cases (89% JAK2-positive and 81% JAK2-negative cases). The most common
mutations besides SF3B1 in the JAK2+ group were TET2 (2/9; 22%), DNMT3A (2/9;
22%), TP53 (1/9; 11%), NRAS (1/9; 11%). The most common mutations
besides SF3B1 in the JAK2-negative group were TET2 (3/11; 27%), DNMT3A (2/11;
18%), TP53 (1/11; 9%), U2AF1 (1/11; 9%), and ASXL1 (1/11; 9%). The JAK2+ group
was predominantly male (7/9; 78%), while those in the JAK2-negative group were
predominantly female (9/11; 81%) (p=0.0216). The mean survival between JAK2+
and JAK- groups was 34.5 and 31 months (p=0.75) and mean age was similar (70
year versus 71 years). The highest mean platelet count in the JAK2+ and JAKgroups was 1004 x 109/L and 870 x 109/L respectively (p= 0.5393). Conclusions:
The most common mutations besides SF3B1 (prevalent in both groups) in RARS-T
were in TET2, DNMT3A, and TP53 with similar incidence between both groups.
Mean survival was longer and highest mean platelet count was higher in the JAK2
group, but the differences were not statistically significant. Interestingly, we found
that JAK2+ RARS-T tends to occur mostly in men and JAK2-negative RARS-T
occurs mostly in women (p=0.0216). This novel finding elucidates the genomic
epidemiology of RARS-T and offers insight into possible gender specific biological
differences in disease.
H72. Next Generation Sequencing Studies in Early Myeloid Neoplasms
C. Soderquist, D. Frank, J. Morrissette
University of Pennsylvania, Philadelphia, PA.
Introduction: Myelodysplastic syndrome (MDS) is a clonal neoplasm characterized
by morphologic dysplasia, peripheral cytopenias, and a variable expansion of blasts.
These abnormalities are manifestations of corrupted subcellular processes
secondary to somatically acquired genetic alterations. There exist less well
characterized “pre-MDS” states in which individuals demonstrate exuberant clonal
hematopoiesis and carry mutations in hematopoietic genes in the absence of overt
morphologic dysplasia. The clinical risk of these “pre-MDS” states remains poorly
971
AMP Abstracts
characterized; however, recent studies suggest the following molecular features may
differentiate patients at “low-risk” from “high-risk” of progression: 1) the number and
combination of mutated genes, 2) the functional class of mutated genes, and 3) the
average mutant allele frequency. We evaluated the use of our targeted myeloid nextgeneration sequencing (NGS) panel in this diagnostically challenging
setting. Methods: We performed a retrospective institutional study of patients with at
least one documented chronic cytopenia but without diagnosable bone marrow
dysplasia. Patients with normal NGS studies were classified as idiopathic cytopenia
of undetermined significance (ICUS). Patients with at least one mutation were
classified as “low-risk” or “high-risk” clonal cytopenia of undetermined significance
(CCUS). "High-risk" patients were defined as having both of the following features: 1)
at least 2 mutated genes from at least 2 functional classes (e.g. DNA methylation,
RNA splicing, etc) and 2) average mutant allele frequencies
>35%. Results: Between 2/2013 and 12/2016, 27 patient samples fulfilled criteria.
20/27 (74 %) patients were classified as having ICUS. 3/27 (11%) patients were
classified as “low-risk” CCUS and 4/27 (15%) as “high-risk” CCUS. “Low-risk” CCUS
patients showed one mutated gene each, all in DNA methylators, with an average
mutant allele frequency of 17%. “High risk” CCUS patients showed an average of 3.8
mutations each, in diverse functional classes (6 DNA methylators, 3 chromatin
modifiers, 3 spliceosome proteins, 2 transcription factors), with an average mutant
allele frequency of 42%. While follow-up is limited (median 5.5 months), 1/4 “high
risk” patients progressed to AML while 0/3 “low-risk” CCUS and 0/20 ICUS
progressed. Conclusions: Though the sample size is small, our data suggest NGS
studies may provide valuable prognostic data in that challenging subset of cytopenic
patients without overt morphologic dysplasia. In particular, those patients with
multiple mutations in multiple gene classes with elevated mutant allele frequency
may be at higher risk of progression. A larger cohort with long term follow-up is
needed to substantiate these findings.
gene involvement of these 9q34 gene rearrangements in T-ALL and 2) determine
whether FISH can help identify most of these rearrangements. Method: CytoScan
HD chromosome microarray analysis (CMA), RNA sequencing analysis and FISH
(using an ASS1/ABL1/BCR probe set or a 4 gene target probe set that includes
NUP214 coverage) were used to evaluate T-ALL cases for DNA copy number
changes, gene fusion partner and FISH signal patterns. Results: Of the 7 cases
analyzed, 4 SET-NUP214 fusions were detected and confirmed by RNA sequencing,
all of which had similar DNA start and end linear positions in the array analyses
(chr9: hg19:131,459,252-134,029,134). The cytogenetically cryptic 2.55 Mb deletions
between the 2 genes that produce in SET-NUP214 gene fusion are known to result
in leukemogenic transformation. The only ABL1-NUP214 gene fusion in the cohort
was observed secondary to a complex duplication with 2 small flanking deletions that
truncated the 2 genes centromeric to ABL1 and telomeric to NUP214. Two novel
gene fusions were identified by the array and confirmed by RNA sequencing. The
first rearrangement resulted in deletion of ASS1 and part of ABL1 and TNRC6B with
the single fusion on chromosome 22. The second resulted in an SPTAN1-ABL1 gene
fusion with deletion of the intervening region, including ASS1. Both gene fusions, to
our knowledge, have not been previous reported, but like other genes that fuse with
ABL1, it is reasonable to expect increased tyrosine kinase activity. Four of the 7
cases identified by microarray analysis also had FISH studies. The FISH probe set
containing NUP214 can differentiate NUP214 fusions from the ABL1 fusions while
both show the key deletion of the ASS1 gene probe. Conclusions: This study
demonstrates that CMA is a powerful tool to identify 9q34 interstitial deletions
resulting in gene fusions associated with T-ALL may respond to TKI therapy. CMA
identified all these gene fusions due to the presence of the gene identifying copy
number changes. Additionally, FISH deletion of ASS1 is a reliable indicator of T-ALL
9q34 gene fusions and if deleted, should be re-flexed to CMA to clarify specific gene
fusions associated with T-ALL.
H73. Correlation of Mutational Burden Detected by Targeted Next-Generation
Sequencing with Pathological Disease Burden in Hematological Malignancies.
S.L. Kang, V.A. Arboleda, T.D. Lee, R.R. Xian
University of California, Los Angeles, CA.
Introduction: Clinical next-generation sequencing (NGS) has become critical to aid
diagnosis, prognosis and treatment selection in patients with hematologic
malignancies. Targeted NGS panels with high depth of coverage allows for accurate
determination of mutant allele burden with a high degree of certainty. Extrapolating
variant allele frequency (VAF) to a proportion of mutated cells is a commonly
accepted assumption, yet, few studies have performed direct correlation of VAF with
tumor burden. We hypothesize that mutant allele frequency will significantly correlate
with disease burden and highlight the importance of integrating VAF when
determining clinical significance of somatic mutations. Methods: Retrospective
review of clinical hematologic malignancy mutation panels performed at UCLA
identified 258 cases with reported mutations collected between March 2016 and
January 2017. Analysis of corresponding pathologic and laboratory data was
performed to assess disease burden. When a single mutation was detected, the
corresponding VAF was used. When multiple mutations were detected, the mutation
with the highest VAF was used. An adjusted VAF was calculated in each case.
When the reported VAF was <50%, it was multiplied by 2 with the assumption that it
is mono-allelic. When the reported VAF was >50%, it was not altered with the
assumption that it may be bi-allelic or hemizygous. Linear regression analysis and
Pearson Correlation analysis were performed. Results: Out of 539 tested cases,
281 cases (52%) had reportable mutations. Excluding molecular disease detection in
morphologically negative bone marrows, 258 cases were analyzed. Cases with the
highest VAFs often had the most mutations. The overall correlation of VAF with
disease burden was statistically significant (R2=0.24, Pearson r 0.49, p<0.0001).
When analysis was subdivided by disease, acute lymphoblastic leukemia (ALL)
showed the highest correlation (R2=0.66, Pearson r 0.81, p<0.0001), followed by
lymphoma (R2=0.41, Pearson r 0.64, p <0.0001), acute myeloid leukemia (AML)
(R2=0.27, Pearson r 0.51, p<0.0001), and myeloid neoplasm inclusive of
myelodysplasia and/or myeloproliferative neoplasms (R2=0.09, Pearson r 0.30, p=
0.0002). Conclusions: Our analysis demonstrates while there is an overall
correlation between VAF and disease burden, the strongest correlation occurs when
there is a clearly defined neoplastic population such as in ALL and lymphoma.
Myeloid disorders show a more scattered correlation pattern, which may highlight the
heterogeneity in disease composition. Our findings support the practice of correlating
VAF with disease burden when interpreting the clinical significance of a mutation.
Further analysis is needed to reconcile the weaker correlation observed in myeloid
disorders.
H75. Comparative Study of the Panel Based Validation with Method Based
Validation in Myeloid Panel
R. Wu1, A. Rangel2, D. Andrews1, Y. Fan1
University of Miami, Miami, FL; 2Memorial Hospital Miramar, Miami, FL.
Introduction: Validation of NGS is challenging due to the complexity of the next
generation sequencing test, which involves both wet and dry labs: DNA processing,
library preparation, sequencing and large data analysis. Not only the complexity of
the NGS processes, but also incredible diversities of the possible mutations of
targeted genes involved, contribute to the challenges of validation faced by
molecular pathologist. It is not realistic to validate the every single possible mutation
in the targeted genes. Therefore, CAP adopts the method based validation for
NGS. Here, we compared the panel based validation with the method based
validation. Although both methods have its own pros and cons, the panel based
validation improved the accuracy and provided the confidence in the NGS assay.
Methods: Method based validation: DNAs from 23 patients were extracted and
separated into 2 portions. Each portion of the DNAs was submitted to different CLIA
labs and sequenced and analyzed using different panels and different platforms by
these 2 different CLIA labs. Panel based validation: DNAs from 20 patients were
extracted and separated into 2 portions. Each portion was submitted to different
CLIA labs but analyzed using exact same panel and same platform in these 2
different CLIA labs. For precision and analytic sensitiveness, the validation was
performed in one CLIA lab. However, different control DNAs from various sources,
including different types of pathogenic variants and non-pathogenic variants, was
utilized to validate the precision and analytic sensitiveness. Results: For method
based validation of 23 cases, 45 variants among total 47 variants were detected by
both methods with 2 variants were not picked up by the lab after comparison to
reference lab results. The concordance rate in method based validation is 96%. In
panel based validation of 20 cases, 81 variants among the 82 variants were detected
by both labs. Only one variant was missed by the lab due to the lower coverage
compared to reference lab. The concordance rate is 99%. The lowest detection rate
of allelic frequency for SNP, small indels, large indels is 5%, 5%, 10% respectively.
All mutations with respective allelic frequencies were detected in different
samples. The concordance rate of variants with 3 different types of control DNAs
was 100% both intra-run test and inter-run test. Conclusions: Although method
based validation is widely acceptable for NGS validation, the panel based validation
improved the accuracy by comparing the NGS settings among different CLIA labs
using exact same platform and same panel and provided more confidence in the
validation of the NGS assay.
H74. Unique 9q34 Rearrangements in T-ALL: Elucidation and Characterization
by Microarray Analysis, RNA Sequencing and FISH
J. Tepperberg, S. Schwartz, I. Gadi, A. Penton, P. Papenhausen
LabCorp, Research Triangle Park, NC.
Introduction: T-cell acute lymphoblastic leukemia is (T-ALL) is an aggressive
hematological malignancy affecting the development of T-lymphocytes.
Approximately 8% of T-ALL has a rearrangement that involves the ABL1 tyrosine
kinase gene at 9q34 and although BCR-ABL1 is a common fusion in B-ALL, it is not
a typical finding in T-ALL. We present 7 cases of T-ALL involving chromosome
rearrangements of 9q34. The goal of this study was to, 1) better characterize the
972
Infectious Diseases
ID01. Evaluation of Cepheid Xpert HIV-1 Qual Assay in Whole Blood for
Diagnosis of HIV-1 Infection
S. Lim, X. Chan, W. Tan, K. Chan, L. Oon
Singapore General Hospital, Bukit Merah, Singapore.
Introduction: While ultrasensitive HIV-1 quantitative real-time PCR assays are
increasingly being used in the diagnosis of early acute HIV-1 infections with
inconclusive serological results, HIV DNA PCR is still the preferred test for infants
born to HIV-infected mothers. HIV-1 RNA in plasma may become undetectable due
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
to perinatal antiretroviral therapy (ART) and neonatal prophylaxis, while HIV-1 DNA
in peripheral blood mononuclear cells (PBMCs) remains detectable. In this study, we
evaluated the performance of the Cepheid Xpert HIV-1 Qual assay, a relatively new
nucleic acid amplification test for the identification of HIV infection in whole and dried
blood, requiring minimal sample manipulation and with a time to result of only 90
min. Methods: A total of 45 non-duplicate whole blood specimens were included in
this study to evaluate the performance of the Xpert HIV-1 Qual assay for detection of
HIV-1 infection, including 30 known positives and 15 known negatives. The known
positive samples were from patients with documented history of positive HIV-1
screen and positive western blot and/or patients with previous positive HIV-1
quantitative PCR results. The known negative samples were either samples tested
negative with the COBAS AmpliPrep/COBAS TaqMan HIV-1 Qual Test, v2.0, or
samples tested negative on the COBAS AmpliPrep/COBAS TaqMan HIV-1 viral load
test v2.0 with concurrent negative HIV screen and negative western blot results.
Results obtained were then tabulated to determine the assay’s sensitivity and
specificity. Results: The study yielded 28 concordant positive and 15 concordant
negative results. Surprisingly, 2 specimens from the known positive group returned
as “not detected” when tested using the Xpert HIV-1 Qual assay. Further testing was
done on these 2 discordant specimens with COBAS AmpliPrep/COBAS TaqMan
HIV-1 Qual Test v2.0, and both samples returned as “detected”. These 2 false
negative cases on the Xpert HIV-1 Qual assay were noted to have undetectable
plasma HIV-1 RNA viral loads on the COBAS AmpliPrep/COBAS TaqMan HIV-1 viral
load test v2.0. Therefore, the sensitivity of the Xpert HIV-1 Qual assay was only
deemed to be 93% (28 out of 30 concordance), while the specificity was 100% (15
out of 15 concordance). Conclusions: This study demonstrated that although the
Xpert HIV-1 Qual assay was simple to run and offered a fast turnaround time, its
sensitivity was less than ideal. The suboptimal sensitivity could be attributed to the
single-target testing (HIV-1 LTR region) in the Xpert HIV-1 Qual assay in contrast to
other commercially available assays with dual-target testing (HIV-1 gag gene and
HIV-1 LTR region), e.g. the COBAS AmpliPrep/COBAS TaqMan HIV-1 Qual Test,
v2.0.
ID02. Pathogen Detection by Metagenomic Next Generation Sequencing of
Purulent Body Fluids
W. Gu, M. Lee, S. Arevalo, S. Federman, J. Whitman, L. Khan, C. Chiu, S. Miller, J.
DeRisi
University of California, San Francisco, CA
Introduction: Patients with infected tissue or body fluids often receive antibiotics
prior to sample acquisition, limiting the yield of traditional culture
methods. Metagenomic next-generation sequencing (mNGS) is an unbiased
molecular technique that can detect virtually all known pathogens through highthroughput shotgun DNA and RNA sequencing. At University of California San
Francisco (UCSF), we recently launched a clinically-validated mNGS test for
diagnosis of the infectious causes of meningitis and encephalitis from cerebrospinal
fluid. In this report, we explore adapting this approach for more challenging body
fluid specimens: purulent secretions that have high human nucleic acid background
due to abundant leukocytes. Methods: Clinical samples consisting of patient body
fluids that were positive for bacterial infection by culture were identified. We reduced
the majority of leukocyte human DNA in the body fluids by allowing the cells to settle
under gravity and discarding the vast majority of white blood cells. We then
performed shotgun sequencing with size selection for cell-free short DNA fragments
between approximately 100 and 500 base pairs, with the goal of excluding higher
molecular weight fragments that were more likely to correspond to human genomic
DNA. Results: Of 3 culture-positive samples tested (2 Staphylococcus aureus,
one Streptococcus pyogenes), mNGS showed pathogens in all samples after using
the size selection method for cell-free DNA. For one of the samples, we
detected Streptococcus anginosus, suggestive of a polymicrobial infection, although
this organism did not grow by culture. Confirmatory molecular testing is pending.
Conclusions: Sequence reads for bacterial pathogens were enriched in purulent
body fluids by analyzing cell-free DNA in this small cohort. Successful pathogen
detection and a reduction in host background were observed after removal of
leukocytes from the sample. Targeting cell-free, short, fragmented DNA is a
potential method to increase the signal-to-noise (pathogen-to-human) ratio in highthroughput DNA sequencing reads. Additional studies are needed to validate this
finding across a broader range of pathogens and sample types.
ID03. Utilization of a Cost-effective High-Throughput Sequencing Approach for
Comprehensive Metagenomic Surveillance of Viral Pathogens in Respiratory
Specimens
S. Png1, C. Chai1, D. Loh2, J. Teo2, L. Chiu1, H. Lee1, E. Koay3, T. Loh1, C. Lee1
1National University Hospital, Singapore; 2Ngee Ann Polytechnic, Singapore;
3National University of Singapore, Singapore.
Introduction: High-throughput sequencing holds great promise for the unbiased
detection of pathogenic viruses in human specimens. However, there remain several
hurdles to be overcome prior to routine use such as: 1) cheap and effective library
preparation for various specimen types; 2) computational tool for rapid and accurate
data inference; and 3) accepted guidelines to confirm metagenomic findings. Here,
we describe an unbiased massive sequencing strategy (uMASS) for the detection of
RNA viruses on the Illumina MiSeq at a cost of SGD150 per sample. We also
developed and validated a companion bioinformatics tool (sensitive unbiased virus
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
identification; SUVI) for automated virus identification. Methods: Sixty samples,
previously tested by the NxTAG respiratory pathogen panel (RPP) were retrieved for
performance comparison with uMASS. Of these, 28 were tested positive for RNA
viruses. A mix of positive and negative samples were combined to form 10 pools (6
samples per pool). Each pool underwent human host DNA and RNA depletion before
the remnant nucleic acids were converted into double-stranded cDNA. Each MiSeq
run comprised of 5 pools that were tagmented, indexed, and normalized with the
Nextera XT DNA sample prep kit. Paired-end sequencing was performed using the
300-cycle MiSeq Reagent Kit v2. The data were analysed by SUVI and the results
were compared against the RPP. Discordant results were resolved using targetspecific PCRs. We also evaluated the performance of SUVI with 7 test datasets from
the Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra). Results: uMASS
had excellent agreement with the RPP in the detection of RNA viruses (25/28).
uMASS missed 2 metapneumoviruses (MPV) and 1 enterovirus/rhinovirus called by
RPP. These samples were tested by target-specific PCRs which corroborated with
the uMASS result. We then retested the samples with RPP and they were found to
be negative too. The discrepancy could be due to decreased sample integrity or an
initial false-positive result. uMASS found an additional MPV in a sample which was
previously found to contain compounds inhibiting the RPP. The positive result was
confirmed by PCR. Also, SUVI was able to identify all known viruses in the test
datasets (SRR453448: dengue virus 2; SRR453458: hepatovirus A; SRR1106123:
GB virus C and hepatitis C; SRR1106548: human immunodeficiency virus;
SRR1170820: bovine coronavirus; SRR1170797: bovine viral diarrhea virus;
SRR1553464: ebola virus). Conclusions: We developed uMASS which can
sequence up to 30 samples per MiSeq run. The method is cost-effective and had
excellent agreement with the FDA-cleared RPP, making it suitable for the
surveillance of RNA viruses. We recommend that a positive result should always be
confirmed by target-specific PCR.
ID04. Analytical Validation of an Analyte Specific Reagent (ASR) for
Mycoplasma genitalium Detection and Point Prevalence Assessment
S. McClellan, W. Lebar, D. Newton, M. Bachman
Michigan Medicine, Ann Arbor, MI.
Introduction: Mycoplasma genitalium is associated with non-chlamydial, nongonococcal urethritis in men, as well as with pelvic inflammatory disease in
women. Assessing Mycoplasma genitalium prevalence has been hampered by the
lack of a suitable diagnostic method. Hologic has developed ASRs that use
transcription mediated amplification (TMA) for detection of the 16s rRNA of M.
genitalium on the automated Panther system. Methods: A total of 985 specimens
submitted for CT/GC and TV testing were analyzed for M. genitalium. Specimen
types included urine (female and male), vaginal, cervical, endocervical, urogenital,
rectal and oropharyngeal. Positive specimens were confirmed by testing with an
alternate M. genitalium target TMA ASR. Urine specimens previously tested for M.
genitalium with the Hologic ASRs were obtained from an outside laboratory for
accuracy evaluation: 41 female urines (16 positive, 25 negative) and 41 male urines
(16 positive, 25 negative). Analytical sensitivity utilized replicate testing of a serially
diluted in vitro RNA Transcript (50,000 copies/mL) with probit analysis (SPSS 23,
IBM). Precision was assessed by testing 1,000 copies/mL replicates of 5 on 4
separate days and using Analyse-it Method Validation Edition software (v.4.0,
Analyse-it Software, Leeds, UK) for variance component calculations. Analytical
specificity and interference used a panel of viruses, bacteria and fungi that could
potentially be found in the urogenital tract were tested in neat and spiked positive M.
genitalium Aptima Urine Transport tubes. Specimen matrix interference was
assessed by selecting 10 vaginal, 10 cervical and 10 endocervical specimens that
had tested negative with the Aptima Combo 2. The specimens were tested neat and
again after spiking with Mycoplasma genitalium. Results: Prevalence in the
population tested was determined to be 5.2%, which ranked second to Trichomonas
vaginalis (6.0% from April 2016 to March 2017). Alternate target testing confirmed
41/43 positive results. Concordance with outside laboratory specimens was
97.6%. Analytical sensitivity was calculated to be 14.8 copies/mL at 95% confidence
of detection. Testing of potential urogenital organisms found no cross reactivity, nor
did they interfere with detection of M. genitalium. Within day, between day and total
coefficients of variance were 2.5%, 2.7% and 3.7%, respectively. There was no
matrix interference for vaginal, cervical or endocervical specimens.
Conclusions: The performance characteristics of the Hologic ASRs for the
qualitative detection of M. genitalium indicate it is a sensitive, specific and
reproducible molecular assay. It is robust enough to be used with a variety of
specimen types.
ID05. Evaluation of RealStar Pneumocystis Jirovecii PCR Kit 1.0 for Qualitative
Detection of Pneumocystis jirovecii Pneumonia (PCP) Specific DNA in
Respiratory Sample Types
K. Rottengatter1, D. Wolff2, A. Gerritzen2
1Altona Diagnostics, Hamburg, Germany; 2Medizinisches Labor Bremen, Germany.
Introduction: Goal of the study is the evaluation of the performance of
RealStar Pneumocystis jirovecii PCR Kit 1.0 (altona Diagnostics, Germany) for the
reliable detection of Pneumocystis jirovecii specific DNA in different respiratory
sample types after automated extraction with NucliSENS easyMAG system and
amplification on the LC480 (Roche). The evaluation is based on the comparison of
performance data of the RealStar Pneumocystis jirovecii PCR Kit 1.0 and the RIDA
973
AMP Abstracts
GENE Pneumocystis jirovecii (r-biopharm) in respiratory samples. Methods: An
initial comparison in qualitative detection of Pneumocystis jirovecii QCMD panel
sample material (QAF114144 QCMD 2016 Pneumocystis jirovecii pneumonia (PCP)
DNA EQA Programme) by RealStar Pneumocystis jirovecii PCR and RIDA
GENE Pneumocystis jirovecii real-time PCR was used to show general comparability
in sensitivity and specificity of both assays. Fifty human clinical nasopharyngeal
swabs were processed by NucliSENS easyMAG system using the generic protocol
for DNA and RNA extraction (BioMeriéux). Pneumocystis jirovecii retrospective
positive and negative tested routine samples were included in this study and newly
extracted. All sample eluates were analyzed in parallel with the RealStar
Pneumocystis jirovecii PCR and the RIDA GENE Pneumocystis jirovecii PCR on
LC480 (Roche). Results: Qualitative testing of both, RealStar Pneumocystis
jirovecii and RIDA GENE Pneumocystis jirovecii were generally concordant (90%):
14 out of 50 samples were detected positive by both assays, 31 out of 50 samples
were detected negative by both assays. 3 samples were detected positive by the
RealStar Pneumocystis jirovecii PCR Kit 1.0 only, whereas another 2 samples were
detected positive by the RIDAGENE Pneumocystis jirovecii Kit only. Conclusions:
Both assays show very high concordance in sensitivity and specificity of
Pneumocystis jirovecii pneumonia DNA detection in respiratory samples. Discordant
results can be explained by very low pathogen load, where both assays are close to
their respective limit of detection.
ID06. Evaluation of RealStar Bordetella pertussis PCR Kit 1.0 for Qualitative
Detection and Differentiation of Bordetella pertussis and Bordetella
parapertussis Specific DNA in Respiratory Samples
K. Rottengatter1, D. Wolff2, A. Gerritzen2
1Altona Diagnostics Inc., San Francisco, CA; 2Medizinisches Labor Bremen,
Germany.
Introduction: Goal of the study is the evaluation of the performance of the
RealStar Bordetella pertussis PCR Kit 1.0 (altona Diagnostics, Germany) for the
reliable detection and differentiation of Bordetella pertussis and Bordetella
parapertussis specific DNA in different respiratory sample types after automated
extraction with NucliSENS easyMAG system and amplification on the LC480
(Roche). The evaluation is based on the comparison of performance data of the
RealStar Bordetella pertussis PCR Kit 1.0 and the RIDA GENE Bordetella (rbiopharm) in respiratory samples. Methods: An initial comparison in qualitative
detection of Bordetella pertussis QCMD panel sample material (QAB094132 QCMD
2016 Bordetella pertussis DNA EQA Programme) by RealStar Bordetella pertussis
PCR and RIDA GENE Bordetella PCR was used to show general comparability in
sensitivity and specificity of both assays. 50 human clinical nasopharyngeal swabs
were processed by NucliSENS easyMAG system using the generic protocol for DNA
and RNA extraction (BioMeriéux). Bordetella pertussis retrospective positive and
negative tested routine samples were included in this study and newly extracted. All
sample eluates were analyzed in parallel with the RealStar Bordetella pertussis PCR
and the RIDA GENE Bordetella PCR on LC480 (Roche). Results: Qualitative testing
of both, RealStar Bordetella pertussis and RIDA GENE Bordetella were generally
concordant (88%): 17 out of 50 samples were detected positive by both assays, 27
out of 50 samples were detected negative by both assays. 3 samples were detected
positive by RealStar Bordetella pertussis PCR Kit 1.0 only, whereas another 3
samples were detected positive by RIDAGENE Bordetella Kit only. In total 6 samples
out of 50 samples stayed discordant. Five out of these 6 samples show ct values
higher than 36. Conclusions: Both assays show very high concordance in
sensitivity and specificity of Bordetella pertussis DNA detection in respiratory
samples. Discordant results can be explained by very low pathogen load, where both
assays are close to their respective limit of detection. This assumption is based on
the ct values of the discordant samples which are at cycle 36 or higher.
ID07. Development of a Quantitative BK Virus PCR Assay on the Luminex
ARIES Molecular Diagnostics Platform
T. Her, T.E. Schutzbank
St. John Hospital and Medical Center, Grosse Pointe Woods, MI.
Introduction: BK virus (BKV) is ubiquitous; most infections occur in early childhood
with no known associated disease. BKV establishes a latent infection in the
kidneys. Seroprevalence in adults is ~90%. Roughly 40% of renal allograft
recipients shed BKV in the urine, either transiently or continuously over weeks to
months. BKV Nephropathy is asymptomatic, and usually discovered due to increase
in serum creatinine levels. Nearly 50% of renal transplant patients with BKV
nephropathy experience a significant loss of function of the transplanted kidney. It is
routine practice to screen renal transplant recipients regularly for BK viremia. In this
study we compared the performance of BKV qPCR analyte specific reagents (ASR)
by EliTech, and Luminex for measuring BKV viral load in plasma using the Roche
Cobas z480 instrument, and the Luminex ARIES platform respectively. Methods:
BKV DNA qPCR testing on the Roche z480 used ELITech 20X MGB Alert BK Virus
Primers and 20X MGB Alert BK Virus Probe ASR reagents. The 25 µL reaction
consisted of 1X of the primers, probe, MGB Alert Hot Start Master and MGB Alert BK
Virus Internal Control. DNA extracted from 1.0 mL of plasma using the Roche
MagNA Pure was eluted into 50 µL of elution buffer, 5 µL of which was added to the
qPCR reaction. Thermocycling conditions were 1 cycle each of 50°C for 2 minutes,
93°C for 2 minutes, and 40 cycles of 93°C for 15 seconds, 56°C for 30 seconds, and
72°C for 30 seconds. Sample volume for the ARIES platform was 200µL; the final
974
elution volume post-extraction was 150 µL, 50µL of which was added to the reaction
tube containing the BKV MultiCode primers and the reaction master mix (Ready
Mix). Results: A BKV DNA linearity panel (Acrometrix) was analyzed on both
platforms. The ARIES and Roche z480 gave very similar results demonstrating a
lower limit of detection of 250 and 500 copies/mL respectively for both platforms. A
total of 37 patients previously tested on the z480, with results spanning the test's
linear range, were tested on the ARIES. The BKV DNA copy number correlation
between the 2 methods was very good with an R squared value of 0.95. The
average difference in log copy number between the 2 methods was -0.22, indicating
that the ARIES method may have slightly greater analytical sensitivity.
Conclusions: BKV quantification results were closely matched between the 2
different methods. The workflow with the ARIES System is greatly simplified by
elimination of DNA extraction and most hands-on steps. The high degree of
automation allows samples to be tested as they arrive. The possibility of amplicon
contamination is significantly reduced with the ARIES System by using a fully selfcontained test cassette that is not open to the environment.
ID08. Development of a Panfungal Next Generation Sequencing Assay
K.D. Tardif, K.E. Hanson
ARUP Laboratories, Salt Lake City, UT.
Introduction: The detection and accurate identification of fungal pathogens is critical
for initiating treatment in the earliest stages of infection and for guiding antifungal
therapy. Pathogen identification using targeted next generation sequencing (NGS)
has the potential for more accurate discrimination of species than is possible with
routine laboratory methods. Furthermore, targeted NGS also has the benefit of
producing a smaller, more manageable data set with greater target sequencing
coverage compared to unbiased whole genome approaches. This study examined
the critical components of a targeted NGS assay for fungal identification. Methods:
Fungal culture was subjected to bead beading using a variety of protocols before
extraction using the Maxwell 16 Viral Total Nucleic Acid Purification Kit (Promega).
DNA was amplified with FastStart PCR Master Mix (Roche) and 5 conserved PCR
primer pairs targeting 4 genes (ITS, 28S, EF-1α, and β-tubulin) were optimized for a
multiplexed reaction. Sequencing adapters and indices were added in 2 additional
rounds of PCR to construct the amplicon library. Sequencing was performed using
the MiSeq Reagent Kit v2 (300 cycle) and MiSeq System (Illumina). Sequencing
data was analyzed by BLAST against target reference sequences in Geneious R9
(Biomatters). An internal control (E. gossypii) was added to each sample. Results:
Using the optimized procedure, an average of 850,000 reads (SD 232,000) were
generated per sample. Assay sensitivity was determined using serial dilutions
of Aspergillus fumigatus and Candida glabrata tested at known genomic copy
numbers. As few as 10 genomic copies were detected through sequencing reads
generated from the multicopy genes, ITS and 28S. PCR targeting the single copy
genes, β-tubulin and EF-1α, did not produce detectable sequence, likely because
multicopy genes outcompeted single copy genes in the multiplexed PCR reaction.
Conclusions: Organism lysis and primer design had the greatest impact on the
sequence quality. Once optimized, targeted NGS provided accurate species level
identification with detection sensitivity comparable to standard PCR.
ID09. Next Generation Nucleic Acid Extraction System: NucliSens eMAG
A.M. McClernon, A.B. Freeman, A.H. Sargent, J.B. Sargent, D.R. McClernon
bioMONTR Labs, Research Triangle Park, NC.
Introduction: With the advent of new, sensitive technologies for clinical diagnostic
testing, laboratories face challenges with processing higher sample volumes and
providing rapid turnaround times. Efficient extraction is a key workflow element, but
automation that allows for flexibility and adaptability still presents challenges. Current
manual or semi-automated methods, such as QIAamp Circulating Nucleic Acid Kit
(Qiagen), enable efficient purification from multiple sample types, volumes to 5mL
and flexible elution volumes but are laborious and time consuming. bioMerieux’s new
NucliSens eMAG system provides standardized workflow with full automation from
primary sample input to output directly into PCR tubes (<90 minutes for 48 samples).
The eMAG utilizes proven BOOM chemistry, sample tracking and improved flexibility
with 2 independent sections of 24 reactions each. The system offers increased
traceability and parallel processing of various samples types with flexible
input/elution volumes. Here we examined the ability of the eMAG to extract nucleic
acid from multiple sample types to be utilized in downstream molecular applications.
Methods: To assess recovery of viral RNA, low titer HIV-1 samples (n = 55, plasma
and CSF) were extracted on eMAG and analyzed with bioMONTR Labs’ HIV-1
SuperLow Assay, a single copy RUO assay utilizing NASBA amplification and
proprietary algorithm for HIV-1 RNA quantitation. For comparison purposes,
duplicate samples were analyzed by Abbott RealTime HIV-1 Assay (1.0mL protocol).
To assess linearity, VQA stock was serially diluted in plasma (4 levels), extracted in
triplicate and tested. To assess free circulating DNA recovery, plasma was spiked
with multiple concentrations of 100bp DNA ladder (New England BioLabs). Samples
were processed with eMAG (generic protocol) and QIAamp Kit and analyzed on
Agilent 2100 BioAnalyzer G2938C using DNA 7500 chip kit to evaluate quality and
quantity. Results: Twenty-nine percent (29%) of low titer samples assayed were <1
c/mL with HIV-1 SuperLow assay while 71% were quantifiable (1–47 c/mL). The
Abbott RealTime HIV-1 Assay did not provide quantifiable results for any of these
samples. The linearity results demonstrated a direct proportional relationship
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
between dilution factor and number of HIV-1 RNA copies (R2 > 0.97). For all
concentrations of spiked DNA, the recovery and electrophoretic pattern from 1001,500bp was similar regardless of extraction system. Conclusions: These results
demonstrate that the eMAG system can efficiently recover nucleic acid from multiple
sample types and yields high quality RNA and DNA for downstream molecular
assays. The eMAG system offers the full benefits of automation and flexibility
utilizing the gold standard BOOM extraction chemistry.
ID10. HPV: The Use of Full Process Controls to Monitor Extraction Variation
A. Ricketts, F. Opdam, C. Netti
Qnostics, Glasgow, Scotland, United Kingdom.
Introduction: Reproducible and reliable screening processes for human papilloma
virus (HPV) plays a critical role in the early identification of patients at risk of going
on to develop HPV associated cancers. Variation in both the extraction and
amplification processes could lead to false negative results being reported so
appropriate controls both within the run and between runs are necessary to identify
systematic errors and shifts in reagent lot performance. The use of well
characterised run controls that simulate patient samples as closely as possible are
the ideal as they cover the whole process. In this study, 2 control preparations were
prepared and used to monitor the run to run reproducibility of the Roche Cobas 4800
HPV detection system. The data was analysed to show the use of a full process
control covering the extraction and amplification could provide information on assay
variability. Methods: The assay qualitatively detects 14 high risk genotypes,
reporting HPV 16 and HPV 18 independently and detecting the other high-risk types
on a third channel. The fourth channel monitors betaglobin (βG) as an internal
cellular control. Two control preparations were formulated in Preservcyt. The first
contained human cells positive for HPV16 and 18 with background negative human
cells. The second contained cells positive for HPV16, 45 and background cells. Over
100 of each control were run and the Ct values for each channel recorded. The data
were normalised by subtracting the average Ct value for βG from each of the 3
values. Results: The data showed that the controls performed well with good
positive detection rates (0.09% of amplifications undetected for control 1 and 0.54%
in control 2). 3.3% of raw data could be considered outliers. The controls gave
reproducible and consistent performance (Control 1 HPV 16 ±0.65 Ct, HPV 18
±1.12, βG ±0.81, Control 2 16 ±0.86 Ct, HPV 45 ±0.61, βG ±0.85). The performance
of the 3 markers in each control showed a highly similar response relative to the βG
average. When the values are normalised and overlaid the relationship between the
3 measurements shows that all 3 channels are similarly affected during processing.
Conclusions: The use of full process controls rather than amplification only controls
(e.g. plasmid based controls) allows the monitoring of the process from extraction to
amplification and more closely replicates the performance of a true clinical sample.
While further work would be required, this preliminary data suggests that extraction
may be a significant contributor to the variation seen between runs in an individual
laboratory.
ID11. Detection of Borrelia burgdorferi DNA by Loop Mediated Isothermal
Amplification (LAMP) in Pediatric Synovial Fluids
R.V. Ponaka1, C. Gomez1, D. Patel1, V. Elagin1, E. Graf2, V. Slepnev1
1Meridian Bioscience Inc., Cincinnati, OH; 2Children’s Hospital of Philadelphia,
Philadelphia, PA.
Introduction: Lyme disease, caused by Borrelia burgdorferi, is the most prevalent
vector-borne disease in the United States. Lyme arthritis is the most common feature
of late stage infection. Distinguishing children who have lyme arthritis from those
who have septic arthritis can be a challenge, but doing so is essential because their
treatment is different. The diagnostic value of antibody assays is unsatisfactory due
to the persistence of antibody after antibiotic therapy or spontaneous resolution of
infection. Molecular methods like polymerase chain reaction (PCR) and LAMP can
be used for the direct and specific detection of B. burgdorferi. We report on the
performance of a simplified B. burgdorferi detection assay in an easy to use LAMP
platform. Methods: illumigene Lyme by Meridian Bioscience (Research Use Only
(RUO), not cleared for use in USA) uses LAMP to detect B. burgdorferi DNA. The
simple DNA extraction procedure relies on chemical lysis, and produces amplifiable
DNA in approximately 7-10 minutes. Clinical sensitivity was evaluated using synovial
fluid specimens from 46 pediatric patients with known clinical history and, if ordered,
Lyme IgM/IgG EIA and, if reflexed, western blot data. Analytical sensitivity of the
assay, was determined using a real-time PCR quantified B. burgdorferi strain. SF
specimens (100 µL) were mixed with 2 buffers each followed by a brief incubation.
Prepared samples were transferred to an M-Prep column, the column was drained,
washed and eluate collected. The collected eluates were directly added to
the illumigene Lyme test device and loaded into the Meridian illumipro-10. Results
were reported in 40 minutes. Results: Retrospective clinical samples were tested at
Children’s Hospital of Philadelphia. illumigene Lyme (RUO) detected 18 positive and
28 negative samples. All LAMP positive samples were Lyme IgG positive (except
one where no antibody data was available) and 8 were Lyme IgM negative, likely
indicating a later stage infection. Of the 28 illumigene Lyme negative samples, 7
were IgG/IgM negative and 3 were IgG Positive/IgM Negative. Lyme antibody testing
was not performed on 14 samples as Lyme was not clinically suspected.
The illumigene negative, Lyme IgG positive results may be due to the persistence of
Lyme IgG antibodies in the body. The analytical Limit of Detection (LoD) of B.
burgdorferi was determined to be 1.0 x 103 copies/mL. Conclusions: The Meridian
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
Bioscience illumigene Lyme (RUO) assay is a simple and sensitive test that
detects B. burgdorferi DNA directly from as little as 100 µL of synovial fluid in
approximately one hour. This provides a specific and faster alternative to serologic
tests for the diagnosis of lyme arthritis.
ID12. ITS1 (Internal Transcribed Spacer) Primer Binding Site Polymorphism in
Clinical Fungal Isolates
J.F. Mele, N.P. Wiederhold, C. Gibas, D.A. Sutton, Y. Wang, K. Vadlamudi, C.
Sanders, W.B. Furmaga, H. Fan
University of Texas Health San Antonio, San Antonio, TX.
Introduction: Molecular identification of fungi by DNA sequencing is an important
tool for accurate fungal identification and clinical diagnosis of fungal diseases. DNA
sequencing of the ITS region using the “universal” ITS primers are widely applied in
clinical laboratories (White et al., 1990; Clinical and Laboratory Standards Institute
Guideline MM18-A). These universal ITS primers have been validated and
are routinely offered in our CLIA certified laboratory. In a 5-year period, 9,000 clinical
fungal isolates were identified using the ITS region. In a retrospective analysis of the
ITS sequence data, we discovered 2 single nucleotide polymorphisms (SNPs)
[TCCGTtGGTGAACCaGCGG] from the canonical ITS1 primer
sequence TCCGTAGGTGAACCTGCGG. To our knowledge, the frequency of these
SNPs in clinical fungal isolates has not been reported and the finding may influence
the ability to effectively detect fungal species in a clinical setting. Methods: A total of
862 clinical isolates received from August 2016 to May 2017 were included in this
study. Fungal DNA was extracted using the EZ1 DNA Tissue Kit (Qiagen). The ITS
region was amplified using the BMB-CR forward primer (Lane et al., 1985) and NL4
reverse primer (White et al., 1990). The BMB-CR primer is about 150 base pairs
upstream of the ITS1 priming site. The amplicons were sequenced by the Sanger
method (BigDye Terminator Cycle Sequencing Kit), using the BMB-CR and ITS4
primers. Sequences were aligned and edited using the DNAStar Lasergene 10
software (DNASTAR, Inc). Edited contigs were queried through the GenBank
database using BLASTn. The top 50 scoring hits were combined with phenotypic
data to identify each isolate to the species level. Results: Nineteen percent (165) of
the isolates contained the 2 SNPs in their ITS1 primer binding site. Among these
isolates, clinically significant fungi were included, such as: Acremonium persicinum,
Fusarium falciforme, Fusarium incarnatum-equiseti species complex, Fusarium
keratoplasticum, Fusarium oxysporum species complex, Fusarium
proliferatum, Fusarium solani/Fusarium solani species complex, Lomentospora
prolificans, Nigrospora oryzae, Nigrospora sphaerica, Purpureocillium
lilacinum, Scedosporium apiospermum, and Scedosporium boydii.
Conclusions: Sequencing the ITS region is an important
molecular identification tool for fungi species. The ITS1 primer has been widely
accepted as the universal primer for the ITS region. However, in our clinical
laboratory, we determined that SNPs within the ITS1 primer binding site are not
rare events (about 20% of clinical isolates carry these SNPs) and are found in many
clinically important fungi. Thus, we recommend suitable alternative forward primer be
considered if sequencing fails with the conventional ITS1 primer.
ID13. The Film Array Global Fever Panel: Goal of Quick Diagnosis of Infectious
Diseases Presenting with Acute Febrile Illness
C. Toxopeus1, J.R. Helm1, W.A. Smith1, N. Batty1, O. Davidson1, B. Marble1, L.
Border1, A. Kelley1, B.T. Gnade2, S. Fernandez2, C. Smith1, E. Pollard1, J.
Stewart1, J. Rouillard1, C.L. Phillips1
1BioFire Defense LLC., Salt Lake City, UT; 2U.S. Army, Fort Detrick, MD.
Introduction: The FilmArray Global Fever Panel is a multiplexed investigational
molecular diagnostics tool that tests whole blood specimens for the presence of 17
different pathogens that cause Acute Febrile Illness (AFI) and is being developed in
collaboration with the Department of Defense (MCS-JPEO and USAMMDA Contract
No. W911QY-13-D-0080, under the NGDS program and NIAID (NIAID Contract
No. HHSN272201600002C, “Advanced Development of Multiplex Diagnostic
Platforms for Infectious Diseases (Global Fever Panel)”). The Global Fever Panel
requires minimal handling of blood specimens, only about 200 µL specimen volume,
and test results are available within one hour. Quick diagnosis of the cause of AFI in
a patient allows for early intervention using the appropriate treatment; this has
several benefits including prevention of disease progression to a stage that would
require more expensive intervention (e.g. hospitalization, additional medication),
preventing potential negative consequences from using the wrong medication (e.g.
use of aspirin in the case of hemorrhagic fever), and preventing the spread of
disease. The Global Fever Panel detects the following pathogens: i)
Bacteria: Bacillus anthracis, Francisella tularensis, Leptospira spp., Salmonella
enterica serovar Paratyphi and serovar Typhi, Yersinia pestis, ii) Viruses:
chikungunya virus, Crimean-Congo Hemorrhagic Fever (CCHF), dengue
virus, (serotypes 1-4), Ebola virus, Lassa virus, Marburg virus, West Nile virus,
yellow fever virus, Zika virus, and iii) Protozoa: Leishmania spp., Plasmodium spp.
with additional identification for P. falciparum, P. vivax, and P. ovale.
Methods: Estimated LoD and an initial evaluation of intra- and extra-panel crossreactivity of the Global Fever Panel were performed. Whole blood (EDTA) samples
from healthy donors were spiked with each pathogen at multiple levels. Live
organism was used when possible in the BSL2; otherwise inactivated organism was
used. Nucleic acid concentration for all isolates used in the Global Fever Panel
studies was quantified using commercially available kits. Results: All organisms
975
AMP Abstracts
were detected by the Global Fever Panel at levels expected to be clinically relevant,
and at comparable sensitivity observed for existing FilmArray panels. No crossreactivity was observed when on-panel and off-panel organisms were spiked at high
levels.
Conclusions: The FilmArray Global Fever Panel is a promising investigational tool
for the early detection and identification of 17 pathogens that cause AFI.
ID14. Evaluation of Two Molecular Diagnostic Assays for Clostridium difficile
Infection
G.A. Capraro, J.W. Longshore
Carolinas Pathology Group, Carolinas HealthCare System, Charlotte NC.
Introduction: Clostridium difficile infection (CDI) is a potentially life-threatening
disease in patients who have had recent medical care and been exposed to
antibiotics. In the US, there are nearly 500,000 infections per year, with an estimated
15,000 deaths. Laboratory testing for CDI involves detection of pathogenic toxins.
Molecular testing targets the genes that give rise to these toxins, and is associated
with high sensitivity and specificity. The purpose of this study was to compare 2
recently FDA-cleared commercially-available molecular tests for diagnosis of CDI for
performance and workflow efficiency. Methods: Liquid stools submitted as part of
routine patient care were tested using our conventional 2-step algorithm: GDH/ToxinA/B EIA, followed by PCR for GDH-pos/toxin-neg stools. We tested 58 stools using
the Quidel Solana (QS) and DiaSorin Simplexa (DS) Cdiff assays according to
manufacturer instructions and compared the results to those obtained by our in-use
BD Max (M) Cdiff assay. Five GDH-pos/toxin-pos and 5 GDH-neg/toxin-neg stools
were also tested using the D and Q assays. Discordant specimens were repeated
once, and the consensus of 2 out of 3 molecular assays was used to establish the
true result. Results: Of the 58 GDH-pos/toxin-neg stools tested by the M assay, 39
of these were positive and 19 were negative. Of the 39 positive stools, 31 (79%) and
37 (95%) were positive by the QS and DS assays, respectively. Of the 19 negative
stools, 19 (100%) were negative by each of the QS and DS assays. There were 8
and 2 stools that were discordant between the QS and M assays and the DS and M
assays, respectively. In 1 case, the QS and DS assays were in agreement
(negative), in 1 case, the QS and M assays were in agreement (positive), and in the
remaining 7 cases, the DS and M assays were in agreement (positive). The QS and
DS assays correlated with the 5 double-pos and 5 double-neg stools tested. Both
assays require transfer of the specimen 3 times. The QS assay was performed with
an average of 2 minutes of hands-on time per specimen. The DS assay was
performed with an average of 1 minute of hands-on time per specimen.
Conclusions: The QS and DS Cdiff assays performed with a sensitivity of 79% and
95%, respectively, and a negative predictive value of 70% and 90%, respectively.
Both assays demonstrated specificity and positive predictive values of 100%. Both
assays had relative short hands-on time, with the time to result for the QS assay
being 29 minutes and the DS assay being 68 minutes. Both assays lend themselves
to efficient small batch processing, with the QS assay accommodating batches of up
to 12, and the DS assay accommodating batches of up to 8. Continuing studies
already in process should further resolve discrepant results.
ID15. Molecular-Based HPV Screening in Resource Limited Countries
A. Atkinson1, C.M. Studwell2, E.P. LaRochelle3, L.S. Kennedy4, S.J.
Deharvengt4, J.A. Espinal5, S. Bejarano5, G.J. Tsongalis1
1Dartmouth Hitchcock Medical Center and Geisel School of Medicine at Dartmouth,
Lebanon, NH; 2University of Connecticut, Storrs, CT; 3Dartmouth College, Hanover,
NH; 4Dartmouth Hitchcock Medical Center, Lebanon, NH; 5La Liga Contra el Cancer,
San Pedro Sula, Honduras.
Introduction: Cervical cancer rates in low- and middle-income countries (LMICs) are
3 times higher than those in developed countries. In Honduras, cervical cancer is
the most common cancer affecting women, accounting for approximately 417 deaths
annually in a population of less than 8 million. Nearly all cervical cancers are caused
by human papillomavirus (HPV) infection and implementing a low-cost, rapid, near
patient HPV screen would greatly improve cervical cancer outcomes. In collaboration
with La Liga Contra el Cancer in Honduras we have adapted HPV tests from
QuanDx for both high and low-risk HPV to provide rapid results that can assist in
screening. We sought to provide insight on the prevalence of HPV infection in the
population by identifying the high-risk and low-risk genotypes present. Methods: We
used the MeltPro High Risk and Low Risk HPV Genotyping assays (QuanDx/Zeesan
Biotech, San Jose, CA) to detect 14 high-risk and 14 low-risk HPV subtypes through
melt curve analysis. DNA was extracted by an alkaline lysis boiling of cervical swabs
collected from 111 women in remote La Mosquitia, Honduras. Once isolated, DNA
samples were added to the lyophilized MeltPro reagents, amplified on the SLAN-96
real time PCR instrument (QuanDx/Zeesan Biotech, San Jose, CA), and analyzed
for HPV genotypes. DNA samples that failed to yield a melt-curve were re-extracted
with the Lab-Aid 824 and re-analyzed. The results were reported to physicians in
Honduras to provide follow up care for women at risk for cervical cancer from HPV
infection. Results: Approximately 35% of the population examined had an HPV
infection; 27 samples tested positive for high-risk HPV strains and 18 samples were
positive for low-risk HPV strains. The most prevalent high-risk HPV genotypes were
HPV-52 found in 29% of the samples, and HPV-16 identified in 19% of the samples.
The most prevalent low-risk HPV genotype was HPV-72, which represented 18% of
the low-risk infections. Re-extraction of DNA samples with the Lab-Aid 824 increased
the ability to detect HPV and resulted in the discovery of 9 additional high-risk HPV
976
infections and 3 additional low-risk HPV infections reducing the “invalid” rate from
10-13% to 0%. Conclusion: HPV infections are prevalent in the La Mosquitia region
of Honduras and the genotype distribution differs from that of developed countries
and our previous data of less isolated communities in Honduras. In the future, the
implementation of a cervical screening program utilizing molecular HPV testing
would greatly improve the identification of at risk individuals and could help reduce
the cervical cancer rate in Honduras.
ID16. Association of Clostridium difficile Molecular Typing with Colonization
and Development of Clostridium difficile Infection (CDI)
T. Theparee1, D. Schora1, J.L. Grant1, B.A. Smith2, L.R. Peterson1, S. Das1
1NorthShore University Health System, Evanston, IL; 2Duke University Medical
Center, Durham, NC.
Introduction: Clostridium difficile infection (CDI) is a leading cause of hospital
acquired infection in the United States with an estimated 453,000 cases per year.
While there are a significant number of asymptomatic carriers, their role in hospital
transmission leading to clinical disease is uncertain. The aim of this study is to
compare C. difficile carriage and CDI using molecular ribotyping. Methods: A subset
of adult patients at a community hospital were screened upon admission for C.
difficile between July 2016 and January 2017 as part of a quality improvement
initiative. All patients who were admitted to the general medical service, GI surgery,
and from a nursing home on any service were screened. We excluded neutropenic
patients, patients admitted for CDI, and those who declined screening. Rectal swabs
were collected using a double headed rayon swab. A swab was processed using the
Cdiff qPCR Test on the cobas 4800 system (Roche). All specimens with positive
qPCR and every specimen with a negative qPCR following a positive were submitted
for anaerobic culture on cycloserine-cefoxitin-fructose agar plates with horse blood
and taurocholate for 5 days. PCR ribotyping was then performed. Briefly, DNA was
extracted from C. difficile colonies and the 16S-23S intergenic spacer region was
amplified. Fragment analysis was performed on the 3500xl Genetic Analyzer
(Applied Biosystems) with LIZ 600 standard and sized using GeneMapper v. 4.0
(ThermoFisher). Ribotypes were assigned using the WebRibo Database v. 2.2
(AGES). CDI was defined according to the CDC’s National Healthcare Safety
Network (NHSN) reporting guidelines of significant diarrhea and C. difficile toxin
gene detection. Results: Of 1566 screening tests, 87 (5.5%) tested positive for C.
difficile and had colonies available for ribotyping. We identified 16 cases of CDI: 4
were excluded for lack of post-infection culture, 6 had prior colonization, and 6 had
no prior colonization. There was a wide distribution of ribotypes among colonized
patients with pathogenic ribotypes 027 (14.9%) and 106 (12.6%) occurring most
frequently. Colonized patients with repeat testing had the same ribotype in 9 of 10
patients. In CDI patients, ribotypes 020 (16.7%) and 078 (16.7%) were most frequent
and 9 of 12 CDI patients were infected with documented pathogenic ribotypes. Only
6 (50%) had prior colonization by the same ribotype and 3 (25%) CDI patients had
unique ribotypes which was not found in the screened population.
Conclusions: While C. difficile colonization among hospitalized patients is diverse,
pathogenic ribotypes occur commonly. Further studies are needed to understand the
role of colonization by specific ribotypes and development of CDI.
ID17. Detection of Gram-Negative Bacteria and Antimicrobial Resistance
Markers Using the iCubate iC-GN Assay
M.S. Conover, K. Heflin, H. Liu, L. Wagner, C. Wells
iCubate, Huntsville, AL.
Introduction: Bloodstream infections and sepsis, with ensuing complications, are
currently the 10th leading cause of death in the United States with over one million
cases reported annually. Time to proper diagnosis and treatment of these infections
is crucial for patient survival, yet traditional culture techniques can require 1-4 days
to identify both the causative organism and antibiotic resistance markers. Delay in
treatment by only a few hours dramatically increases mortality risks and has also
been shown to increase treatment duration and medical costs. Therefore, to improve
time to treatment, iCubate has developed an automated, molecular-based, in vitro
diagnostic system, the iC-GN Assay. This assay detects the 8 most common gramnegative pathogens associated with sepsis as well as 2 carbapenemases and an
extended spectrum β-lactamase. Methods: In this study we developed and validated
the iC-GN Assay, a multiplexed, in vitro diagnostic test that detects 11 gene targets
in a single reaction. This assay is designed for use with the iC-System, which
automates sample preparation, amplicon-rescued multiplex PCR (ARM-PCR), and
microarray detection in a closed cassette. The iC-GN Assay is performed directly on
positive blood culture samples, without the need for additional culture methods.
Using clinically relevant blood culture systems, we determined the limit of detection
(LoD), inclusivity, and exclusivity of the iC-GN Assay. Additionally, we performed a
small beta study testing approximately 100 clinical samples. Results: We
determined the LoD for all targets in the iC-GN to be below 1x107 CFU/ml, far below
concentrations at blood culture bottle positivity. Additionally, the inclusivity of the
assay was determined by testing a minimum of 10 genetically diverse isolates of
each species at near LoD concentrations. Antibiotic resistance marker inclusivity was
evaluated in all obtainable species or through in silico analyses when necessary. In
conjunction with its ability to recognize a diverse range of targets, the assay is also
specific, demonstrating limited cross reactivity when tested against a panel of 116
off-target organisms. Finally, the iC-GN Assay displayed greater than 95%
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
performance during small-scale clinical testing. Conclusions: These analytical
validations demonstrate that the iC-GN Assay is capable of accurately and rapidly
identifying bacterial species and resistance determinants present in blood cultures
containing gram-negative bacteria. Utilizing molecular diagnostics like the iC-GN
Assay can decrease time to treatment, medical costs and patient mortality in a userfriendly manner.
ID18. Real-time Gastrointestinal Illness Surveillance Through Cloud Based
Epidemiology Network of Clinical Laboratories
J.M. Ruzante1 , R. Selvarangan2 , J. Jones3 , L. Meyers3
1Research Triangle Park, NC 2Children’s Mercy Kansas City, Kansas City
MO; 3BioFire Diagnostics, LLC, Salt Lake City UT.
Introduction: The Centers for Disease Control and Prevention estimates that
annually there are about 179 million cases of acute gastroenteritis in the United
States, resulting in approximately 600,000 hospitalizations and 5,000 deaths. About
80% of those cases have unknown etiology, making the control and treatment of GI
cases extremely difficult. However, new detection methods are increasing the ability
to rapidly identify these microorganisms and monitory their activity through cloud
based medical device connectivity in real-time. Methods: The FilmArray GI panel
developed by BioFire Diagnostics detects 22 GI pathogens, including 13 bacterial, 5
viruses, and 4 parasites. The assay utilizes a self-contained pouch where 200μl of
stool specimen in Cary-Blair medium is placed. The analysis consists of automated
nucleic acid extraction, reverse transcription, amplification and detection. Each
pouch contains an internal nucleic acid extraction and a polymerase chain reaction
control. All de-identified test results from a research cohort, are exported to a cloud
database that stores and aggregates all results. This system, FilmArray Trend, has
been adopted by 20 clinical laboratories across the United States, including 5
pediatric hospitals. Results: Since 2015, there are over 20,000 GI test results stored
in the database and 46% of those are positive for one or more pathogens. The top
ten pathogens detected are respectively, Clostridium difficile (13%), Norovirus (7%),
Enteropathogenic Escherichia coli (EPEC) (6%), Enteroaggregative E. coli (EAEC)
(3%), Rotavirus A (3%), Sapovirus (3%), Salmonella (2%), Campylobacter (XX%),
Cryptosporidium (2%), and Astrovirus (2%). Further, 12% of the positive results have
between 2 and 4 pathogens detected in the same sample, and out of those, C.
difficile the most frequent pathogen identified, representing 20% of the co-detections.
However, the rates and frequency for co-detection differs when comparing the
results for the 2 pediatric hospitals with the remaining sites (mixed population). In the
pediatric sites, Vibrio, E. coli O157 and C. difficile are the most common pathogens
identified co-infections, while in the mixed population sites, those consist of Yersinia
enterocolitica, C. difficile and Astrovirus. Conclusions: Beyond providing critical
information to physicians to aid patient management, the advent of this novel
technology, the increase of its adoption by clinical laboratories nation-wide, and the
centralization of results into a single database, provides an unique opportunity for the
public health community to improve outbreak detection, and better understand the
risk factors and the burden of gastrointestinal diseases in the community.
ID19. WITHDRAWN
ID20. Testing High-Risk Human Papillomavirus on Head and Neck Tumor
Tissue Squamous Cell Carcinoma Using a Modified Commercial PCR Assay
A.N. Huho1, N. Yadak2, T. Bocklage2, M.N. Vasef2, S. Yang3
1University of New Mexico- TriCore Labs, Albuquerque, NM; 2University of New
Mexico, Albuquerque NM; 3TriCore-University of New Mexico, Albuquerque NM.
Introduction: Increase in the prevalence of high-risk human papillomavirus (HRHPV) associated head and neck squamous cell carcinoma (HNSCC), specifically
oropharyngeal squamous cell carcinoma (OPSCC), has prompted clinical demands
for testing HR-HPV directly on the tumor tissue. p16 has been shown to have a high
sensitivity and specificity in OPSCC. The widely available Roche COBAS 4800 HRHPV PCR assay was approved for the primary screening of cervical cancer. In this
study, we developed a modified method based on the COBAS 4800 to detect HRHPV in formalin fixed paraffin embedded (FFPE) tissues. We set out to demonstrate
the test performance characteristics and the potential clinical utility of this laboratorydeveloped assay for direct HR-HPV molecular testing in the HNSCC FFPE
specimens. Methods: One hundred and eighteen FFPE blocks of HNSCC were
retrieved from archives (41 Excisions, 61 biopsies and 16 FNA). Tissue slices were
trimmed from the blocks and directly placed in microcentrifuge tubes, deparaffinized
and lysed followed by lysate dilution in 50% ethanol and loaded on the COBAS 4800
instrument. Results were recorded as negative or positive for HR-HPV. All
specimens had a concurrent p16 Immunohistochemical stain (IHC) performed and
interpreted as positive, negative, or equivocal by a pathologist. The limit of detection
(LOD) was determined by selecting a positive sample (by PCR and P16) and
enriching for >95% tumor percent by macro-dissection of unstained slides. The
scrapings were processed as above, with serial dilutions. The sensitivity and
specificity were determined using p16 IHC staining as a reference, the precision and
sample stability were also determined. Results: The assay showed a sensitivity of
91.9% and Specificity of 100% compared to p16 results in OPSCC (48 case). The
limit of detection was equivalent to 3 tumor cells per ml of lysate dilution. The Intraassay and inter-assay precision were both 100%. Excellent sample stability was
demonstrated by reproducibly positive results on the lysate dilutions with HR-HPV
positive tumor concentrations close to the LOD at room temperature for at least 2
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
months. Discrepant results between HR-HPV PCR and p16 IHC were most
commonly observed in FNA specimens (43.8%), which in general have relatively
fewer cells. Conclusions: The COBAS 4800 is a reliable method for detecting HRHPV in FFPE samples including FNA cell blocks. The assay has good specificity,
sensitivity, precision, very low LOD with minimum hands-on time. FFPE- FNA cell
blocks had the most equivocal p16 IHC results, and we demonstrate, for the first
time to our knowledge, that testing HR-HPV in the FFPE FNA cell blocks by the PCR
assay using the COBAS 4800 may be very useful to solve these inconclusive cases
due to insufficient tumor cells.
ID21. Investigation of Differences in Gene Expression by Kanamycin Stress in
Multidrug-Resistant Mycobacterium tuberculosis with / without rrs Mutation
Using RNA-Seq
Y. Kim, H. Kang, J. Yang, H. Gu, S. Kim, Y. Han, S. Cho, T. Park, H. Lee
Kyung Hee University School of Medicine, Seoul, South Korea.
Introduction Drug resistance of Mycobacterium tuberculosis is a major problem in
the world's healthcare and much of the mechanism of resistance in secondary drugs
used in MDR tuberculosis remains unknown. An efflux pump could be an explanation
for many drug resistances however, the relationship between the expression of
genes and phenotypic resistance is complicated due to the post-translational
regulatory network. In this study, we investigated differences in gene expression
levels according to kanamycin stress in M. tuberculosis without associated
mutations. Methods Two kanamycin resistant M. tuberculosis strains with (strain no.
M114) or without (strain no. M125) rrs A1401G mutation were prepared after single
colony isolation. There were no eis mutations in both strains. The MIC for kanamycin
of 2 strains were >640 ㎍/㎖ and >2560 ㎍/㎖, respectively. M. tuberculosis H37Rv
was also prepared as reference. McFarland 1 suspension of MTB strains were
incubated in 7H9 broth for 24 hours with 25% and 50% of MIC on the shaker. RNA
extraction was performed on the culture medium and RNA seq was performed with
illumine Hiseq 2500 through purification and library preparation using the Illumina
TruSeq RNA Sample Preparation Kit. RNA-Seq reads were mapped to the genomic
DNA reference (GCF_000195955.2) using a Bowtie aligner and Reads per Kilobase
of gene per Million mapped reads (RPKM) were calculated. Significantly differentially
expressed genes (|fold change|≥2) were screened based on their normalized value
after Log2 transformation. KEGG PATHWAY Database and volume plot using R
program were used for comparison. Results The KEGG enrichment map showed
significant differences in gene expression fold changes in M114 compared to H37Rv
according to drug stress while M125 did not. In the comparisons using volume plot
analysis between M114 (incubation with 50%, 25% and 0% of MIC, respectively) and
H37Rv, the Top 5 genes showing significant fold change were selected. Among
them, the Rv2031c expression was commonly low. Rv2395A also showed a
significant fold change compared to the drug incubated M114 (50% and 25% of MIC)
and H37Rv. Interestingly, the numerically largest positive fold change (2^24) from
the comparison between M114 25% MIC of kanamycin and H37rv was WhiB7 which
is known to be associated with the up-regulation of eis and efflux pump tap
(Rv1258c) which affect kanamycin and streptomycin resistance. WhiB7 also showed
the fourth largest fold change (2^36) in M114 50% MIC. Conclusions Our study
showed significant changes in gene expression according to kanamycin stress in M.
tuberculosis without rrs mutation. The large change in expression of WhiB7 seen
with kanamycin exposure may be a helpful finding to identify a new resistance
mechanism.
ID22. Evaluation of a Molecular Point of Care System for the Detection of
Clostridium difficile
I.O. Op den Buijs, R.T. Roymans, J.H. van de Bovenkamp
PAMM Laboratories, Veldhoven, The Netherlands.
Introduction: Clostridium difficile infection (CDI) is the most important cause of
nosocomial diarrhea in developed countries, especially in patients with antibiotic
therapy. The utilization of a fast and sensitive diagnostic method is essential for
targeted treatment and to restrain a single infection from becoming an outbreak. In
this study we performed an evaluation of the novel Revogene system
(genePOC). This system provides a point-of-care solution that can be used for the
detection of C.difficile toxin B in 1 to 8 samples within 75 minutes. Also, the footprint
of the system is as small as a microcentrifuge and the use of disposables is brought
back to minimum. Methods: In a 10 week period 401 clinical stool samples were
tested for genotypical CDI using qPCR on the Revogene system. As a reference
test, toxigenic culture (TC) was performed on all samples using Cdiff ChromID plates
followed by in vitro toxin production testing by cell cytotoxicity neutralisation assay
(CCNA). For discrepancy analysis, a CCNA was performed directly on discrepant
stools. Results: Out of the 401 tested samples, 4% was excluded due to inhibition
(N=12) or instrument failure (N=3) in the Revogene system. If focusing on inhibition,
8 out of 12 inhibited samples were found within one week. Due to the uptake of too
much sample the inhibition rate that week increased from 1% to 3%. Of the
remaining 386 samples, 23 were positive in both qPCR as TC. However, 8 samples
were only positive in qPCR and not in TC. Of these samples, 4 were CCNA positive,
proving the presence of free toxin B in these stool samples. This could indicate that
these 4 patients did suffer a CDI but the bacteria was no longer viable for culture.
For the remaining 4 samples, the presence of toxin B could not be confirmed with
CCNA indicating either a false positive result, or a lower detection limit for tcdB in
qPCR. Two samples were negative in qPCR but positive with TC. Neither of these
977
AMP Abstracts
samples could be confirmed positive in CCNA, indicating that the C.difficile has the
capability to produce toxin, but no free toxins were present in the stool. With TC as a
reference method, sensitivity, specificity, positive and negative predictive value of the
qPCR assay on the Revogene system is 92% (74-99%), 98% (96-99%), 74% and
99% respectively. Conclusions: The Revogene system provides an easy and
sensitive solution for CDI testing in the lab. The inhibition-rate highly depends on the
amount of stool that is taken. Therefore, a good instruction is necessary, making this
assay less acceptable as a point-of-care solution for CDI testing. Nevertheless, the
system itself is easy to handle by any, non-laboratory educated person and will
therefore have the potential as a true point-of-care system for other sample types.
ID23. Dermatomycosis – a Novel and Rapid Detection of Causative Fungal
Agents with a DNA-Based Microarray (EUROArray Dermatomycosis)
S. Kosanke1, M. Harder2, J. Brasch3, P. Nenoff4, M. Cavalar2, S. Uhrlaß4, Y. Gräser1
1Universitätsmedizin Berlin - Charité, Berlin, Germany; 2EUROIMMUN Medizinische
Labordiagnostika AG, Lübeck, Germany; 3Universitätsklinikum Schleswig-Holstein,
Kiel, Germany; 4Labor für medizinische Mikrobiologie Bakteriologie, Mykologie,
Virologie & Infektionsserologie, Rötha, Germany.
Introduction: Fungal infections of the feet are a common disorder in the present
age. Infected sites are either the nails (Onychomycosis) and/or the dermal tissue of
the feet (Tinea pedis). Dermatophytes species of the
genera Trichophyton and Epidermophyton are the most prevalent agents. Very
rarely, also non-dermatophytes like yeasts (e.g. Candida guilliermondii) or mould
fungi (e.g. Scopulariopsis brevicaulis, Fusarium spp.) are able to cause fungal foot
and nail diseases. Besides fungal foot disease other clinical presentations of
dermatophyte infections become more relevant and the number of local outbreaks
has increased in the last years. The diagnostic gold standard of the identification of
dermatophytes are the direct microscopy of skin or nail scales mounted with KOH on
objective slides, the setup of a fungal culture and its subsequent analysis. But this
procedure can be time-consuming and it requires a well-skilled professional to
identify the species according to its micromorphological characteristics because of
pleomorphy of micro- and macroscopical features. To improve the conventional
diagnostic a novel DNA-based microarray for the rapid and sensitive identification of
dermatophyte species was developed by the EUROIMMUN AG. The test was used
for the analysis of clinical nail, hair and skin samples. Herein, we present the
performance of this new multiplex PCR as compared to conventional methods (direct
microscopy, culture and other molecular techniques) for the diagnosis of
dermatomycosis. Methods: DNA extraction was performed using QIAamp DNA Mini
Kit (Qiagen GmbH, Germany) according to the manufacturer’s protocol for tissue
samples with the modifications that the samples were treated prior DNA purification
with proteinase K and lysis buffer ATL for at least 12 hours. PCR was performed by
using 5µl DNA-extract and 20µl PCR components (EUROArray Dermatomycosis,
EUROIMMUN AG) according to manufacturer’s protocol. The PCR products were
mixed with 65µl hybridization-buffer B (EUROIMMUN AG) and hybridized for 1h at
55°C. The final evaluation was done by using the automatically EUROArray Scanner
System (EUROIMMUN AG). Results: A total of 100 samples obtained with clinically
suspected dermatomycosis were tested by conventional methods and PCR. All in all
the PCR showed higher sensitivity and specificity compared to culture and direct
microscopy, especially in cases of mixed infections. Conclusions: The new invented
EUROArray Dermatomycosis test (EUROIMMUN AG) provides to be a reliable
method which offers a rapid identification of the most relevant dermatomycosis
pathogens within 3 hours (excluding the DNA preparation) with high sensitivity
and specificity.
ID24. Performance Comparison of the DiaSorin Simplexa C. difficile Direct
Assay with the Illumigene C. difficile DNA Amplification Assay in Unformed
Stool Samples
B.C. Sutton1, K. Maggert1, E. Rowell1, Q. Vu1, K. Gonzalez2, M. Tabb2
1The Medical Foundation, South Bend, IN; 2DiaSorin Molecular LLC, Cypress, CA.
Introduction: In the United States, Clostridium difficile infection (CDI) occurs largely
in hospitalized patients, where it causes up to 3 million cases of diarrhea and colitis
and is linked to 14,000 deaths a year. Current laboratory diagnosis often relies on
the detection of C. difficile toxins A and/or B and antigen by enzyme immunoassay
(EIA), both specific but insensitive tests. However, use of PCR-based assays for
detection of toxin-producing C. difficile is an attractive alternative in terms of
specificity, sensitivity, and rapid diagnosis of CDI, and facilitates prompt patient
management decisions. In this study we compared the performance of 2 PCRbased tests for toxigenic C. difficile, the illumigene C. difficile DNA Amplification
Assay (Meridian Bioscience, Inc., Cincinnati, OH), and the Simplexa C. difficile Direct
assay (DiaSorin Molecular LLC, Cypress, CA). Methods: Two hundred and two
unformed stool samples were evaluated by The Medical Foundation microbiology
laboratory. Seventeen samples were referred by medical offices and were likely from
outpatients; the rest came from hospitalized patients. Most specimens (187/202,
93%) had orders for C. difficile PCR testing. In the remaining specimens PCR was
performed (Simplexa and illumigene) as a confirmatory test if an EIA toxin assay
showed discrepant results between C. difficile antigen (glutamate dehydrogenase)
and toxins A or B (C. diff Quik Chek Complete, Alere, Waltham, MA). Stools were
stored at 2-8oC and were <48 hours old prior to PCR testing. Samples were tested
first with the illumigene assay that requires sample preparation (mixing and heat
treatment) prior to PCR on an illumipro-10 analyzer (Meridian Bioscience, Inc.),
978
followed by the Simplexa Direct assay, which is a direct test performed on the
LIAISON MDX instrument (DiaSorin Molecular LLC, Cypress, CA). Results: The
agreement of Simplexa Direct versus illumigene was 98% (198/202 samples), with
50 samples showing a positive result and 148 that were negative in both assays. Of
the 4 discordant cases, 3 were positive with the Simplexa test but negative with
the illumigene assay, while 1 was positive with illumigene and negative with
Simplexa. Specimen handing time was longer for the illumigene assay (25 minutes
for 8 samples), versus the Simplexa test (15 minutes for 8 samples). Total
turnaround time was shorter for the Simplexa test (8 samples to result: 75 minutes),
versus illumigene (8 samples to result: 85 minutes). Conclusions. 1. Both PCR
assays detected toxigenic C. difficile in stool with a high degree of concordance
(98%). 2. The Simplexa Direct assay requires less technical specimen handling,
and may reduce cost and improve laboratory turnaround time for routine C.
difficile testing.
ID25. Comparison of the Accula Influenza A/B PCR Assay and Alere i Influenza
A/B Isothermal Nucleic Amplification Assay for the Detection of Influenza in
Adult and Pediatric Populations
S. Young1,4, H. Cai2,4, R.B. Cary2, G. Cai2, S. Gastelum2, J. Giles2, A. Grkovich2, C.
Lindberg2, J. Monroe2, D. Thomas2, L. Vo2, J. Zeis3
1TriCore Reference Laboratories, Albuquerque, NM; 2Mesa Biotech, San Diego,
CA; 3Toolbox Medical Innovations, Carlsbad, CA; 4 University of New Mexico,
Albuquerque, NM.
Introduction: Physicians manage patients with influenza like illness (ILI) based on
history, clinical symptoms and/or rapid point- of- care (POC) antigen detection tests.
These antigen methods are fast but have lower sensitivities compared to central lab
based diagnostic tests. Central laboratory based testing with moderately complex
polymerase chain reaction (PCR) influenza tests is both sensitive and specific. The
time delay in providing a result is the limitation of central laboratory. This reporting
delay is the result of: 1) the transport time of the sample to the lab, and 2) the test
performance time. The Mesa Biotech Influenza A/B assay (Accula) is a rapid PCR
based test designed to be performed at the POC. The performance of the Accula Flu
A/B Test in detecting and differentiating Influenza A and Influenza B in fresh nasal
swab specimens from patients with signs and symptoms of ILI, was compared with
the CLIA waived Alere i Influenza A & B Assay. Methods: This study was a
prospective, multi-center study to evaluate the performance of the Accula Flu A/B
Test, an investigational Nucleic Acid Amplification Test (NAAT) designed to report
rapid test results at the POC locations during the 2016/2017 flu season at 16
investigational sites throughout the U.S. The Accula Flu A/B test was compared to
the Alere i Influenza A & B Assay in the detection and differentiation of Influenza A
and Influenza B virus in nasal swabs from patients with signs and symptoms of
ILI. The study was performed in POC locations, such as physician office
laboratories, urgent care and outpatient clinics settings. A total of 1332 subjects were
enrolled in this study. Of those, 74 specimens are un-evaluable, leaving a total of
1258 specimens that were considered evaluable. Results: Compared to the Alere i
Influenza A & B Assay, the performance of the Accula Flu A/B test for influenza A
was as follows: sensitivity: (289/298) 97% (95% CI: 94.1% - 98.5%), specificity:
(900/960) 94% (95% CI: 92% - 95.2%) and accuracy: (289+900/1258) 95% (95%
CI: 93.1% - 95.6%). Compared to the Alere i Influenza A & B Assay, the performance
of the Accula Flu A/B test for influenza B was as follows: sensitivity: (126/134) 94%
(95% CI: 88.2% - 97.2%), specificity: (1109/1123) 99% (95% CI: 97.9% - 99.3%) and
accuracy: (126+1109/1257) 98% (95% CI: 97.5% - 99.0%). The Quidel Lyra
Influenza A+B Assay was used to adjudicate the 91 discrepant tests.
Conclusions: These studies demonstrated equivalent performance of the Accula
Flu A/B test to the Alere i Influenza A & B Assay with the intended users, achieving
97% sensitivity and 94% specificity for Flu A and 94% sensitivity and 99% specificity
for Flu B.
ID26. Developing High Throughput Urinary Tract Microbiota Profiling Using
TaqMan and OpenArray Technologies
K. Li, J. Li, B. Huang, S. Patel, I. Pagani, E. Zeringer, N. Puri, J. Fonseca, E.
Diamond, L. Nyuyen, K. Varma
Thermo Fisher Scientific, South San Francisco, CA.
Introduction: Each year around 150 million people are affected by urinary tract
infections (UTIs), which could present serious health issues whether symptomatic or
asymptomatic. Currently UTIs are diagnosed based on clinical symptoms and urine
analysis (bacteria culture, and presence of white blood cells) and treated with
antibiotics. However, the urinary tract hosts a diverse and complex microbial
community and emerging evidences show that urinary microbiota may exert a
profound effect on urologic health, both positive and negative. Current
methodologies either lack target throughput or rely on bacterial culture. Therefore,
panel-based molecular testing should not only identify the presence of specific
species but also profile urinary microbiota, which would help explain its biological
significance and may potentially provide guidance for the proper antibiotics and
reduce overtreatment. Methods: To help address these unmet needs, we have
developed a new research application for urinary tract microbiota profiling by
leveraging our TaqMan technology and high throughput OpenArray platform. First
we select 17 different urinary tract relevant microorganisms (16 bacteria,
plus Candida albicans) that are important in urinary tract health. We identify unique
gene targets that are species-specific and design a collection of assays targeting
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
these signature genes for each of the species using our robust assay design
pipeline. These assays are vigorously evaluated with plasmid artificial templates and
ATCC gDNA samples on OpenArray, a microscope slide–sized plate with 3,072
through-holes (48 subarrays/plate and 64 through-hole/subarray). Results: We first
evaluated assay accuracy with plasmids and ATCC controls and all assays show
expected on-target signals. Then we tested specificity by running each assay with a
collection of ATCC gDNA samples (both inclusivity and exclusivity panels) for 17
targeted species and no cross-reactivity was observed. We also did serial dilution
for each assay and demonstrated 7 log linear dynamic range (with R2 >0.99) with
limit of detection (LOD) down to ~10-100 copies. Good PCR efficiency and
reproducibility were achieved. Conclusions: We have developed a new research
application for urinary tract microbiota profiling, including a large collection of
relevant assays. We have demonstrated excellent assay performance - high
sensitivity, specificity, accuracy and reproducibility on OpenArray platform. The
application provides researchers a powerful and cost-effective tool with simple
workflow, fast turnaround time, and high throughput yet flexible sample/target
combinations. This application may help better understanding pathogenicity in the
urinary tract and potentially improve in urologic health in the future.
ID27. Rapid Detection of Clostridium difficile with the GenePOC CDiff Assay
A. Zumoberhaus1, S. Steiner1, M. Arnold2, M. Altwegg1
1Bioanalytica, Lucerne, Switzerland; 2Cantonal Hospital, Lucerne, Switzerland.
Introduction: Rapid and accurate diagnosis of C. difficile (Cd) infection is important
for treatment and infection control measures. PCR meets these requirements and is
accepted as standalone method especially in the United States whereas in Europe
testing for the presence of GDH and free toxin B (CDT) by immunological
procedures is more common. In this (ongoing) study we analyze the performance of
the GenePOC CDiff assay for the detection of the toxin B gene (tcdB) of Cd in
comparison to other molecular and immunological methods. Methods: Stool
specimens with a request for Cd sent to Lab A (a hospital laboratory; N=123) or Lab
B (serving general practitioners (GPs) and institutions; N=124) are included in the
study. The routine procedure for Lab A consists of a GDH/CDT antigen assay (Quik
Check Complete, Alere) followed by PCR (GeneXpert, Cepheid; detects tcdB) if
GDH- and CDT-results are discordant. Specimens are then sent to Lab B for
GenePOC analysis. Lab B routinely uses the BD MAX Cdiff assay (Becton
Dickinson; detects tcdB) as standalone test. For the time of the study, all specimens
are also analyzed using the Quik Check Complete and the GenePOC test. If the
latter differs from the routine result, the Quidel AmpliVue C. difficile test is used for
final assignment as true positive or true negative. Prevalence of tcdB and
percentage of positive CDT-tests among tcdB-positive specimens were calculated in
relation to whether the specimen was submitted by an institution or by a private
practitioner (PP). Results: As compared to the true result, the GenePOC CDiff
assay exhibits a sensitivity and a specificity of 100%, whereas sensitivity/specificity
for the BD MAX (based on the 124 specimens of Lab B only), GDH and CDT assays
are 100%/98.0%, 88.2%/98.6% and 50%/100%, respectively. None of the false
negative GDH tests occurred in a specimen positive for CDT. The overall prevalence
of tcdB is 7.3% for Lab A (all specimens institutional) as compared to 15.7% for
institutional specimens (N=51) and 23.3% for specimens from GPs (N=73) in Lab B.
Irrespective of the origin of the specimens, CDT was detected with very similar
frequencies (50.0 to 55.6%) in tcdB-positive specimens of these 3 groups.
Conclusions: The GenePOC CDiff assay is an easy to use, sensitive and specific
tool for the detection of tcdB-carrying Cd in stool specimens. It matches the
performance of other molecular assays and is analytically more sensitive, but
possibly clinically less specific, than the GDH and CDT antigen tests. Astonishingly,
Cd with and without free toxin can be found in similar frequencies in the various
patient groups. It might be interesting to analyze the prevalence of tcdB, GDH and
CDT in patients not suspected of having Cd-associated diarrhea.
ID28. Multicenter Evaluation of Cobas HBV Real-Time PCR Assay on the Roche
Cobas 4800 System in Comparison with COBAS AmpliPrep/COBAS TaqMan
HBV Test: Leading Circle for Cobas 4800 Virology (LCCV) Project
H. Kim1, M. Hur2, E. Bae3, K. Lee4, W. Lee5
1Konkuk University Hospital, Seoul, South Korea 2 Konkuk University School of
Medicine, Konkuk University Hospital, Seoul, South Korea; 3Veterans Healthcare
System Medical Center, Seoul, South Korea; 4Gangnam Severance Hospital, Yonsei
University College of Medicine, Seoul, South Korea; 5Kyung Hee University Hospital
at Gangdong, Seoul, South Korea.
Introduction: Hepatitis B virus (HBV) nucleic acid amplification test (NAAT) is
important for diagnosis and management of HBV infection. We evaluated analytical
performance of the newly launched cobas HBV (Roche Diagnostics GmbH,
Mannheim, Germany) on the cobas 4800 System in comparison with COBAS
AmplPrep/COBAS TaqMan HBV Test (CAP/CTM HBV) in 4 clinical laboratories.
Methods: Precision was evaluated using 3 levels of cobas HBV/HCV/HIV-1 Control
Kit, and linearity was evaluated across the anticipated measuring range (10.0 IU/mL
to 1.0 x 109 IU/mL) at 8 levels using clinical samples. Correlation using 212 clinical
samples, turn-around-time (TAT) for 24 tests, and limit of quantification (LOQ)
according to the processing volume were compared between the 2 systems.
Results: Within-laboratory imprecisions (standard deviation [SD] and coefficient of
variation [%CV]) ranged from 0.06 to 0.09 and from 1.29% to 2.63%, respectively.
Between-laboratory imprecisions was < 0.06 and < 1.57%, respectively. Linearity
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
(coefficient of determination, R2) ranged from 0.999 to 1.000, and overall correlation
was very high (r = 0.996) between the 2 systems. TAT in cobas 4800 System and
COBAS CAP/CTM were 3.5 hours and 5 hours, respectively. LOQ for cobas HBV
and CAP/CTM HBV were < 10 IU/mL with 400 uL and < 20 IU/mL with 500 uL,
respectively. Conclusions: The new cobas HBV real-time PCR assay on the cobas
4800 System showed reliable analytical performance and improved workflow. It
could be a good laboratory option for HBV NAAT.
ID29. A Model for Detection of Novel Influenza Incidence in the United States
J.D. Jones1 , J. Nawrocki1 , B. Galvin1 , M. Poritz2 , C. Ginocchio1 , L. Meyers1
1BioFire Diagnostics, LLC, Salt Lake City, UT; 2BioFire Defense, Salt Lake City, UT.
Introduction: Early detection of novel influenza could facilitate detection of
pandemics. The BioFire FilmArray Respiratory Panel (RP) can detect 20 respiratory
pathogens including Influenza A, subtyping H1 and H3. FilmArray Trend is an
epidemiology system that automatically receives de-identified test results from the
FilmArray Instruments. This project has connected 20 sites across the US and
collected 300,000 test results since 2012. In this dataset 362 tests were FluA
positive with no sub-type determined. All of these need further evaluation by public
health, suggesting that 0.5% FilmArray tests were sent to public health laboratories
for further novel Flu evaluation. Although Flu A is not subtyped beyond H1, H1-2009
and H3 by the FilmArray RP, we investigated whether the (masked) PCR data within
the dataset could be used to detect novel influenza samples. Methods: More than
17,000 Flu A positive test results were found in the FilmArray Trend dataset with 362
tests resulting in no subtype detected. We applied the Pathogen Extended
Resolution (PER) test, a model designed to detect pathogen subtypes using PCR
amplification metrics, to the 5 Influenza-A assays. Two control datasets were used to
train the PER test and to evaluate its performance: 1) Isolates from 8 strains of avian
influenza, and 2) Common Flu Strains of known concentrations, including those near
the FilmArray Limit of Detection (LoD). The 362 no-subtype detected tests were
then classified as either a common or novel strain using the FluA-PER test. Results:
The percent positivity for the PER test, trained to identify common influenza,
demonstrated an overall sensitivity of 100% and specificity of 92.3% when evaluated
against 28 isolates of common and 13 isolates of novel influenza. The FluA-PER test
associated 99.4% (n=360) of FilmArray Trend tests having no-subtype detected as
common Flu, near LOD. The remaining 0.6% (n=2) were classified as novel flu.
These samples classified as novel flu have not been confirmed and will be
investigated further. Conclusion: We have demonstrated that the PCR amplification
data available to BioFire through the FilmArray Trend database can be used to
identify signatures associated with common and novel influenza strains.
Development of the Pathogen Expanded Resolution (PER) test to identify patterns
masked within the FilmArray software has the potential to provide alerts to network
participants of anomalous patterns in influenza trends. Further validation,
development and refinement of these alert algorithms and how to use them must be
explored before broad implementation.
ID30. Comparative Evaluation of ARIES Flu A/B & RSV and Xpert Flu/RSV XC
for Simultaneous Detection and Identification of Influenza Viruses A, B and
Respiratory Syncytial Virus in Cancer Patients
L. Ling1, S.E. Kaplan2, J.C. Lopez2, J. Stiles2, X. Lu1, Y. Tang2
1The Eighth Affiliated Hospital of Sun Yat-sen University, Shenzhen Guangdong,
China; 2Memorial Sloan Kettering Cancer Center, New York, NY.
Introduction: Rapid identification of respiratory pathogens such as influenza viruses
A (FluA), B (FluB) and respiratory syncytial virus (RSV) helps limit unnecessary
antimicrobial use and enhance infection control practices. We performed a
comparative evaluation of ARIES Flu A/B & RSV (Luminex Corp, Austin, TX)
and Xpert Flu/RSV XC (Cepheid, Sunnyvale, CA) assays. Methods: Limits of
detection (LODs) of the ARIES and the Xpert assays were determined using titerknown FluA, FluB, and RSV control strains. Clinical performance was evaluated
using 200 remnant nasopharyngeal swab (NPS) specimens received for a FilmArray
Respiratory Panel (BioFire, Salt Lake City, UT) testing during the 2015-2016
influenza season against a standard that combines the 3 assays. Results: The 95%
LODs (TCID50) of the ARIES and the Xpert assays were 0.027 and 0.003 for FluAH1, 0.003 and 0.003 for FluA-H1N1, 0.027 and 0.027 for FluA-H3, 0.240 and 0.024
for FluB, 0.538 and 0.538 for RSV-A, and 0.170 and 1.702 for RSV-B, respectively.
Among the 200 specimens included, the ARIES and the Xpert detected 48 (24.0%)
and 50 (25.0%) FluA, 45 (22.5%) and 44 (22.0%) FluB, and 48 (24.0%) and 49
(24.5%) RSV. Sensitivities of the ARIES and the Xpert were 96.0% (95% CI, 86.5 to
98.9%) and 100.0% (95% CI, 92.9 to100.0%) for FluA, 97.8% (95% CI, 88.7 to
99.6%) and 95.7% (95% CI, 85.5 to 98.8%) for FluB, and 98.0% (95% CI, 89.3 to
99.6%) and 100% (95% CI, 92.7 to 100%) for RSV, respectively. Specificities of the
both assays were comparable at 99.3 to 100% for all 3 viruses. The intra- and intertest coefficients of variation ranged from 0.5 to 3.3% and 1.9 to 2.2% for the ARIES
and 0.7 to 6.1% and 2.6-3.2% for the Xpert. Hands-on-time and test turnaround time
(TAT, mean±SD) were 1.2±0.2 and 121.2±0.2 minutes for the ARIES and 1.1±0.3
and 31.1±0.3 minutes for the Xpert. Conclusions: Both ARIES and Xpert assays
demonstrated excellent sensitivities and specificities for simultaneous detection and
identification of FluA, FluB and RSV from NPS specimens in cancer specimens. TAT
was significantly shorter on the Xpert when instrument scalability is unlimited.
979
AMP Abstracts
ID31. Rapid Diagnosis of Bloodstream Infections Through Identification of
Pathogens and Resistance Markers Directly from Whole Human Blood at 1
CFU/ml
N. Casali1, T. Mach1, Z. Lang1, F. Paillier1, B.A. Brown2, L. Clarizia2, A. Bahrami1, S.
Barakat1, N. Brown1, M. Davila2, A. Day1, J. Gallagher1, C. Icely1, J. Killpack1, S.
McBader1, G. McCabe1, M. Norvell2, S. Pradhan1
1DNA Electronics, London, England, United Kingdom; 2DNA Electronics, Carlsbad,
CA.
Introduction: Bloodstream infections (BSIs) are a major cause of morbidity and
mortality worldwide. Due to the lengthy turnaround time required to identify the
causative organisms by current standard of care culture-based diagnostics, patients
typically receive empirical treatment with broad-spectrum agents, impacting survival
rates and hospital stay. The DNA Electronics (DNAe) LiDia platform is a rapid bloodto-result diagnostic system that addresses this unmet need in the treatment of
serious BSIs leading to sepsis. We report the detection of BSI pathogens and
antibiotic resistance genes in 1 CFU/ml spiked samples and clinical specimens
directly from whole blood using a modular workflow which is the basis of the
integrated LiDia BSI cartridge-based test. Methods: Contrived samples were
prepared by inoculating 10 ml of whole human blood with pathogens at 1 CFU/ml.
Clinical specimens were collected from hospitalized patients with BSI detected by
blood culture. Pathogens from contrived and clinical samples were extracted with
superparamagnetic beads conjugated to polyclonal antibodies raised against a
diverse panel of bloodstream pathogens. The DNA extracted from isolated
pathogens was identified via a nested PCR approach, in which the specific pathogen
gene targets and resistance markers were detected using time-resolved electronic
measurement of pH changes during amplification by ISFET sensors on a CMOS
sensor chip. Results: Contrived whole blood specimens were correctly identified as
positive for A. baumannii, P. mirabilis, K. pneumoniae, K. oxytoca, E. faecalis, E.
faecium, S. aureus, C. glabrata, and C. albicans spiked at 1 CFU/ml. The absence of
resistance markers in these samples was correctly confirmed in K. pneumoniae, E.
faecalis, E. faecium, and S. aureus. In addition, the presence of both pathogen and
resistance marker was confirmed for contrived whole blood specimens with
vancomycin-resistant E. faecalis and methicillin-resistant S. aureus (MRSA) spiked
at 1 CFU/ml. Corresponding whole blood samples with no input spike were correctly
identified as negatives. Clinical specimens of MRSA and methicillin-sensitive S.
aureus were correctly identified direct from whole blood via detection of S.
aureus and the presence or absence of mecA/C, concordant with trigger blood
culture results. Conclusions: We have demonstrated detection of bacterial and
fungal pathogens and resistance markers direct from whole blood in clinical
specimens and spiked samples at 1 CFU/ml. This modular workflow delivers
clinically actionable results in less than 3 hours and is the basis of DNAe's
automated walkaway LiDia platform. Once cleared or approved, the LiDia platform
will enable swifter treatment of BSIs with targeted therapeutics, improving patient
outcomes.
ID32. A Clinical Performance Evaluation of QPLEX STI Detection Kit
S. Cho
Seoul National University Hospital, Seoul, South Korea.
Introduction: Eight STI (Sexually Transmitted Infections) types genotyping kit was
developed based on the QMAP (QuantaMatrix multiplex assay platform) technology
which is suspension bead array system. Methods: This kit’s core technology is a
coded bead system for a multiplex bioassay. Coded beads were fabricated from
lithography process and tetra-ethyl-orthosilicate polymerization. After that the beads
were coated with silica shell and then carboxyl groups were generated chemically on
the surface of the beads. QuantaMatrix Inc. has also developed decoding software
and reader system. To assess the validity of the clinical detection of QPLEX STI
Detection Kit compared to AccuPower STI8A-Plex Real-Time PCR Kit (Bioneer,
Korea). Results: Eight microorganisms (C. trachomatis, M. genitalium, M.
hominis, U. urealyticum, T. vaginalis, N. gonorrhoeae, herpes simplex virus type1
and type2) were evaluated on 737 vaginal swabs and 673 urine samples. Between 2
kits, agreement was excellent regardless species both for the swab samples (kappa
= 0.982 ~ 1) and urine samples (kappa = 0.977 ~ 1). Conclusions: Clinical detection
of sexually transmitted diseases by QPLEX STI Detection Kit is as accurate as
AccuPower STI8A-Plex Real-Time PCR Kit. Our novel coded bead technology and
QMAP system provides multiplex solutions for molecular diagnosis of STI.
ID33. Rapid Detection of Respiratory Pathogens with GenMark’s ePlex RP
Panel
K. Henthorn, A. Schroeder, D. Osato, G. Borillo, P. Le, N. Mohamed, R.
VanSickle, J. Moberly, C. Hoeft, K. Wong, S. Kulkarni, A. Thornberg, L. FreemanCook
GenMark Diagnostics, Carlsbad, CA.
Introduction: The disease burden from respiratory infection is greater than that of
any other cause of disease (Mizgerd, J. P. 2006. Lung infection—a public health
priority. PLoS Med. 3(2):e76. doi:10.1371/journal.pmed.0030076). Communityacquired respiratory tract infections are a leading cause of hospitalization
especially in infants, elderly and immunocompromised patients. With viral and
bacterial infections having similar clinical presentations and more than one causative
agent potentially being involved, diagnosis can be challenging. To ensure
appropriate treatment of respiratory illness, rapid and accurate diagnosis is
980
important. The ePlex Respiratory Pathogen (RP) Panel is a qualitative nucleic acid
multiplex in vitro diagnostic test intended for use on the ePlex instrument. It
simultaneously detects 21 of the most common respiratory viral and bacterial
pathogens in nasopharyngeal swabs (NPS) in approximately 1.5 hours. The singleuse cartridge contains reagents required to enable sample-to-answer testing,
including sample extraction, PCR amplification, and detection via GenMark’s
proprietary eSensor technology. The ePlex RP Panel is for investigational use only.
Methods: To assess performance characteristics, analytical studies to determine
sensitivity, cross-reactivity and interference were conducted. In addition a
prospective, multicenter study was completed to evaluate the clinical performance of
the RP Panel in nasopharyngeal swab samples. The performance of the RP Panel
was compared to an FDA-cleared molecular respiratory panel as well as validated
PCR tests with bi-directional sequencing. Results: The limit of detection was
identified and verified in a minimum of 20 replicates for each target on the RP Panel
using quantified reference strains. Comprehensive inclusivity testing showed high
analytical reactivity and it was shown that the assay is able to detect mixed infections
with more than one clinically relevant organism. No potentially interfering substances
were identified and no cross-reactivity with other organisms was found at the
concentrations tested. Comparison of the results from the RP Panel in the clinical
study with the results from the FDA-cleared molecular respiratory panel and
validated PCR tests showed an agreement of > 92% in 2,908 samples after
discordant resolution. Conclusions: The ePlex RP Panel demonstrated high
analytical sensitivity and specificity in detecting a broad range of respiratory
pathogens and showed excellent clinical performance. The system provides rapid
results with minimal hands-on steps, helping to improve laboratory efficiency and
deliver faster results to the clinician for improved patient care.
ID34. Challenges Associated with Developing Rapid Molecular Diagnostics for
Detection of Antibiotic Resistance
M. Stonebraker, L.L. Malone, E. Grigorenko
Diatherix Laboratories, Huntsville, AL.
Introduction: Widespread use of antibiotics has led to an increased number of
multidrug resistant bacteria all over the world. New resistance mechanisms emerge
and spread globally, hindering the ability of healthcare professionals to treat common
infectious diseases. Because accurate and timely detection of antibiotic resistance
(AR) patterns are crucial for treatment of disease, as well as antimicrobial
stewardship, the field is moving toward molecular-based testing. However, the
development of novel diagnostic tests based on gene detection faces several
challenges including a multiplicity of mechanisms and gene targets, lack of common
nomenclature for gene classes, and a scarcity of organisms with well-defined
phenotypic and genotypic profiles. This study highlights these challenges with
emphasis on a subset of problematic gene targets. Methods: Extensive NCBI and
literature searches were performed to determine relevant gene targets and
corresponding gene sequences responsible for resistance to common classes of
antibiotics. For discrimination of extended-spectrum beta-lactamases (ESBL) and
non-ESBL TEM variants, 4590 sequences were used to design TaqMan assays
targeting 4 specific point mutations (E104K, R164S, G238S, E240K). For coverage
of relevant CTX-M enzymes, 4 assays were designed using 839 sequences. Two
nomenclature systems to identify aminoglycoside modifying enzymes were used in
primer design. Characterized bacterial isolates were collected from commercial
(n=231), public (n=376), and academic (n=113) sources. Results: Over 80 assays
that detect genes responsible for resistance to 8 commonly used classes of
antibiotics were developed. Out of 47 strains characterized for TEM ESBL mutations,
7 publicly-sourced isolates produced discrepant results, which were confirmed by
phenotypic testing. The CTX-M assays were inclusive of 69 variants that can be
found across 33 bacterial species, and assay specificity and inclusivity were
confirmed with 63 characterized isolates. For less common AR genes, isolates were
not available; therefore, assays for select aminoglycoside, macrolide, and quinolone
resistance genes were designed but could not be validated. Conclusions: Several
factors that can influence test development were identified in this study, illustrating
the challenges associated with the molecular detection of AR. The complexity of
mechanisms involved in AR is indisputable, and prevalence rates are needed to
discern the most important genes of interest. As the field moves toward molecularbased testing, it would be beneficial for commercial and public sources to have more
comprehensive, accurate genetic characterization of isolates, and a consensus on
gene nomenclature is required.
ID35. Quantitative Detection of HCV and HBV on NeuMoDx Molecular System
C. Couture, E. Craig, M. Carey, H. Lee, J. Zhu, M. Mastronardi, B. Wu, S.
Brahmasandra
NeuMoDx Molecular, Ann Arbor, MI.
Introduction: Determining HBV DNA or HCV RNA levels in serum and/or plasma is
an important tool to characterize viral loads in infected patients to monitor disease
progression, efficacy of antiviral therapies, as well as to detect drug resistant
mutants and identify relapse upon discontinuation of an antiviral therapy. The
NeuMoDx HCV and NeuMoDx HBV Tests are in-vitro diagnostic assays
incorporating a universal nucleic acid isolation chemistry enabling extraction of
qPCR ready RNA and DNA from serum and plasma specimens, combined with a
sensitive real-time PCR assay to deliver highly accurate results in a completely
automated manner on the NeuMoDx Molecular System. In addition, all reagents and
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
disposables are room temperature stable and are intended to remain on-board the
system to provide a seamless, on-demand testing workflow. Methods: The
NeuMoDx Molecular System automates and integrates the extraction, purification,
quantification, and results interpretation of infectious disease nucleic acid targets
using quantitative RT-PCR. The objective of this study was to test and report
performance of the NeuMoDx HCV and NeuMoDx HBV Tests in key analytical
performance metrics. Studies were performed to characterize the analytical
sensitivity, linearity, precision, inclusivity, turnaround time, and results processing
accuracy and are reported here. Results: Evaluation of the analytical sensitivity
using a Probit style analysis indicates that both the tests demonstrated excellent
limits of detection. The NeuMoDx HCV Test was able to detect ~12IU/mL (95% CI)
and the NeuMoDx HBV Test showed a detection limit of ~10IU/mL (95% CI). Both
the NeuMoDx HCV and NeuMoDx HBV Tests demonstrated excellent linearity
across typical, clinically relevant measurement dynamic ranges (R2 ˃ 0.99 across 7
logs), as well as reproducibility/precision across multiple systems, operators, and
reagent lots. In addition, the NeuMoDx HCV Test showed equivalent detection
performance across all relevant HCV genotypes. Turnaround time (TAT) for the
NeuMoDx HBV Test is ~65 min and the TAT for NeuMoDx HCV Test is ~75 min and
the results interpretation module incorporated for automated processing of data
provides extremely accurate results. A comparison study was also conducted
between the NeuMoDx HCV Test and the COBAS HCV TaqMan Assay using split
samples. Bias of HCV viral load output from the 2 tests was less than 0.5 Log IU/mL.
Conclusions: Based on these data, we believe that the NeuMoDx HCV and
NeuMoDx HBV Tests are extremely easy to use, rapid, automated molecular tests
for the sensitive, and accurate viral load monitoring required for effective patient
management of blood borne virus infections.
of respiratory pathogens in a 96-well format. BioCode Respiratory Pathogen Panel is
a molecular assay for detection of 20 respiratory pathogens including influenza
(Influenza A, H1, H3, H1N1pdm09, Influenza B), RSV (A, B), parainfluenza (type 1,
2, 3, 4A, 4B), metapneumovirus (A, B), rhinovirus, coronavirus (OC43, NL63, 229E)
and adenovirus. Methods: Biocode MDx 3000, automated system for PCR, postPCR sample handling and detection in a 96-well format, was used for the study. An
off board automated nucleic acid extraction system was used and viral targets were
amplified by one-step RT-PCR. PCR products were captured by target-specific
probes coupled to barcoded magnetic beads (BMBs), and fluorescent signal was
generated by incubation with a streptavidin conjugate. Qualitative results were
determined by Median Fluorescent Index (MFI) threshold for each analyte.
Results: Primers designed for each target were evaluated for amplification efficiency
by SyBr green and melt curve analysis. A serial dilution of target nucleic acid was
performed and tested. Primer pairs with high specificity and amplification efficiency
were selected and evaluated first in subpanels, and then in the multiplex panel
format. The Applied BioCode multiplex Respiratory Pathogen Panel detected
intended viral targets and did not cross-react with organisms tested in this study.
Negative controls showed no target signals, and well-to-well carry-over
contamination was not observed. Preliminary limit of detection was ≥ 102 TCID50/mL
for Influenza A, Influenza B, 2009 H1N1, coronavirus, metapneumovirus, RSV,
rhinovirus and parainfluenza viruses. Assay performance for 99 nasopharyngeal
swab samples showed 91% positive agreement with an FDA-cleared multiplex
assay. Conclusions: The BioCode Respiratory Pathogen Panel is highly specific for
its intended target viruses. This assay together with BioCode MDx 3000 system is
user friendly, simplifying the workflow and reducing contamination risk as well as
hands on time for the users.
ID36. Validation and Performance of Sequencing-Based Reference Assays for
Biocode Gi Pathogen Panel
A. Pham, D. Mantzke, C. Knoth, M. Aye
Applied BioCode, Inc., Santa Fe Springs, CA.
Introduction: Applied BioCode has developed an automated high-throughput
molecular diagnostic system. The BioCode GI Pathogen Panel is a 18-plex
molecular assay for detection of gastrointestinal pathogens including bacteria
(Campylobacter, C. difficile toxin A/B, Salmonella, Shigella/EIEC, EAEC, EPEC,
ETEC, shiga toxin-producing E. coli, E. coli O157, Vibrio, Yersinia enterocolitica),
viruses (norovirus group I/II, adenovirus F, rotavirus A), and parasites
(Cryptosporidium, Entamoeba histolytica, Giardia lamblia). To support the clinical
trials of GI Panel, we developed assays with PCR followed by sequencing
(Reference Assays) for method comparison for 7 targets of the BioCode GI Panel.
Methods: Two different SYBR Green PCR/Sequencing assays were validated for
each of the 6 targets while ETEC required 3 assays. NucliSENS easyMAG was used
for extraction. Targets were amplified with real-time SYBR Green PCR using ABI
7500 system. Presumptive positives, based on assay-specific Tm ranges, were
confirmed by bi-directional sequencing with ABI 3500 Analyzer. Sequence data was
analyzed with ABI Sequence Scanner Software and the Lasergene Software to
generate PHRED scores and contig length and ambiguous nucleotides, respectively.
NCBI BLAST of each contig is performed to generate Identity to Reference, Query
Coverage, and E-Value. Results: No cross-reactivity was observed with 89
organisms tested in this study. Limit of detection was determined for each assay
with both unpreserved stool and Cary-Blair stool: 5.0E+2 CFU/mL for both EAEC
assays, 1.0E+3 and 5.0E+3 CFU/mL for EPEC assays, 1.0E+4 CFU/mL for
ETEC (LT & ST1a assays), 1.0E+5 CFU/mL for ETEC (ST1b Assay), 5.0E+4
oocysts/mL for C. parvum (both assays), 1.0E+3 and 1.5 E+3 oocysts/mL
for Giardia assays, 5 cysts/mL for both E. histolytica assays, 1.0E+1 TCID50/mL for
Adenovirus 40 (both assays), and 1.0E+0 TCID50/mL for Adenovirus 41 (both
assays). Inclusivity study showed that each assay accurately detected different
strains representing various temporal, geographic and genetic diversity of target
pathogens. Retrospective clinical samples were used for the Method Comparison. A
total of 88 specimens (23 EAEC, 36 EPEC, 13 ETEC, 3 Adenovirus 40/41, 6 Giardia,
6 Cryptosporidium, and 1 E. histolytica) were tested and compared with results from
FDA-cleared assays. Agreement for EPEC, ETEC, Adenovirus, Giardia, and E.
histolytica targets were 100% while agreement for EAEC and Cryptosporidium were
90% and 67%, respectively. Conclusions: The Reference Assays for select GI
pathogens are sensitive, specific, and highly accurate, and can be used as
composite comparator PCR/Sequencing assays to rapidly detect and differentiate 7
select targets of the BioCode 18-Plex Gastrointestinal Pathogen Panel.
ID38. Comparison of Luminex ARIES Vaginosis Panel and BD AFFIRM VPIII for
the Detection of Candida spp., Gardnerella vaginalis, and Trichomonas
vaginalis
J. Barry, J. Barnett, J. Sizemore, A. Smith, S.S. Talwalkar
CPA Lab / Norton Healthcare, Louisville, KY.
Introduction: Luminex ARIES is a closed tube sample-to-answer real-time PCR
system. ARIES system also has ability to run ASRs by using either ASR primers,
ARIES extraction cassettes or ARIES Ready Mix. BD AFFIRM VPIII is one of the
most widely used methods for patients that are clinically suspected of having
vaginitis secondary to sexually transmitted disease. It is a DNA-probe based method
that can simultaneously detect 3 organisms commonly associated with
vaginitis; Candida albicans (CA), Gardnerella vaginalis (GV) and Trichomonas
vaginalis (TV). Objective of this study was to compare the 2 methods in terms of
overall sensitivity, ease of use & operational cost for detection of vaginosis panel
(VP) targets. Methods: ARIES uses Multicode RTx real-time PCR technology to
execute sample extraction, amplification & detection within 2 cassettes. VP
comprised of CA / Candida glabrata (CG) run in one ARIES extraction cassette and
GV/TV run in second cassette. Primers were loaded into ARIES Ready Mix tubes
prior to snapping them onto the ARIES extraction cassette. Both assays used 200µL
input volume. Results were reviewed using cycle threshold (Ct) and melt curve (TM)
analysis in the SYNCT software using the User Defined Protocol. Being an IVD
assay, AFFIRM was set-up per manufacturer package insert. Results interpretation
relied on an individual to observe visual color change within a probe analysis card
and compare it to the inbuilt positive control for each target. Eight months of data on
patient samples tested by both methods was reviewed and compared for this study.
Results: A total of 4773 samples were reviewed. ARIES technology provided a more
sensitive detection of VP targets than the BD AFFIRM VPIII. ARIES allowed for
speciation of Candida sp. since the ASR probes are specific for CA and CG. Table 1
outlines the sample positive rate by individual methods. Hands-on time was
comparable between the 2 methods (approximately 30-45 min for a batch of 6
samples). Based on average monthly volume of 550 samples, the operational cost
for ARIES was similar to AFFIRM
ID37. Development of a Respiratory Pathogen Panel with an Automated HighThroughput System
S. Mir, J. Granados, M. Aye
Applied BioCode, Inc., Santa Fe Springs, CA.
Introduction: Acute respiratory tract infections (ARIs) are the most common causes
of childhood morbidity and mortality worldwide, accounts about 30% of all childhood
deaths in developing world. However in developed countries, ARIs cause enormous
direct and indirect health costs (Hinman, A. R. 1998. Global progress in infectious
diseases control. Vaccine 16:1116-1121). Rapid identification of respiratory viruses
can help in avoiding indiscriminate use of antibiotics as well as large healthcare
costs. Using Applied BioCode’s proprietary barcoded magnetic bead (BMB)
technology, we are developing a multiplex molecular diagnostic assay for detection
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
Table 1: Positivity by Platform (n=4773)
Target
BD AFFIRM VPIII
Luminex ARIES
Candida sp.
11.78%
32.23%
G. vaginalis
45.14%
56.19%
T. vaginalis
4.59%
10.09%
Conclusion: The VP assay on ARIES is easy to perform and has better sensitivity
compared to BD Affirm with ability of also discriminating Candida sp. especially since
treatment differs. It also provides ability to add other ASR targets to the panel,
especially Atopobium vaginae which may be clinically more relevant. ARIES offers a
sample to answer solution with versatility, flexibility and scalability for a growing
molecular lab.
981
AMP Abstracts
ID39. Clinical Evaluation of the xMAP MultiFLEX ZIKA RNA Assay
A. Walden, K. Curless, V. Winn, R. Relich
Indiana University Health, Indianapolis, IN.
Introduction: Until recently, Zika virus (ZIKV) remained a relatively obscure
mosquito-borne pathogen that sporadically caused outbreaks of febrile illness in
tropical regions of the Eastern Hemisphere. In 2015, ZIKV was detected in clinical
samples from patients experiencing a dengue fever-like illness in Brazil and, since
then, sustained autochthonous transmission of ZIKV has been reported in nearly all
nations in the Western Hemisphere. Infection with ZIKV has been linked to
numerous sequelae, including birth defects and stillbirth. The diagnosis of ZIKV
infection is confounded by symptom overlap with numerous, similar diseases and by
the fact that antibodies raised against ZIKV cross react in serological tests with
antigens of related flaviviruses. To that end, a small number of commercial ZIKV
molecular diagnostics have been developed and granted emergency use
authorization (EUA). In this study, we evaluated the performance characteristics of
the xMAP MultiFLEX ZIKA RNA Assay. Methods: A combination of 74
serum/plasma and urine samples were tested for ZIKV RNA. Extraction was
performed on the bioMérieux NucliSENS easyMag instrument. RT-PCR (reverse
transcription – polymerase chain reaction) and subsequent steps were set up using
the Qiagen OneStep RT-PCR Kit and the Luminex xMAP MultiFlex Zika RNA Assay.
All samples were ran on the Luminex MAGPIX instrument. Limit of Detection (LOD)
and reproducibility were performed using the Exact Diagnostics Zika Verification
Panel. For specificity testing, 4 related flaviviruses and 4 strains of ZIKV were
passaged 1 time in Vero cells and, following harvesting, cell culture supernatants
were frozen until testing. Results: Seventy of the 74 samples tested were sent to a
reference lab for comparison testing. The correlation between in-house testing and
the reference lab testing was 100%. The LOD was found to be 100 cp/mL. There
was no variation in reproducibility between 3 technologists. Specificity study results
showed the assay detected RNA from all 4 cultured ZIKV strains, but did not detect
RNA from any of the non-ZIKV flaviviruses tested. Conclusions: Following the
emergence of ZIKV in the Western Hemisphere in 2015, the need for diagnostic
tools to detect this virus in human clinical samples emerged as well. Unfortunately,
clinical observation and traditional serological methods do not always provide a
definitive diagnosis. Through challenge with clinical samples from patients with
suspected Zika fever and with culture supernatants of various ZIKV and non-ZIKV
flaviviruses, we evaluated the performance characteristics of the xMAP MultiFLEX
ZIKA RNA Assay and determined it to be a very reliable test for the diagnosis of
ZIKV infection.
ID40. Evaluation of ELITech Group’s MGB Alert HSV-1 and HSV-2 ASRs on the
Abbott m2000sp/rt Platform
K. Stepaniants1, I. Ankoudinova1, M. Metcalf1, W. Mahoney1, D. Lucic2
1Elitech Group, Bothell, WA; 2Abbott Molecular, Des Plaines, IL.
Introduction: Herpes simplex virus 1 (HSV-1) is normally associated with orofacial
infections and encephalitis, whereas herpes simplex virus 2 (HSV-2) usually causes
genital infections and can be transmitted from infected mothers to neonates. In the
US, HSV is the most common cause of acquired, sporadic encephalitis. PCR
identification of HSV in plasma, spinal fluid and genital lesions has become a gold
standard diagnostic tool due to its sensitivity and improved turnaround time over
culture. Here we evaluate the implementation of the MGB Alert HSV -1 and HSV-2
ASRs reagents (ELITechGroup Inc Logan Utah) using the open mode and
Laboratory Developed Test (LDT) capabilities of the m2000sp/rt platform (Abbott
Molecular, Des Plaines, IL). Methods: Limit of detection (LOD) and assay linearity
were determined by using HSV commercially available panels from ExactDx (Ft.
Worth, TX). Clinical accuracy was assessed using CSF, oral and genital swab
specimens in comparison to ARUP test of record HSV-1 and HSV-2 assays. DNA
was extracted using an open mode m2000sp protocol (m2000-DNA-Urine-BA-20070 for 96 samples). Master mix was assembled using an open
mode m2000sp protocol with the following reaction volumes: 15uL eluate and 25uL
master mix (15e25m) Tfi (Thermo Fisher Inc., Carlsbad, CA). Amplification and
detection were performed using the m2000rt Qualitative Laboratory Defined
Application software. The following cycling conditions were applied for 45
cycles: 50°C 2min.; 93°C 2min.; 93 °C 15 sec.; 56 °C (fluorescence acquisition) 35
sec.; 72 °C 30 sec. Fixed threshold with value of 20,000 was used for result analysis
of HSV-1, HSV-2 and IC. Results: Analytical sensitivity (LOD) was determined to be
169 copies/mL for both HSV-1 and HSV-2 analytes. Linear regression of both assays
was R2>0.99. Positive Percent Agreement (PPA) for the clinical accuracy was
determined as positive when positivity was confirmed with MGB Alert HSV ASR on
the m2000 platform. Similarly, Negative Percent Agreement (NPA) was determined
as negative when negativity was confirmed with MGB Alert HSV ASR
on m2000platform. In this study, PPA and NPA were both 100% where 25 patient
samples were positive and 52 patient samples were negative. The average
turnaround time for the assay tested on the m2000 platform was 4.5hrs for 24
samples. Conclusions: MGB Alert HSV ASRs were successfully implemented on
the m2000 platform with excellent performance in comparison to ARUP test of
record. The ability to perform these markers on the m2000 platform as either
qualitative STI markers or quantitative markers for immunocompromised patient
populations would allow laboratories to consolidate testing, reduce send-out costs,
and in turn improve turnaround times.
982
ID41. Comprehensive Women’s Health Diagnostic Testing Using Innovative
Multiplex PCR Assays
W. Hauser1, G. Wang2, D. Schapfel1
1Enzo Clinical Laboratories, Farmingdale, NY; 2Westchester Medical Center,
Valhalla, NY.
Introduction: Vaginitis includes a wide array of conditions and is a general
diagnosis, but it is one of the most common reasons medical care is sought by
women. The exact incidence of vaginitis is not known, however it is so common that
most women experience at least one episode in their lifetime, and more than half
have multiple episodes. Bacterial vaginosis (BV), vulvovaginal candidiasis, and
trichomonas account for 70% of vaginitis diagnoses, with 40-50% of those being BV,
20-25% being vaginal candidiasis, and 15-20% being trichomoniasis. Proper
identification of the cause of vaginitis is imperative to manage the infection effectively
and avoid treatment failure, which can lead to prolonged symptoms for the
patient. Conventional microbiology methods for vaginitis diagnosis can require
extended turnaround times and can utilize subjective decision making points, which
leaves a strong need for faster, more specific, and more sensitive molecular based
vaginitis testing. Methods: The AMPIPROBE Candida, NG/CT/Trich, Bacterial
Vaginosis, and UP/Mg/Mh Assay Kits (Enzo Life Sciences, Farmingdale, NY)
provide multiplex real-time PCR testing that utilizes paired fluorophore- and
quencher-labeled primers. Each assay was evaluated for sensitivity, specificity,
reproducibility, and accuracy with experiments designed in accordance with the
requirements set forth by the New York State Department of Health Microbiology
Molecular Checklist (02/2011). The analytic performance for each assay was
examined using the QIAGEN QIAsymphony SP and Rotor-Gene Q systems with
APTIMA vaginal swab specimens spiked with whole organisms, genomic DNA, or
plasmid copy targets. Results: The limits of detection (LOD) for the
AMPIPROBE Candida assay in APTIMA vaginal swab specimens by a 95% Probit
analysis ranged from 10.5 CFU/mL to 130.4 CFU/mL for the C. albicans, C. glabrata,
C. krusei, C. parapsilosis, and C. tropicalis targets. The LOD for the AMPIPROBE N.
gonorrhoeae/C. trachomatis/T. vaginalis assay was 2.1 cells/mL, 13.6 EB/mL and
1.4 cells/mL respectively. The LOD was 389.6 CFU/mL for the Lactobacillus spp.,
320.1 copies/mL for the A. vaginae, 237.6 copies/mL for the G. vaginalis, 755.8
copies/mL for the Megasphaera, and 538.6 copies/mL for the BVAB2 targets of the
AMPIPROBE bacterial vaginosis assay. Finally, the AMPIPROBE UP/Mg/Mh
assay had an LOD of 13.1 CFU/mL for U. parvum, 82.8 CFU/mL for U. urealyticum,
65.3 copies/mL for M. genitalium, and 58.5 copies/mL for M. hominis.
Conclusions: The AMPIPROBE technology can provide a new proprietary approach
to create clinically sensitive and specific multiplex PCR assays and, when run in
tandem, serve as a comprehensive panel for the accurate detection and
identification of vaginitis associated microorganisms.
ID42. Sensitive Detection of Bacterial Targets on the NeuMoDx Molecular
System
J. Zhu, M. Carey, L. Gong, C. Couture, K. Rutila, B. Wu, M. Mastronardi, S.
Brahmasandra
NeuMoDx Molecular, Ann Arbor, MI.
Introduction: The NeuMoDx Molecular System is a revolutionary molecular
diagnostic platform providing a continuous loading, sample to answer workflow with
on-board, ambient storage of all reagents and disposables. The System incorporates
a universal nucleic acid extraction chemistry for isolation of DNA from a wide variety
of clinical specimens, combined with a sensitive real-time PCR module to provide a
seamless workflow for on-demand, random access in-vitro diagnostic testing for
single and multi-target implementations. The performance of this System for DNA
target detection was evaluated with 2 representative assays – the NeuMoDx GBS
Test (Group B Streptococcus) and the NeuMoDx CTNG Test (Chlamydia
trachomatis/Neisseria gonorrhoeae). Methods: Analytical studies – sensitivity, crossreactivity, inclusivity, interfering substances, and a pilot method comparison study –
were performed to elucidate the typical performance metrics for both tests. Testing
was performed with clinical urine specimens for CTNG, and clinical enriched Lim
broth for GBS. Analytical sensitivity was determined via a Probit style analysis;
additionally, sensitivity assessment of the NeuMoDx CTNG Test included
demonstration of low level detection of each target in the presence of a high level
challenge of the other target. Specificity of both tests was confirmed by screening
>125 known organisms either phylogenetically similar to CTNG and/or GBS or
typically cohabitating in these specimens. Both tests were also performed in the
presence of typical interfering moieties to assess the efficacy of the nucleic acid
isolation process in preparing real-time PCR ready DNA. Finally, a small scale
method comparison study involving 100 specimens was performed for each of the
tests to assess concordance with reference tests. Results: The NeuMoDx CTNG
Test demonstrated a LoD of 10 EB/mL for CT and 3 CFU/mL for NG in urine, while
the NeuMoDx GBS Test exhibited a LoD of 256 CFU/mL. Sensitivity of neither CT
nor NG target was affected by the presence of high titers of the complementary
target. Both tests showed no cross-reactivity across the >125 organisms screened,
and demonstrated robustness in the presence of interfering agents. Turnaround time
(TAT) for the NeuMoDx GBS Test was <60 min and the TAT for the NeuMoDx
CTNG Test was <65 min. Both assays showed excellent inclusivity by correctly
detecting all clinically relevant variants tested with no significant variation in
sensitivity. Finally, excellent positive and negative concordance was demonstrated in
the method comparison study performed for both tests. Conclusions: The
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
NeuMoDx Molecular System proved to be a very easy to use, reliable, and rapid
method to screen for bacterial pathogens using single or multi-target assays.
ID43. Direct Detection of VZV from Cutaneous, Mucocutaneous and CSF
Specimens Using the Simplexa VZV Direct Assay on the DiaSorin LIAISON
MDX System
E. Eleazar, A. Bologa, J. Hugo, P. Naranatt, M. Tabb
DiaSorin Molecular, LLC, Cypress, CA.
Introduction: Varicella zoster virus (VZV) is the infectious agent responsible for
causing chicken pox, with primary infection usually occurring in childhood. Upon
reactivation, VZV also causes herpes zoster or shingles. VZV is also associated with
other harmful conditions including Mollaret's meningitis, inflammation of arteries in
the brain leading to stroke, myelitis and death. Prompt detection of VZV infections
can assist in patient management, and is especially important in cases of debilitating
conditions. The Simplexa VZV Direct assay is in development as a sample-toanswer detection system performed on the DiaSorin LIAISON MDX system. Swab
samples collected in viral transport media (VTM) or cerebrospinal fluid (CSF)
specimens are loaded directly onto a Direct Amplification Disc without extraction or
other specimen preparation. The goal of this study was to compare the performance
of the Simplexa VZV Direct assay relative to the DiaSorin VZV ASR used with
extracted specimens and a FDA-cleared nucleic acid amplification test (NAAT) that
requires pre-treatment of specimens. Methods: Relative sensitivity and specificity
were determined by testing panels of de-identified patient specimens and comparing
results with those obtained via NAAT methods requiring nucleic acid extraction or
sample pre-treatment. Whole blood interference was evaluated by utilizing contrived
samples containing the VZV Ellen Strain at 3X LoD. Cross-reactivity was tested
using a panel of 53 bacteria and viruses. Analytical reactivity was performed by
testing 5 different VZV strains. Performance of several types of VTM (M4, M4RT,
M5, M6 and UTM) were evaluated using the VZV Ellen Strain at 3X LoD.
Results: Relative sensitivity and specificity for CSF samples was 95.5% (21/22) and
100% (90/90), respectively. Relative sensitivity and specificity for all swab samples
was 100% (58/58) and 98.8% (81/82), respectively. The whole blood interference
study demonstrated the assay’s ability to detect VZV Ellen Strain at 3X LoD in either
UTM or synthetic CSF with up to 15% K2 EDTA whole blood (v/v). No crossreactivity was observed for the bacteria and viruses tested and the assay detected 5
different VZV strains. The VZV Ellen Strain was detected at 3X LoD in all 5 types of
VTM. Conclusion: The Simplexa VZV Direct assay was capable of directly detecting
VZV from un-extracted clinical specimens with performance comparable to either the
DiaSorin VZV ASR with upfront nucleic acid extraction or another FDA-cleared NAAT
assay that uses upfront sample treatment. The assay and instrumentation provide a
compact system for rapid (~1 hour) detection of VZV directly from cutaneous,
mucocutaneous and CSF samples.
ID44. Validation of Human Papilloma Virus Detection in Anal Cytology
Specimens on the Cobas 4800 System
L. Helander1, S. Favaloro2, C. McGowin1, G. Love1, M. Hagansee1, C. Kletecka2
1Louisiana State University Health Sciences Center, New Orleans, LA; 2Southeast
Louisiana Veterans Healthcare System (SLVHCS), New Orleans, LA.
Introduction: Human papillomavirus (HPV) is the most common sexually
transmitted infection worldwide and is linked to cervical, anal and oropharyngeal
cancers. The introduction of cervical screening has significantly reduced the
incidence of HPV related cervical cancers, while HPV-related anal and
oropharyngeal cancer rates continue to rise. Men who have sex with men (MSM)
have a 15-35 fold increased risk of developing anal cancer when compared to the
general population. Currently, there are no definitive screening guidelines for anal
carcinoma. Cytologic screening for precancerous lesions is used, but there is
considerable inter-observer variability in interpretation, with a higher incidence and
lower specificity of atypical squamous cells of undetermined significance (ASC-US)
diagnosed in these specimens. Early data suggests that cytologic screening of highrisk, HIV negative individuals particularly may be useful in triaging to more invasive
procedures and HPV testing may provide some utility in screening. Due to the large
MSM population at this treating facility, validation of high risk HPV (HRHPV) in anal
cytology specimens was undertaken. Methods: Validation was performed using the
Roche cobas 4800 HPV test, a multiplex real-time polymerase chain reaction (PCR)
and nucleic acid hybridization test that concurrently detects HPV16 and HPV18 as
individual reactions, and HPV31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68 as a
pooled result under Other HRHPV (OHR). Beta-globin is used as a
control. Detection was performed on 41 anal cytology samples from the local HIV
Outpatient Clinic. A comparative assessment using the same sample was
performed by a reference laboratory, also using the Roche cobas 4800 HPV testing
platform. HPV spiking experiments were performed to further assess assay
performance. Results were analyzed using Cohen’s kappa
statistics. Results: Sensitivity and specificity were 100% and 94.1%, respectively, for
all HRHPV types between the Roche cobas 4800 HPV testing platforms (HPV16:
100% and 94.1%; HPV18: 100% and 100%; OHR: 100% and 88.9%). One anal
specimen was HPV-negative at the reference laboratory and HPV16 and OHR
positive in our testing. A second specimen deemed only HPV16 positive by the
reference laboratory was both HPV16 and OHR positive in our testing. Five
specimens were excluded due to an inability of the reference or our facility to obtain
valid results. HPV spiking experiments showed excellent inter-assay reproducibility
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
with 100% agreement between expected and obtained results. Kappa coefficients
were classified as good to excellent for all comparisons. Conclusions: The results
confirm that our Roche cobas 4800 HPV test can successfully detect HRHPV from
collected anal cytology specimens.
ID45. Multicenter Evaluation of Cobas HCV Real-Time PCR Assay on the Roche
Cobas 4800 System in Comparison with COBAS AmpliPrep/COBAS TaqMan
HCV Test: Leading Circle for Cobas 4800 Virology (LCCV) Project
W. Lee1, J. Baek1, H. Kim2, M. Hur2, E. Bae3, K. Lee4
1Kyung Hee University Hospital at Gangdong, Seoul, South Korea; 2Konkuk
University School of Medicine, Seoul, South Korea; 3Veterans Healthcare System
Medical Center, Seoul, South Korea; 4Gangnam Severance Hospital, Seoul, South
Korea.
Introduction: Hepatitis C virus (HCV) nucleic acid amplification test (NAAT) is
important for diagnosis and management of HCV infection. We evaluated analytical
performance of the newly launched cobas HCV (Roche Diagnostics GmbH,
Mannheim, Germany) on the cobas 4800 System in comparison with COBAS
AmplPrep/COBAS TaqMan HCV Test (CAP/CTM HCV) in 4 clinical laboratories.
Methods: Precision was evaluated using 3 levels of cobas HBV/HCV/HIV-1 Control
Kit, and linearity was evaluated across the anticipated measuring range (15.0 IU/mL
to 1.0 x 108 IU/mL) at 7 levels using clinical samples. Correlation using 171 clinical
samples, turn-around-time (TAT) for 24 tests, and limit of quantification (LOQ)
according to the processing volume were compared between the 2 systems.
Results: Within-laboratory imprecisions (standard deviation [SD] and coefficient of
variation [%CV]) ranged from 0.03 to 0.08 and from 0.70% to 1.76%, respectively.
Between-laboratory imprecisions was < 0.06 and < 1.38%, respectively. Linearity
(coefficient of determination, R2) ranged from 0.992 to 0.998, and overall correlation
was very high (r = 0.997) between the 2 systems. TAT in cobas 4800 System and
COBAS CAP/CTM were 3.5 hours and 5 hours, respectively. LOQ for cobas HCV
and CAP/CTM HCV were < 15 IU/mL with 400 uL and/or 500 uL. Conclusions: The
new cobas HCV real-time PCR assay on the cobas 4800 System showed reliable
analytical performance and improved workflow. It could be a good laboratory option
for HCV NAAT.
ID46. Comparison of Cobas HCV GT Against Versant HCV Genotype 2.0
Assays with Confirmation by Sequencing
T. Png, F. Yusrina, C. Chua, M. Khoo, L. Chiu, C. Lee, G. Lee, B. Yan, H. Lee, H.
Lee
National University Hospital, Singapore.
Introduction: Inaccurate hepatitis C virus (HCV) genotyping may lead to
unnecessary failed therapy or high cost due to retreatment. The gold standard in
identifying the genotype is by sequencing the HCV genome. However, this method is
laborious and impractical for large-scale testing, especially in the setting of a clinical
laboratory. Here, the cobas HCV GT (Roche Molecular Diagnostics) was evaluated
and compared with the Versant HCV Genotype 2.0 Assay (LiPA; Siemen
Healthcare). Methods: A total of 100 archived serum samples were used in this
study. All samples were received at the National University Hospital for routine
diagnostic testing between March and July 2016. In the event of discrepant results
produced by LiPA and cobas or indeterminate result produced by either assay, the
core and NS5B regions of the HCV genome were sequenced using a published
method. Results: The LiPA assay successfully genotyped 86 (86%) out of 100
samples. In contrast, the cobas successfully genotyped 73 (73%) samples. Overall,
the assays produced concordant results for 66 (66%) samples. Notably, all the 7
samples with genotype 1 with unavailable/inconclusive subtyping results were either
genotype 3 (1, 14%) or genotype 6 (6, 86%) by sequencing analyses. Moreover, 2
samples were incorrectly genotyped as types 4 (Sequencing: type 3; cobas:
indeterminate) and 6 (Sequencing: type 3; cobas: type 3), respectively, highlighting
the assay limitation in misclassifying the HCV genotype. Of the 99 samples
successfully tested on the cobas assay, 25 (25%) and 1 (%) results were deemed
indeterminate and invalid, respectively. The remaining 73 results have agreed
completely with LiPA and/or sequencing results. Upon performing sequencing, 18
(72%) of the 25 indeterminate samples were determined as either subtype 6m, 6n,
6v, or 6xa. The remaining 4 consisted of 3 genotype 6 with unknown subtype and 1
possible recombinant virus from subtypes 1b and 6v. Conclusions: The cobas
assay is based on genotype-/subtype-specific real-time PCRs for qualitative
identification of genotypes 1 to 6 and subtype 1a and 1b. However, the identification
of genotype 6 by this assay is only limited to subtypes 6a and 6b, resulting in the
relatively high amount of indeterminate results (25%) in this study population.
However, the cobas assay showed good specificity performance with a high level of
automation, with much faster and more reliable results than the LiPA assay. These
indeterminate results can be re-tested using the core and/or NS5B sequencing
methods described in this study. The initial screening by cobas followed by
sequencing of indeterminate samples should minimize unnecessary failed therapy or
high cost incurred due to wrong genotyping results.
983
AMP Abstracts
ID47. A Multi-Center Clinical Evaluation of a Sample to Answer Real-Time PCR
Assay for Toxigenic C. difficile in Symptomatic Subjects
S. Young1,7, P. Pancholi2,7, R. Widen3, B. Schmitt4, R. Dunn5, A. Drain5, A.
Weissfield6
1Tricore Reference Laboratories, Albuquerque, NM; 2The Ohio State University,
Columbus, OH; 3Tampa General Hospital, Tampa, FL; 4Indiana University,
Indianapolis, IN; 5Luminex, Austin, TX; 6Microbiology Specialist, Inc., Houston,
TX; 7University of New Mexico, Albuquerque, NM.
Introduction: C. difficile is an intestinal bacterium that causes symptoms ranging
from mild diarrhea to life-threatening inflammation of the colon. C. difficile infection
(CDI) often occurs as a complication of antibiotic treatment and is the most common
hospital acquired bacterial infection. The Luminex ARIES C. difficile Assay is a
qualitative real-time PCR assay for detection of toxigenic C. difficile in raw stools.
The assay is used with ARIES Systems, multiplex test systems capable of
automated nucleic acid extraction and purification from a clinical sample, real-time
PCR detection and data analysis in less than 2 hours. In this study, we assessed the
clinical performance of the ARIES C. difficile Assay in prospectively collected, deidentified, remnant, stool specimens from patients suspected of having CDI.
Methods: The clinical performance of the ARIES C. difficile Assay was evaluated on
984 stool specimens prospectively collected from October 2016 to February 2017 at
4 US clinical sites. All specimens were tested by both ARIES C. difficile Assay and
reference method (direct and enriched toxigenic culture). Reference method testing
was performed at a centralized facility. ARIES C. difficile Assay testing was
performed at each clinical site on specimens that were either kept refrigerated (28°C) or stored frozen (-70°C) prior to testing. Results: Nine hundred seventy-nine
(979) of 984 specimens tested (99.5%) generated valid ARIES C. difficile Assay
results. Five specimens were invalid by ARIES C. difficile Assay upon re-test. Clinical
sensitivity of ARIES C. difficile Assay against direct toxigenic culture was 98.1%
(103/105; 95% confidence interval [CI], 93.3%-99.8%). When compared to direct and
enriched toxigenic culture, sensitivity of the ARIES C. difficile Assay was 90.5%
(133/147; 95% CI, 84.5%-94.7%). Clinical specificity of the ARIES C. difficile Assay
against direct and against direct and enriched toxigenic culture were 92.6%
(809/874, 95% CI, 90.6%-94.2%) and 95.8% (797/832, 95% CI, 94.2%-97.1%),
respectively. Conclusion: The ARIES C. difficile Assay is a sensitive and specific
diagnostic tool for the detection of toxigenic C. difficile in stool specimens from
subjects suspected of having CDI.
ID48. Evaluation of the Focus Diagnostics Simplexa HSV 1 & 2 Direct for
Detection and Differentiation of Herpes Simplex Virus 1 and 2 in Neonatal
Swab Specimens
K. Gvozdjan1, A.T. Harrington2
1University of Illinois Hospital and Health Sciences System, Chicago, IL; 2Loyola
University Medical Center, Maywood, IL.
Introduction: Herpes simplex virus (HSV) testing is routinely performed for
newborns at risk of exposure to the virus during labor and vaginal delivery. Though
polymerase chain reaction (PCR) has been accepted as the method of choice for
HSV detection in cerebrospinal fluid, viral culture is still the recommended method
for HSV detection in eye, mouth, and nasopharyngeal swabs. Since HSV can take
from 2-5 days to grow in culture, PCR may serve as a more efficient alternative for
virus detection in the neonatal period. Therefore, the purpose of our study was to
determine the concurrence rate of HSV detection using viral culture versus Simplexa
HSV 1 & 2 Direct, and to determine the limit of detection (LoD) of the PCR-based
method in detecting 4 different strains of HSV in neonatal swab specimen matrix.
Methods: Twenty-eight eye, mouth, and nasopharyngeal swab specimens from
neonates, 1-30 days old, tested for HSV 1 and 2 by culture for clinical diagnostic
purposes from January 1, 2016 - May 31, 2016 were included in the study. All
specimens were kept frozen in viral transport media at -70°C. The specimens were
run on Simplexa HSV 1 & 2 Direct following the manufacturer’s instructions. The
results obtained by PCR and viral culture were compared to determine the
concurrence between the 2 methods. Specimens negative for HSV by PCR and
culture were used as matrix for serial dilutions of titered viral stocks. The highest
dilution at which 12 out of 12 replicates were detectable and the initial 50% tissue
culture infectivity dose (TCID50) were used to determine the LoD for each
strain. Results: All (100%) of the neonatal swab specimens tested were negative for
HSV 1 and 2 by culture and Simplexa HSV 1 & 2 Direct. HSV 1 and HSV 2 strains
were correctly differentiated in 100% of the specimens spiked with known HSV
strains. LoD values (TCID50/mL) for HSV-1 McIntyre, HF, MS, and G were 0.034,
0.5, 0.05, and 0.018, respectively. Ct values (mean ± SD) for HSV-1 McIntyre, HF,
MS, and G were 36.4 ± 0.67, 35.9 ± 0.52, 35.2 ± 0.42, and 36.0 ± 0.62,
respectively. Conclusions: The Simplexa HSV 1 & 2 Direct assay demonstrated
high specificity, sensitivity, and reproducibility in detection of the 4 HSV
strains. Given that results may be reported within a few hours of collection and
therefore facilitate prompt decision making regarding patient treatment and
discharge from the hospital, the assay offers a promising alternative to viral culture in
evaluation of neonatal swab specimens for suspected HSV infection.
984
ID49. Performance of NxTag RPP Assay from Luminex
L. Mazur, B. Baltadjieva, A. Gandhi, J. Chudvasvimol, M. Patel, S. Ahmed, L.
Sok, M. Mihalov
ACL Laboratory, Rosemont, IL.
Introduction: ACL laboratory is offering NAT Respiratory pathogen testing all year
round. Annual Respiratory Pathogen testing volume has been increasing an average
of 15% per year. Clinical test demand during influenza seasons is creating additional
challenges, which require providing testing multiple times per shift. High sample
influx during that time may reach 150-250 samples a day. In order to process high
volume of samples and meet average turnaround time of 12-24 hours ACL
laboratory evaluated performance of NxTag RPP assay from Luminex. ACL study
included also an assessment of workflow variables such as; TAT, instrument
maintenance and contamination risk. Methods: Two hundred eighty-nine previously
tested clinical samples collected in Eswab and UTM media were extracted utilizing
EasyMag instrument and King Fisher. Method 1. Samples tested by Luminex IVD
xTAG RVP multiplex assay with liquid bead array detection by LX 200
instrument. Method 2. Luminex IVD FDA NxTAG RPP multiplex assay with liquid
bead array detection by MagPix instrument. All discrepant or multiple-positive
samples were repeated and tested additionally by FilmArray instrument using RPP
assay from BioFire. Chlamydophila pneumoniae and Mycoplasma pnuemoniae were
tested and compared to ACL in-house LDT assays. Results: IVD FDA NxTAG RPP
multiplex assay demonstrated good agreement with previous version xTAG RVP:
289 samples generated 6069 individual test results: Clinical sensitivity was 97.4
(combined for all positive targets), Clinical specificity was 99.6 (all negative
targets). Overall NxTAG RPP assay detected 19 additional targets. They were
confirmed by alternative method (BioFire); NxTag RPP assay missed 9 targets.
Precision and reproducibility was assessed by testing of 5 pools QC reagents
prepared from Zeptometrix reagents. All pools were run 20 times; NxTag RPP assay
reproducibility was 99.2. Precision was as follow; lot-to-lot %cv (9-35), tech-to-tech
%cv (10-36), instrument-to-instrument %cv (9-39), intra run %cv (10-33). Coinfection rate for flu season 2016/17 showed; dual targets detection at 6.3%, triple
targets at 0.9% and quadruple targets at 0.1%. The NxTAG-RPP assay
demonstrated: 2 hours shorter TAT and less hands-on time allowing multiple runs
per shift, equivalent time for instrument maintenance; and lower area contamination
rate (wipe test positivity rate). Conclusions: This study demonstrates that: IVD FDA
NxTAG RPP multiplex assay from Luminex Diagnostics performed very well in
comparison to previous version of xTag RVP assay. The performance
characteristics, improved workflow and lower level of contamination are good fit for
ACL Laboratory high volume testing.
ID50. Evaluation of the Anyplex MTB/NTM Real-Time Detection (V2.0) for
Detection of Nontuberculous Mycobacteria in Respiratory Specimens
H. Kim
Hallym University Medical Center, Anyang, Gyeonggi-do, South Korea.
Introduction: Molecular detection of nontuberculous mycobacteria (NTM) is
increasingly important. The Anyplex MTB/NTM Reat-time Detection (Seegene) is a
CE-marketed test for direct detection of both MTBC and NTM. Recently a new
version (V2.0) of the Anyplex MTB/NTM was developed. The Anyplex MTB/NTM
Reat-time Detection (V2.0) is an upgraded version for NTM detection. In this study,
the sensitivity and specificity of the Anyplex MTB/NTM Reat-time Detection (V2.0)
were evaluated for detecting NTM. Methods: Two hundred and nineteen NTM
culture positive respiratory specimens and 104 AFB stain and culture negative
respiratory specimens were evaluated. Real-time PCR was performed by the
Anyplex MTB/NTM Reat-time Detection (V2.0). Results: The sensitivity and
specificity of the Anyplex MTB/NTM Reat-time Detection (V2.0) were 75.3%
(165/219) and 98.1% (102/104) for NTM detection, respectively. Conclusions: The
Anyplex MTB/NTM Reat-time Detection (V2.0) could be useful for detecting NTM in
respiratory specimens.
ID51. Direct Detection of mRNA in Whole Blood Samples for Transcriptomic
Profiling
A. Khine, R. Yuan, V. Parmar, S. Talebpour, T. Alavie
Qvella Corporation, Richmond Hill, Ontario, Canada.
Introduction: The transcriptomic profile of leucocytes has been demonstrated to
enable stratification of patient population according to the underlying cause of
disease. In cases of some life threating diseases, such as sepsis, early stratification
can greatly impact the course of treatment. The purpose of the present study was to
demonstrate the ability of a novel sample preparation method to prepare and
process a whole-blood sample ready for mRNA detection via RT-PCR without
requiring nucleic acid extraction. Cationic detergents are mixed with a whole blood
sample to lyse blood cells and form a complex with released nucleic acids, thereby
preventing gene induction and RNA degradation. The subsequent processing of the
complexes’ releases nucleic acids in a medium which is ready for real-time RT-PCR.
The measured threshold cycles are used for profiling the transcription of a gene set.
Methods: Human whole blood was collected into PAXgene Blood RNA tubes and
white blood cell mRNA was stabilized instantly. One mL of sample from the
PAXgene tube was processed by centrifugation to remove cellular debris and
harvest the detergent-nucleic acid complexes. Stabilized mRNA was subsequently
released in 100 µL of phosphate buffer by simultaneously subjecting the complexes
to high amplitude pulsed electric fields and flash heating, thereby providing a
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
processed sample ready for mRNA detection via RT-PCR without requiring nucleic
acid extraction. The mRNA targets of selected genes in the final sample were
detected and identified by RT-qPCR to demonstrate the performance of the method
in quantitative transcriptomic profiling. Results: The mRNA targets were detected in
less than an hour from the whole blood sample. The performance of the method was
demonstrated by measuring the expression levels of a panel of infection markers
relative to housekeeping genes hypoxanthine phosphoribosyltransferase 1 (HPRT1)
and heat shock protein 90 alpha family class B member 1(HSPCB). The
reproducibility of the measured relative threshold cycle values were determined to be
<1 cycle by measuring the standard deviations over different sample preparations
across multiple samples from different donors. Conclusions: A new sample
processing method for use in the quantitative measurement of transcriptomic profiles
of genes in leucocytes of whole blood samples was developed. The performance of
the method demonstrates its capability of potentially enabling disease diagnosis and
prognosis employing biomarker panels. The precision of the quantification may help
to avoid specificity loss due to the error accumulation across large biomarker panels.
ID52. Comparison Between BD Maxwell VP and ACL LDT SwabOne Assay
L.J. Mazur, B. Baltadjieva, I. Chaplyk, E. Ochoa, L. Ma, M. Patel, V. Hernandez, L.
Sok, M. Mihalov
ACL Laboratory, Rosemont, IL.
Introduction: Bacterial vaginosis (BV) is the most commonly reported
microbiological syndrome among women of childbearing age. BV is characterized by
a shift in the vaginal flora from the dominant Lactobacillus to a polymicrobial flora.
Symptomatic BV can be described as a syndrome based on the presence of a
collection of clinical features without a specific etiologic agent defined. The diagnosis
of BV is usually made using a series of clinical criteria collected by a clinician
performing a pelvic examination, or by interpretation of vaginal fluid Gram stains
using Amsel’s clinical criteria or Nugent score. Data emerging from molecular
investigations suggest that PCR-based strategy for the diagnosis for BV is the most
sensitive and specific. This study presents comparison between IVD BD Max Vaginal
Panel and ACL LDT SwabOne assay. Methods: Two hundred twenty-five clinical
samples (Affirm 114, Thin prep plasma 70 and UTM 41) were extracted using
NucliSENS EasyMAG protocol. Samples were characterized by Nugent score prior
to testing by PCR methods. Method1. BD MAX Vaginal Panel IVD assay from
Beckman Dickenson and run on BD MAX System. Method2. ACL LDT SwabOne
multiplex assay based on liquid bead array run on MagPix from Luminex. All
discrepant samples were run by one of the following methods; qPCR or sequencing,
or in case T. vaginalis by TMA Hologic). Analytical sensitivity and LOD were
determined by run dilutions of the bacterial targets from ZaptoMetrix, ATCC and
synthetic oligos from ITD diluted in UTM and Eswab matrix. Results: ACL LDT
SwabOne assay assays demonstrated good agreement with the comparator method:
Bacterial vaginosis (BV) -129/135 (95.5%) positive samples were detected and 84/85
(99%) were not detected by each method. T. vaginalis (TV) - 4/9 (45.4%) positive
samples were detected BD Max VP missed 5 positive samples, 216/221 (97.7%)
were not detected by each method. Candida 53/59 (89%) positive samples were
detected ACL SwabOne assay missed 5 positive samples, 163/166 (98.2%) were
not detected by each method. ACL SwabOne assay detected 4 targets; M.
genitalium - 0/0 positive samples and 225/225 (100%) negative samples, M. hominis
– 78/79 (98.7%) positive samples, 145/146 (99.3%) were not detected, U.
urealyticum – 113/114 (99%) positive samples, 109/111(98.2%) were not detected,
U. parvum 40/42(95.2%) positive samples, 183/183 (100%) were
detected. Conclusions: This study demonstrates that ACL LDT SwabOne assay
performed very well in comparison to comparator IVD assay and based on
performance characteristics is suitable for clinical testing of patient population.
ID53. Performance Evaluation of the Abbott RealTime CMV IUO Assay on the
m2000 Platform Compared to the Roche COBAS AmpliPrep/TaqMan CMV
Assay in Transplant and Immunocompromised Individuals
P.M. Kulling, D.C. Dirks, M.E. Waddell, M.D. Poulter
University of Virginia, Charlottesville, VA.
Introduction: Human cytomegalovirus (CMV) is associated with increased mortality
and morbidity in solid organ transplant (SOT) recipients, hematopoietic stem cell
transplant (HSCT) recipients and immunocompromised individuals. Accurate and
timely identification of CMV is critical to guide preemptive strategies, diagnose onset
of conditions, and monitor response to therapy, as well as detect reactivation of
latent virus. However, variability between methodologies used to quantify CMV,
despite the availability of a WHO International Standard (IS), has made it challenging
to standardize clinical cutoffs for patient intervention. Our study compares
performance of the Abbott Real-Time (ART) CMV Investigational Use Only (IUO)
assay performed on the m2000 to the Roche COBAS AmpliPrep/Taqman CMV test
(CAP/CTM). Methods: Assay linearity and precision were assessed using a
commercially available linearity panel obtained from ExactDx (Ft. Worth,
Texas). Clinical comparison between ART and CAP/CTM assays was assessed
using 150 retrospective human EDTA plasma samples from SOT recipients, HSCT
recipients, and non-transplant immunocompromised individuals. Of these, 90
patients had previously detectable or quantifiable CMV by the CAP/CTM assay while
60 patients had no previously detectable CMV. Results: ART exhibited excellent
linearity (R2=0.9933) and within/between run precision (standard deviation of ≤0.13
log IU/mL; range 0.003 to 0.128 IU/ml). ART demonstrated increased viral load
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
sensitivity with detection or quantification of CMV in 4/20 SOT recipients, 7/20 HSCT
recipients, and 2/20 non-transplant immunocompromised individual samples
previously undetectable by CAP/CTM. In total, ART detected CMV in 13 (21.6%)
patients in which CMV was not detected by CAP/CTM. Across patient populations,
samples with quantifiable CMV by CAP/CTM were also quantifiable by ART with a
consistently positive bias. The average bias between ART and CAP/CTM was 0.27
log10IU/mL (range -0.65 to 1.41 log10IU/mL). Conclusions: The difference in lower
limit of detection between ART and CAP/CTM is likely the result of increased
detection of virus in patients under surveillance for CMV. Although the average
observed bias between methods was 0.27 log IU/mL, 33 (37%) patients had a viral
load difference >0.5 log10IU/mL. It is hypothesized that this difference in
quantification may be due to fragmentation of the CMV genome in clinical samples
and differences in CMV amplicon size between these 2 assays. The difference
between these 2 tests in both detection and quantification may result in additional
surveillance testing somewhat earlier than seen with the use of CAP/CTM, and may
necessitate establishing a new baseline for patients if transitioning from surveillance
with CAP/CTM.
ID54. Molecular Analysis of Fungal Populations in Patients with
Onychomycosis Using Next Generation Sequencing (NGS) and Real-Time PCR
E. Gustafson, I. Akinsanmi, L. Bennett, B. Bakotic
Bako Pathology, Alpharetta, GA.
Introduction: Onychomycosis is the clinical term for a fungal infection of the nail. It
constitutes an important public health problem due to its high incidence, increasing
prevalence and an increased risk of complications such as cellulitis and skin
ulcerations, both of which may lead to loss of digits or limb. The most common
fungal pathogens in onychomycosis are keratin metabolizing dermatophytes
including Trichophyton rubrum and Trichopyton mentagrophytes. However, other
fungal organisms implicated include saprophytic molds and Candida species.
Conventional methods for diagnosis of onychomycosis include culture, clinical
examination and direct microscopic examination using Periodic Acid Shift reaction
(PAS) and/or Gomori methenamine silver (GMS). These methods are less specific
(histology) and have extended turnaround times (culture), compared to molecular
methods. Methods: Here we report an analysis of the fungal diversity in
onchomycotic nails using next generation sequencing and real time PCR. We
utilized a novel PCR amplicon targeting the ITS2 and 28S region of the fungal
ribosomal DNA gene which was shown to produce a highly represented and uniform
PCR product size in multiple species tested. Products were sequenced with 50,000
to 100,000 reads using the Illumina MiSeq. The samples were tested for fungi using
as established real time PCR assay used to detect 15 of the most common fungal
organisms implicated in onychomycosis as well as traditional histological methods
using PAS and GMS. Results: Using histology as the reference method, real-time
PCR has a Positive Predictive Value (PPV) of 94%. NGS results are in 94%
agreement with the PCR method at the genus level when evaluating the top 10% of
species identified. Conclusions: A better understanding of the diversity and
distribution of infecting fungal species may lead to the selection of more effective
treatment options. Correlation of fungal organisms in fungal co-infections as well as
fungal bacterial co-infections will be discussed.
ID55. Mosquito Surveillance and Testing for Local Zika Virus in New York City
2016
J. Rakeman
New York City Department of Health and Mental Hygiene, New York, NY.
Introduction: Starting from Brazil in May 2015, Zika virus (ZIKV) infection spread
rapidly throughout Central and South America. ZIKV is an arthropod-borne flavivirus
primarily transmitted through the bite of infected Aedes mosquito species,
primarily A. aegypti, which is mainly present in the southern states of USA. Another
species, A. albopictus, is a potential vector for spreading ZIKV and is well
established throughout New York City (NYC). The high number of NYC residents
and visitors travelling to ZIKV affected areas has resulted in the importation of
numerous travel-related human ZIKV infections and poses the risk of passing the
virus along to possible transmission vectors. This study describes the measures that
the NYC Public Health Laboratory (PHL) and the Office of Vector Surveillance and
Control introduced for local Aedes mosquito surveillance and the results of
monitoring potential mosquito-borne ZIKV transmission for the year 2016. Methods:
The NYC PHL commenced monitoring local Aedes species in the 2016 mosquito
testing season from June 1 to October 31. Prior to this, a validation study was
performed within only 10 days, which included assessment of the sensitivity,
reproducibility and accuracy of the assay using negative mosquito pools spiked with
quantified ZIKV culture fluid (ZeptoMetrix, Buffalo, NY). The ZIKV screening assay is
based on a laboratory-developed real-time reverse transcription polymerase chain
reaction (rRT-PCR) protocol for clinical ZIKV testing of human specimens. City-wide
mosquito trapping at 121 locations, and speciation and pooling of mosquitoes were
performed by the Office of Vector Surveillance and Control. The subsequent
molecular analysis comprised homogenization of frozen pools, RNA extraction using
NucliSENS easyMAG system (bioMérieux, St. Louis, MO) and the rRT-PCR assay
targeting the envelope gene for broader detection of multiple ZIKV strains. A test
result was interpreted as positive if the cycle threshold (Ct) <38 and as negative if Ct
≥38 or undetected. Results: After the successful validation of the ZIKV rRT-PCR for
mosquito screening, the NYC PHL analyzed 2,120 mosquito pools comprised of
985
AMP Abstracts
20,317 individual specimens of A. albopictus. The number of mosquitoes per pool
ranged between 1 and 50 with an average of 10. All pools tested were negative for
ZIKV. Conclusion: Vector surveillance for ZIKV is an effective strategy to identify
potential risk to human populations for local ZIKV transmission by identifying
potentially infected mosquito vectors. The NYC PHL successfully validated and
implemented ZIKV testing on mosquitos collected through surveillance. Our testing
showed no evidence of ZIKV in local mosquito populations collected through
surveillance in NYC.
ID56. Monitor Vaginal Microbiota with One Swab: Copan ESwab as Convenient
Collection and Transport Device for Cross Platform Molecular Tests of
Women’s Health
Z. Huang, D. Rayman, S.S. Beqaj
Patients Choice Laboratories, Indianapolis, IN.
Introduction: Among many women’s healthcare testing are tests for Sexually
Transmitted Infections (STIs). Most common STI tests are Trichomonas
vaginalis (TV), Neisseria gonorrhoeae (NG), Chlamydia trachomatis (CT), Herpes
simplex virus (HSV)-2, bacterial and yeast vaginosis pathogens and Mycoplasma
and Ureaplasma. There are several platforms/technologies currently available to test
for these pathogens; however, most platforms have their own collection device or
allow a variety of collection devices, which can often be confusing for practitioners.
ESwab (Copan Italia, Brescia, Italy) is an FDA approved collection and transport
device. It consists of a nylon Flocked swab for specimen collection and 1mL of Liquid
Amies, a supportive medium that can maintain the viability of clinically important
aerobes, anaerobes, fastidious bacteria, yeast and viruses. The purpose of this
study was to validate ESwab as a universal swab for detection of STI pathogens by
molecular methods. Methods: A total of 376 ESwab specimens were used for this
study. Of these 376 specimens, 165 were patient specimens, 151 previously tested
negative patient specimens spiked with ATCC strains for tested pathogens, and 66
Gibson Tri-valent swabs eluted in ESwab medium. 200uL collection medium was
extracted using Roche MagNA Pure 96 system and cobas 4800 system depending
on individual assay requirements. CT/NG, HSV-1 and -2 and TV were tested on the
cobas 4800 system and results compared to a reference method. Bacterial vaginosis
scoring test (Lactobacillus spp., Gardnerella vaginalis, Atopobium vaginae), yeast
vaginitis (Candida albicans, Candida glabrata, Candida krusei), Ureaplasma and
Mycoplasma panel (Ureaplasma urealyticum, Ureaplasma parvum, Mycoplasma
hominis), were tested using Sacace real-time PCR assays on the Qiagen RotorGene Q and results compared to a reference method. In addition, precision and
specimen stability studies were performed for each test. Results: All tests showed
very good correlation using both patient specimens and spiked specimens: CT/NG:
100%/100% (41 specimens), HSV-1/-2: 100%/100% (44 specimens), bacterial
vaginosis: 97.9% (94 specimens), TV: 98.9% (92 specimens), yeast vaginitis: 96%
(99 specimens), and Ureaplasma and mycoplasma: 100% (78 specimens). For
precision, 5 specimens were run for 4 times for all tests and coefficient of variance of
Cycle threshold (Ct) of all organisms were below 5%, with the highest of 3.77%.
Stability study performed on ESwab for all tested pathogens was 7 days when stored
refrigerated. Conclusions: This validation study demonstrates that ESwab is a
convenient collection and transport device with extended stability and can be used
for molecular testing regardless of the technology or method used.
ID57. Multicenter Clinical Evaluation of a Real-Time PCR Assay for Bordetella
pertussis
T.S. Uphoff1, J. Dien Bard2, G. Denys3, M. Oethinger4, W. Sinclair5, S. Young6
1Marshfield Labs, Marshfield, WI; 2Children's Hospital and University of Southern
California School of Medicine, Los Angeles, CA; 3Indiana University School of
Medicine, Indianapolis, IN; 4Providence Health Services Oregon, Portland,
OR; 5Intermountain Healthcare, Salt Lake City, UT; 6TriCore Reference Laboratories,
Albuquerque, NM.
Introduction: Pertussis is a contagious respiratory disease caused by Bordetella
(B.) pertussis that can lead to life-threatening complications in young children. Early
and accurate diagnosis is critical for positive patient outcomes and to aid public
health officials in controlling spread during outbreaks. The Great Basin Bordetella
Direct Test is a qualitative in vitro diagnostic test for the detection of B.
pertussis DNA from nasopharyngeal (NP) swab specimens obtained from patients
suspected of having a respiratory tract infection from B. pertussis. The assay was
performed on the PA500 Portrait Analyzer and utilizes PCR amplification of the
insertion sequence IS481. Methods: The performance of the Bordetella Direct Test
was evaluated on 1037 NP samples, 915 prospectively collected and 122
retrospective frozen NP samples at 5 US clinical sites from August, 2016 to January,
2017. Results generated by the Bordetella Direct Test were compared to those
obtained with the Nanosphere Verigene RP Flex Test. Discordant samples were
adjudicated by testing with the Quidel Amplivue B. pertussis an FDA approved
molecular assay. Reproducibility studies were performed at 3 sites with 2 operators
at each site. A 3 sample panel consisting of a moderate positive sample (1.9x the
limit of detection {LoD}) a high positive sample (3.8x LoD) and a negative sample
was tested in triplicate over 5 non-consecutive days by each operator (n=270).
Results: In the combined evaluation of the Bordetella Direct Test (prospective and
retrospective samples), the Positive Percent Agreement (PPA) for B. pertussis was
92.5% (74/80; 95% confidence interval [CI], 84.6% to 96.5%). The Negative Percent
Agreement (NPA) for B. pertussis was 99.6% (953/957; 95% CI, 98.9% to
986
99.8%). Six samples were positive by the RP Flex Test and negative by the
Bordetella Direct Test, 4 of the 6 were confirmed positive using the Amplivue
assay. Conversely, 4 samples were Bordetella Direct Test positive and RP Flex
negative and 2 of these were confirmed positive using the Amplivue assay. There
were no discordant results in the reproducibility study, all 1.9x, 3.8x LoD and
negative samples gave the expected results in all testing events (270/270).
Conclusions: The Great Basin Bordetella Direct Test is a sensitive and specific
diagnostic assay for detection of B. pertussis in NP swabs from subjects with
suspected B. pertussis infection.
ID58. A Novel Approach for Sensitive Detection of ZIKV RNA in Whole Blood
and Urine Samples
Y. Chen1, J. Liu1, L. Tran1, S. Gessert1, J. Sapida1, K. Das1, N. Sergeev2, E.
Holmes1
1Theranos, Inc., Palo Alto, CA; 2Theranos, Inc., Newark, CA.
Introduction: The recent Zika virus (ZIKV) outbreak in the Americas has been
shown to cause a spectrum of neurologic complications including microcephaly in
fetuses and neonates. Greater access to testing is needed to improve detection,
clinical outcomes, and aid epidemiologic surveillance in impacted regions. Detection
of ZIKV RNA in clinical specimens is considered a definitive diagnosis of
Zika infection. However, the detection window of viral RNA in serum and
plasma is very narrow (typically 2 weeks from symptom onset). Recent findings
suggest that ZIKV may persist longer in other sample types such as urine and whole
blood. Unfortunately, efficient recovery and detection of viral RNA from whole blood
is challenging due to large amounts of human DNA, RNA and protein
(e.g. hemoglobin) which interfere with sample preparation and downstream analytical
procedures. Methods: The assay we developed utilizes a novel magnetic bead
based method for efficient extraction of ZIKV RNA from whole blood which allows
isolation of ZIKV RNA, removal of PCR inhibitors, and reduction of endogenous
human nucleic acid. The assay incorporates MS2 bacteriophage as an internal
control for sample processing, extraction, and downstream processes. Following a
multiplexed RT-PCR amplification, the products are split for singleplex isothermal
amplification and detection of ZIKV and MS2, respectively. The isothermal
amplification method employs nested primer pairs with pair-wise complementary 5'
ends, which facilitate formation of RT-PCR amplicons with 5' overhangs, allowing
generation of concatemers which are detected by a dsDNA-specific fluorescent dye.
Cut-off on inflection time has been established to allow qualitative results
reporting. Results: Limits of detection (LoD) at 120 RNA copies per mL in both
whole blood and urine are achieved in small volumes of whole blood (~120uL) and
urine samples (~150uL). The assay detects multiple ZIKV strains, including the
recent Puerto Rican PRVABC59 strain and 2 African strains, DakArdD41662 and
MR766. No cross-reactivity is observed at clinically relevant concentrations with 8
closely related pathogens that lead to similar disease manifestations. A high load
(>106 copies per mL) of the same pathogens does not inhibit detection of ZIKV
present at twice the LoD. Conclusions: We have developed a diagnostic assay for
ZIKV RNA which leverages an extended detection window, while retaining
competitive sensitivity and specificity in small volumes of whole blood and urine
samples. This assay is being integrated into an automated sample-to-answer
platform aimed at allowing access to ZIKV testing at the point-of-care.
ID59. Using Independent Run Controls to Monitor Relative Amplification
Efficiency of a HAI Assay
J.C. Yundt-Pacheco1, C.I. Kanis2, W. Chiu2, J. Liu2, K. Barnecut2, V.P. Luu2, T.
Cheever2, M. Zeballos2, R. Stahl2, B. Fukunaga2
1Bio-Rad Laboratories, Plano, CA; 2Bio-Rad Laboratories, Irvine, CA.
Introduction: Amplification efficiency is an important performance metric for qPCR.
This is particularly true on embedded test platforms used for qualitative testing when
there are minimal details about test performance metrics easily accessible. We used
a commercially available independent run control (Amplichek II) to assess and
periodically monitor the amplification efficiency of the Cepheid Xpert Methicillin
Resistant Staphylococcus aureus (MRSA) assay. Methods: The equation below
describes the exponential amplification of PCR,
Where Cn = the copy number at cycle n, Ci = Initial copy number (Sample Size), E =
Efficiency of amplification in decimal form, and n = Number of cycles.
Rearranging terms,
Eight evaluations of Amplichek II were used to determine a mean cycle threshold
value (Ct) to use as a baseline efficiency of 1.
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
We then monitored the relative amplification efficiency every 4 months for 36 months
with a single lot of Amplichek II stored at 2°C - 8°C. We also evaluated the stability
of Amplichek II by comparison against specimens stored at -20°C. Results: The
stability study showed no downward trend in the Amplichek II MRSA
concentration. Relative amplification efficiencies ranged from 0.874 – 1.096.
Conclusion: There were significant variations (±10%, p<0.05) in the relative
amplification efficiency measurements using Amplichek II. Using an independent run
control to monitor relative amplification efficiency could be a useful method for
tracking performance over time.
charts constructed with a range determined by the mean ± 2 SD can be used to
identify control values that exceed expected range and determine the acceptability of
performance of each assay and the reporting of patient results. Conclusions: We
have demonstrated a high quality performance of commercial multianalyte control
materials for transplant viral loads. By incorporating these controls and quality
control tools into the laboratory quality assurance program, systematic errors can be
identified and corrective actions performed before reporting patient results.
ID60. Extreme One-Step RT-PCR: Potential for Point-of-Care Viral Detection.
J.A. Houskeeper, L. Narramore, C.T. Wittwer
University of Utah, Salt Lake City, UT.
Introduction: Reverse transcription (RT) is often coupled with PCR for simplified
detection of RNA. Despite the ease of one-step RT-PCR, some pathogens are still
assessed using a 2-step method. Most two-step methods require over an hour for
pathogen detection. Even the best commercially available one-step RT-PCR kits
require a minimum of 20 minutes for detection. Due to the essential nature of fast
and accurate detection of viral infections, we have developed one-step RT-PCR
assays that can be performed in under 2 minutes and implemented as a point-ofcare technique. Methods: A prototype instrument was used to temperature cycle 5
µL samples in 0.4 to 1.0 seconds at annealing, extension, and denaturation
temperatures. Such rapid PCR requires increased polymerase, primer, and
Mg++ concentrations. Viral RNA from human parainfluenza (VR-93D), human
enterovirus (VR-1775DQ and VR-1825), zika (VR-1838DQ), and human respiratory
syncytial virus (VR-26D) was acquired from the American Type Culture Collection
(ATCC). Primers were designed with melting temperatures between 60C and
75C. Detection sensitivity and efficiency of viral targets were analyzed by real-time
PCR. Sample specificity was demonstrated by high-resolution melting. Results: All
viral targets were amplified as 50 to 100 bp targets in 2 minutes or less. Without the
use of hot-start methods, the extreme RT-PCR resulted in amplification of notemplate controls. This was attributed to the increased reaction component
concentrations resulting in enhanced non-templated additions. However, with the
implementation of hot-start techniques, the Cqs of non-templated additions
were increased by more than 10 cycles as confirmed by real-time PCR and highresolution melting. Conclusions: Use of extreme diagnostic technology has led to
the design of a one-step RT-PCR assay that can be performed in under 2 minutes
with competitive sensitivity to alternative RT-PCR methods. Extreme RT-PCR
illustrates the ability of point-of-care diagnostic testing for infectious disease when
prompt results may be critical.
ID62. Rapid and Sensitive Isothermal Molecular Amplification of Group A
Streptococcus (GAS) with Alere i Molecular Platform
N. Moore
Alere, Scarborough, ME.
Introduction: The diagnosis of patients with suspected Group
A Streptococcus (GAS) can be clinically challenging, however, if detected
and treated quickly, can improve care and antibiotic use. Detection of GAS by lateral
flow assays in clinical practice may be suboptimal based on reported sensitivities of
around 85%. As an alternative, molecular assays show increased sensitivities for
detecting GAS due to sample amplification. These molecular tests are becoming
more prevalent in clinical practice as the ease-of-use and time-to-results have
dramatically improved. Traditional PCR molecular technology uses thermocycling
steps of heating and cooling which adds time to result (20 minutes to hours). The
Alere i molecular platform uses isothermal molecular technology as a means of
amplifying the sample. By using enzymes and constant temperature, time to results
can be significantly reduced. The sensitivity and specificity of the Alere i Strep A
assay has been previously reported. We evaluated the Alere i molecular platform for
its ability to provide molecular results within a typical patient office visit where there
are often fewer than 15 minutes to diagnose and treat. Methods: Time to result data
for the Alere Strep A assay was collected from a study involving 8 US sites and 481
enrolled patients. Patients presenting with pharyngitis had 2 swabs collected. One
swab was evaluated with the Alere system and the second was sent for bacterial
culture. Discrepant analysis was done by PCR. Additional randomized blind-coded
panels of low positive and negative samples were tested by 6 untrained users at 3
sites with no laboratory experience. Results: Time to result for the Alere i Strep A
assay was 8 minutes for a negative result, including test warm-up time. Of the
samples that tested positive for GAS, 75% were reported as positive within 5
minutes of introducing the sample, including test warm up time. Conclusions: The
Alere i Strep A assay is a rapid and highly accurate isothermal molecular
amplification method for the detection of GAS. Alere i Strep A negative rest results
were available within 8 minutes, and 75% of positive results were available within 5
minutes of introducing the sample. The ability to provide negative GAS results within
8 minutes and a 5-minute early call-out for most positive samples with the Alere i
Strep A test helps increase the fit of molecular testing into current standard-of-care
timeframes to help improve the accurate and early diagnosis of GAS and associated
therapy decisions.
ID61. Performance Evaluation of Commercial Multianalyte Control Materials
Calibrated Against the 1st WHO International Standards for Quantification of
CMV, EBV and BKV in Transplant Patients
F. Sabato1, S. Da Rosa Duque1, M. Chhoa1, C. Margaritini1, C. Anaya-Bergman1, L.
Barto1, M. Sene1, A. Ferreira-Gonzalez2
1Virginia Commonwealth University Health, Richmond, VA; 2 Virginia Commonwealth
University, Richmond VA.
Introduction: Professional guidelines and international consensus groups
recommend the quantification of CMV, EBV and BKV DNA in plasma for
management of active infection in transplant patients. Laboratory-developed
procedures (LDP) using real-time PCR have been used for that purpose.
Standardized quality control materials offer an approach to assess analytical process
performance. Recently, commercial multianalyte controls calibrated against the
1st WHO International Standards for CMV, EBV and BKV, using digital PCR have
been developed. Here, we report the evaluation of these controls and the
establishment of quality control parameters to monitor the overall testing
performance of our transplant viral qPCR LDP. Method: Two levels of multianalyte
controls (Exact Diagnostics, TX) at concentrations close to the clinical decisions
levels for each target were included in this study (LPC: low positive control and HPC:
high positive control). They are intact virus diluted in plasma EDTA to best simulate a
true patient sample. Three aliquots of each control were included in clinical runs for 7
days, for a total of 21 replicates. Aliquots were extracted on QIAsymphony using the
DSP Virus/Pathogen Mini kit and DNA targets were quantified by qPCR with
the artus CMV, EBV and BKV reagents on the Rotor Q. Controls were analyzed for
each target and the log transformed data was used to calculate the mean and
standard deviation (SD). Levy-Jennings charts were constructed to monitor the
performance of the assays overtime and to identify test system shifts or trends that
would require applying corrective actions. Results: Mean and SD values in Log
IU/mL were calculated for each of the targets in the controls. For the LPC: 3.73 ±
0.05; 3.37 ± 0.12 and 3.29 ± 0.14 were obtained for CMV, EBV and BKV,
respectively. For the HPC: 4.98 ± 0.05; 4.66 ± 0.05 and 4.64 ± 0.12 were obtained
for CMV, EBV and BKV, respectively. A high level of precision was observed with SD
≤ 0.14 Log10 which indicates that testing controls multiple times would produce
results within ± 0.26 Log10 (± 2SD). This level of variability is within the ± 0.5
Log10 which is not considered to be clinically significant. Therefore, Levy-Jennings
ID63. Improved Cost and Turnaround Time Using an Extraction-Free
Amplification and Detection Method for Respiratory Viruses in Clinical
Specimens
M. Elkan
The Children’s Hospital of Philadelphia, Philadelphia, PA.
Introduction: Molecular detection of respiratory viruses has replaced viral culture in
clinical laboratories. While these methods are more sensitive and faster than
culture, they can be prohibitively expensive. Numerous FDA cleared and laboratory
developed options exist for multi-virus molecular testing. The vast majority of panels
employ nucleic acid extraction and purification prior to amplification and detection of
viral nucleic acids. Data is emerging to suggest this step may be unnecessary for
virus detection, possibly reducing costs further. Our laboratory employs a laboratory
developed multiplex PCR panel for 13 respiratory viruses and an internal control. In
order to shorten turn-around times and reduce costs, we explored the performance
of a rapid heat block treatment, in lieu of nucleic acid extraction, followed by our
standard PCRs. Methods: Previously tested nasopharyngeal aspirates, positive for
either viral DNA or RNA, were selected for this side-by-side comparison study. For
the extraction-free method, a 50µL aliquot was given a heat block treatment, then
rapidly chilled at 4°C. For the standard extraction method, a 200µL aliquot was
extracted on a MagnaPure LC 2.0 employing the Magna Pure LC Total Nucleic Acid
Isolation Kit (both Roche) and eluted in 66µL. The viruses tested included influenza
virus types A (n=8) and B (n=8), respiratory syncytial virus types A (n=8) and B
(n=7), parainfluenza virus types 1 (n=8), 2 (n=8), and 3 (n=8), rhinovirus (n=8),
adenovirus (n=8), human metapneumovirus (n=8), corona virus types NL63 (n=7),
OC43 (n=7), HKU1 (n=8) and 229-E (n=8) and a beta-actin internal RNA control
(n=7). Both heat block treatment samples and MagnaPure extractions were
simultaneously tested on the QuantStudio Dx System (Applied Biosystems).
Results: Sensitivity, calculated as virus detection in the heat-treated specimens
compared to MagnaPure LC extractions, was as follows: influenza-A 88%, influenzaB 88%, RSV-A 100%, RSV-B 100%, PIV1 88%, PIV2 100%, PIV3 100%, rhinovirus
25%, adenovirus 88%, hMPV 100%, NL63 88%, OC43 100%, 229E 100%, HKU1
75% and B. actin 100%. When positive, crossing cycle thresholds (Ct) for heattreated specimen virus detections were within 2 Ct values of corresponding extracted
specimens (except rhinovirus). Conclusions: Our extraction-free approach showed
good sensitivity compared to our standard method for most viruses. Our approach
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
987
AMP Abstracts
reduced reagents cost per test by 48% and shortened test time by 36%. Thus,
extraction-free approaches to respiratory virus detection may be beneficial to clinical
laboratories and manufacturers of respiratory panel tests. More work is needed to
improve sensitivity, particularly for rhinovirus, and could include bead beating to
ensure better lysis.
samples that tested negative on the ABI7500 platform with viral load ranging from 30
copies/mL to 343 copies/mL. Conclusions: ELITe InGenius Sample-to-Answer
system is an easy to use, fast and efficient, automated instrument for BK virus
detection by Quantitative Real-time PCR method to provide sensitive, accurate and
reproducible results.
ID64. Optimization of Metatranscriptomic Method for Rapid and Unbiased
Detection of Microbial Pathogens in Bronchoalveolar Lavage Specimens
C. Yin1, W. Huang1, G. Wang1, J. Rosenblum1, S. Krishnan1, N. Dimitrova2, J.T.
Fallon1
1New York Medical College, Valhalla, NY; 2Philips Research North America,
Cambridge, MA.
Introduction: Bronchoalveolar lavage (BAL) specimens contain important
microbiological and cytological information in respiratory infectious processes.
Compared to conventional culture-based diagnostic methods, next generationsequencing (NGS) provides unbiased identification of a diversity of microbes,
inhabitant or pathogenic microorganisms, especially for those nonculturable and
fastidious ones. Here, we detail an optimized metagenomics RNA sequencing
(RNAseq), or metatranscriptomic, protocol for rapid detection of microbial pathogens
in clinical BAL specimens. Methods: The unused portions of 19 BAL patient
specimens were stored in Trizol LS Reagent at -80oC. An internal RNA control
(20ng) was added into each sample at the beginning of RNA extraction. After beads
disruption, RNA was extracted and human ribosomal RNA (rRNA) was depleted by
using Ribo-Zero kit. NGS library was generated from 1 ng of synthesized cDNA by
using Nextera XT kit, and was sequenced on MiSeq (paired-end, 75x2). Sequence
reads from the internal RNA control and human host source were removed by using
BWA alignment. The remaining sequence reads were used for taxonomic
classification by using the Kraken algorithm and NCBI database (August
2016). Results: The optimized protocol takes about 24 to 36 hours, which is rapid
and sensitive in detection of microorganisms. The outcomes from our
metatranscriptomic method were consistent with culture results from routine clinical
diagnostic methods. All bacterial species reported from the clinical laboratory,
including Enterococcus faecium, Enterococcus faecalis, Pseudomonas aeruginosa,
Neisseria meningitides, Proteus mirabilis, Haemophilus influenza, Staphylococcus
argenteus, Staphylococcus aureus, Streptococcus pneumoniae, Achromobacter
xylosoxidans Bordetella petrii, etc, were detected by metagenomics RNA-seq. In
addition to bacterial pathogens, the metatranscriptomic method simultaneously
detected infectious viruses (WU Polyomavirus) and fungi (Candida dubliniensis),
which demonstrate its advantage. The internal RNA control also played a critical role
in determination of pathogen negatives, monitoring of experimental contaminations
and relative quantification of microbes. Conclusions: Compared to amplicon
metagenomics sequencing and shot-gun metagenomics sequencing previously
used, metagenomic RNAseq provides a wider range of detection of viral, bacterial,
fungal, and parasitic species. Our simplified and optimized metatranscriptomic
method provides a useful tool for rapid and sensitive detection of pathogens in
clinical samples, and is a promising application for clinical microbiological
diagnostics.
ID66. Using Independent Run Controls to Monitor Relative Amplification
Efficiency in a STI Assay
J. Yundt-Pacheco1, C. Kanis2, W. Chiu2, J. Liu2, K. Barnecut2, V.P. Luu2, T.
Cheever2, M. Zeballos2, R. Stahl2, B. Fukunaga2
1Bio-Rad Laboratories, Plano TX; 2Bio-Rad Laboratories, Irvine, CA.
Introduction: Amplification efficiency is an important performance metric for qPCR.
This is particularly true on embedded test platforms used for qualitative testing when
there are minimal details about test performance metrics easily accessible. We used
a commercially available independent run control (Amplichek STI) to assess and
periodically monitor the amplification efficiency of the Abbott m2000 RealTime
CT/NG assay. Methods: The equation below describes the exponential amplification
of PCR,
ID65. Development and Validation of the Alert MGB ASR for BK Virus
Quantitative Viral Load Testing on the ELITEe InGenius Sample-to-Answer
System
D. Banerjee, R. Selvarangan
Children’s Mercy Hospital, Kansas City, MO.
Introduction: BK virus infects children during early-childhood years and remains
quiescent unless immunosuppression is encountered by the host. Virus shedding is
detected in AIDS patients and those receiving organ (kidney) or Bone Marrow
Transplant, where the infection presents as BK Virus nephropathy (BKVN) or
hemorrhagic cystitis respectively. Screening of blood and urine for BK virus in renal
transplant patients is important due to the risk of transplant failure with BK virus
infection and need for prompt patient management. In this study, we compared
performance of ELITech MGB Alert BK virus specific ASRs on ABI7500 and ELITe
Ingenius, a fully automated, sample to result Real Time PCR platform for BK virus
detection using clinical specimens. Methods: The MGB Alert ASR for BK virus and
the Elite InGenius system were calibrated using AcroMetrix BKV Panel and BK
monoreagent prepared using MGB Alert Master Mix, primers and probes. A range of
BK virus dilutions were prepared by diluting BKV Panel with AcroMetrix Plasma
Matrix for limit of detection (LOD), limit of quantification (LOQ) and precision
experiments. For method comparison, 72 frozen specimens (urine - 44 and plasma 26) that were previously detected as positive (60) and negative (12) by ABI7500
system were re-tested on Ingenius and results were compared. Results: The BK
assay developed on InGenius has a linear range of 5e2 copies/mL – 5e6 copies/mL.
LOD and LOQ were both detected to be 100 copies/mL. Intra assay and inter assay
precision of 500, 104 and 105 copies/ml tested twice daily, over 3 days by a single
operator demonstrated highly precise measurements (Total within laboratory SD
ranged from 0.1 to 0.3 copies/mL with % CV’s of 2.3% to 11.5%). A total of 59
samples tested positive and 8 samples tested negative by both methods. Viral load
in urine specimens ranged from 30 copies/mL to 193,197,676 copies/mL and 343
copies/mL to 11,608,997 copies/mL in plasma samples. The correlation coefficient
between InGenius and ABI7500 was 97.7% and 95% for 42 positive urine specimens
and 17 positive plasma specimens respectively. InGenius detected BK virus in 4
988
Where Cn = the copy number at cycle n, Ci = Initial copy number (Sample Size), E =
Efficiency of amplification in decimal form, and n = Number of cycles.
Rearranging terms,
Eight evaluations of Amplichek STI were used to determine a mean cycle threshold
value (Ct) to use as a baseline efficiency of 1 for CT and NG.
We then monitored the relative amplification efficiency every 4 months for 20 months
with a single lot of Amplichek STI stored at -20°C for CT and NG. We also evaluated
the stability of Amplichek STI by comparison against specimens stored at -70°C.
Results: The stability study showed no downward trend in the Amplichek STI CT
and NG concentrations. Relative amplification efficiencies for CT ranged from
0.8605 to 1.059. Relative amplification efficiencies for NG ranged from 0.8216 to
1.014. Conclusion: There were significant variations (±10%, p<0.05) in the relative
amplification efficiency measurements using Amplichek STI. Using an independent
run control to monitor relative amplification efficiency could be a useful method for
tracking performance over time.
ID67. A Two-Step RT-LAMP Provides Improved Sensitivity for Point of Care
Detection of Arboviruses
J. Benzine, J. Koelbl, T. Rockweiler, D. Manna
Lucigen, Middleton, WI.
Introduction: Arboviruses (ARthropod BOrne VIRUSES) are a group of viruses
transmitted by insect vectors, and include Zika (ZIKV), dengue (DENV), and
chikungunya (CHIKV). Symptoms that arise from ZIKV, DENV, or CHIKV infection
(fever, rash, and joint pain) are similar and often benign. However, the recent
outbreak of Zika in Brazil revealed a link between the virus and severe fetal
abnormalities, drawing attention to the increasing global prevalence of arbovirus
infections, and highlighting the need for rapid, low-cost diagnostics. Here we present
an improved RT-LAMP (Reverse-Transcription Loop-mediated isothermal
AMPlification) method for highly sensitive, specific and rapid detection of ZIKV,
DENV types 1-4, and CHIKV. Methods: Conventional “one-step” RT-LAMP reaction
is carried out using a combination of strand displacing DNA polymerase and a
reverse transcriptase but suffers from poor sensitivity. The two-step RT-LAMP,
developed here, utilized a single thermostable DNA polymerase with strand
displacement and reverse transcriptase activities (Lucigen). In the first step, RNA
target is converted to cDNA, and in the second step, resulting cDNA is amplified
through LAMP. Lyophilized reagents for the two-step LAMP contained all necessary
primers, enzymes and fluorescent dye for real time monitoring of amplification.
Assays were carried out at constant temperature, and reactions that
yielded fluorescence signal above a set threshold within 30 minutes were called
positive. Assay sensitivity was determined using serial dilutions of target RNA or
virus while assay specificity was determined against a panel of viral pathogens.
Results: The two-step RT-LAMP proved 1 to 4 log(s) more sensitive compared to
conventional one-step RT-LAMP and was within a log of that obtained using RTPCR. Using the two-step RT-LAMP, all the arbovirus targets were detected within 30
min, 2-4 times faster than RT-PCR. The RT-LAMP assays were highly specific when
tested against a panel of viral targets. Lyophilized two-step RT-LAMP reagents
performed similar to wet reagents. Lastly, using the two-step RT-LAMP, Zika virus
was rapidly detected from spiked urine with high sensitivity without the need for
nucleic acid purification. Conclusions: The two-step RT-LAMP method provides
improved sensitivity over conventional one-step RT-LAMP and is capable of
detecting arboviruses in the presence sample matrix such as human urine.
Furthermore, lyophilized two-step RT-LAMP reagents are room-temperature stable
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
making such assays field deployable without the need for cold chain. Two-step RTLAMP arboviral assays have the potential to improve patient management and
infection control at the point of need and particularly in resource-poor settings.
ID68. Early Detection of Fungi and Yeast Using Species Specific Dual
Amplification PCR (MycoDART) for Clinical Diagnosis.
S.S. Sutton, S.A. Muralidhar, M.S. Shon, V.E. Bolton, D.G. Hooper
AdvaTect Diagnostics, LLC, Carrollton TX.
Introduction: Invasive Candidiasis and Aspergillosis are a major cause of morbidity
and mortality amongst immunosuppressed patients. The major pitfall for diagnosis of
these infections is the difficulty to culture these organisms from infected sites and
body fluids, especially blood components. Animal and human patient PCR data in
potential invasive fungal diseases (IFD) is presented. PCR results from plasma are
compared to blood culture and Platelia Aspergillus EIA (Galactamannan). Method:
Plasma from patients (University of Florida, Gainesville, FL) and from mice in a
controlled study (UTSA, San Antonio, TX) of Aspergillus and Candida infections as
well as known ATCC cultures of 4 Aspergillus spp. and 5 Candida spp. were
examined. The assays utilized proprietary specific hydrolysis probes designed for
species-specific target sequences (MycoDART). Extractions from hyphae and
conidia were tested in a modified dual amplification PCR assay. Geotrichum spp.
spores were used as an internal control. Tests were conducted to demonstrate
specificity, linearity, efficiency, and limit of detection (LOD) for each species.
Results: Dual amplification PCR (MycoDART) isolated 30% more isolates for
Aspergillus and 73% more isolates for Candida species in clinical samples compared
to the traditional PCR methodology. In-vivo mouse studies on injection with A.
fumigatus and C. albicans showed complete infection with the DNA probes/primes in
plasma by Day 3 for A. fumigatus and Day 0 (4 hours) for C. albicans. Clinical
patients with hematologic malignancies or undergoing hematopoietic cell
transplantation from the University of Florida were evaluated for performance and
specificity of the assay. The method successfully detected and identified 33%
more Aspergillus spp. and 20% more Candida spp. in the plasma samples tested
when compared to the traditional blood culture method. Treatment modalities that
differed among patients will be presented. Furthermore, efficiency (>91%), linearity
(>0.99) and precision will be presented. LOD and sensitivity range from 6 copies to
55 copies per microliter of DNA will also be presented for each species. Conclusion:
Clinical sample analysis results indicate evidence for a rapid and reliable
assay. Species-specific dual amplification PCR (MycoDART) for the diagnosis for
IFDs proves optimal to current strategies such as blood cultures and Galactamannan
ELISA. Such tests could be applied to early diagnosis, monitor effectiveness of
therapy, and assist in predicting outcomes.
ID69. Detection of Group B Streptococcus Using the Simplexa GBS Direct
Assay
R. Martin, A. Tran Ha, R. Hazelo, Y. Parocua, H. Gregson, P. Naranatt, M. Tabb
DiaSorin Molecular, Cypress, CA.
Introduction: Group B Streptococcus (GBS) is a gram positive bacteria commonly
found in the gastrointestinal and urogenital tracts of healthy adults. Approximately
10-30% of all pregnant women are colonized with GBS and transmission to the
newborn remains the leading cause of early-onset neonatal sepsis. The Simplexa
GBS Direct assay is currently in development to detect GBS from antepartum swab
samples enriched in Lim broth or a chromogenic broth. Methods: Vaginal-rectal
swabs in various non-nutritive transport media were enriched in Lim broth and Carrot
broth. The enriched culture samples were loaded directly onto a Direct Amplification
Disc without up-front nucleic acid extraction or other specimen preparation followed
by real-time PCR on the DiaSorin LIAISON MDX system. Results were compared to
historical culture results or another FDA cleared Test (NAAT). Analytical reactivity of
the Simplexa GBS Direct assay was assessed with 10 GBS strains spiked into
pooled negative natural matrix. Analytical specificity was also demonstrated by
testing 123 bacteria, viruses, and parasites that may be found in vaginal-rectal
specimens. Results: Two hundred seventeen vaginal-rectal swabs in transport
media were enriched in Lim broth according to CDC guidelines and tested on the
Simplexa GBS Direct assay. For discordant analysis, enriched samples were tested
on the Cepheid Xpert GBS LB assay. When comparing valid results to the reference
method (culture or NAAT), clinical sensitivity for Simplexa GBS Direct was 95.8%
and clinical specificity was 99.2%. Four of the 5 discordant results agreed with
Cepheid Xpert LB results. Additionally, 118 swabs were also enriched in Carrot
broth. Sensitivity and specificity were 90.4% and 97.0%, respectively, with carrot
broth enrichment. Five of the 7 discordant results agreed with Cepheid Xpert LB
results. Analytical reactivity experiments demonstrated that the assay can detect
GBS strains from serotypes I-V and a non-hemolytic strain at ≥460 CFU/reaction
enriched culture. One hundred twenty-three organisms were tested for analytical
specificity and none were detected on the Simplexa GBS assay. An in
silico specificity analysis was also performed on an additional 17 organisms.
Conclusions: The Simplexa GBS Direct assay demonstrated ≥ 90% sensitivity and
≥ 97% specificity using either Lim broth or Carrot broth enrichment. All ten (10)
Group B Streptococcus strains were detected. No cross reactivity was observed
using the bacteria, viruses or parasites tested. The assay and instrumentation
provide a compact system for rapid (~1 hour) detection of GBS directly from Lim
broth or Carrot broth enriched samples.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
ID70. Evaluation of the Elitech MGB Alert Dengue 1,2,3,4 Test on the Elite
InGenius Sample to Result System
P. Ruiz1, J. McCue1, C. Fernandez-Bango2, S. Chao1, O. Ferrer2, A. Pinon-Ramos2
1University of Miami Health System, Miami, FL; 2University of Miami, Miami, FL.
Introduction: Dengue virus is a single stranded RNA (ssRNA) virus of the
Flaviviridae family of viruses, genus Flavivirus which also includes West Nile virus,
yellow fever virus and several other viruses that are causes of encephalitis. Dengue
virus has 4 distinct but closely related serotypes (Dengue 1-4). These are causative
agents of Dengue Fever (aka Breakbone Fever) and the more severe, frequently
deadly Dengue Hemorrhagic Fever. Dengue virus is transmitted by female
mosquitos (species Aedes aegypti) which also transmits chikungunya, yellow fever
and Zika infection. The Aedes mosquito is wide spread throughout the tropics and
Latin America. Over the last several decades, Dengue infections incidences have
grown dramatically. Recent estimates range from 280-520 million infections per year.
Over 100 million infections result in clinical manifestations with severity of
disease. Individuals that survive an infection of any single serotype of the Dengue
Virus acquires lifetime immunity from that particular serotype but not the others
making identification of the type of virus a patient is infected with an important
aspect of the diagnosis. Here we present the results of our evaluation of a novel
assay to detect and differentiate the 4 major serotypes of the Dengue Virus.
Methods: Extraction and PCR were performed on the ElitechGroup MDx ELITe
InGenius sample to result system. The ELITe InGenius utilizes standard magnetic
bead extraction technology for sample processing, sets up the PCR and conducts
the thermalcycling in a load and walk away fashion. The ElitechGroup Dengue
1,2,3,4 RUO was used for the PCR portion of the test. Quantifast one step RT-PCR
Master Mix (Qiagen) was used for reverse transcription and amplification
of extracted viral RNA. 12 previously characterized samples were run (6 Positive, 6
Negative) with Synthetic RNA Positive Control material supplied by the ElitechGroup
specific for each serotype. Results: Initial results show good correlation between
samples run on InGenius and the previously used method. One sample that was
previously determined as positive did not amplify. All negative samples were
confirmed by both methods (PPA 83.3%, NPA 100%). Blinded positive control
samples performed as expected. Limit of Detection and specificity experimental
results to be presented with poster are still pending at this time. Conclusions: The
ElitechGroup Dengue 1,2,3,4 RUO run on the ELITe InGenius Sample to Answer
system shows robust amplification of samples that are positive for all 4 serotypes of
the virus. Furthermore, the ability of the assay to differentiate each serotype proves
to be a useful tool in patient management. The Assay we have developed is an
accurate and sensitive method for the detection of Dengue virus infection.
ID71. WITHDRAWN
ID72. Evaluation of the Abbott Real-Time RT-PCR EBV Assay for EBV
Detection and Quantification
M. Yoon
Seoul National University Hospital, Seoul, South Korea.
Introduction: There has been a need to develop reliable diagnostic assays for
quantifying EBV load. We evaluated the performance of a commercial Abbott Realtime EBV assay, which targets BLLF1 gene. Limit of detection, linearity and
reproducibility were assessed using the Qnostics EBV analytical panel (Qnostics,
Ltd., United Kingdom). Methods: A total of 120 whole blood samples were analyzed
to evaluate reproducibility the agreement between the Abbott Real-time EBV assay
and the Nanogen EBV REAL-TIME Alert Q-PCR assay (Nanogen Advanced
Diagnostics, Italy), used in our laboratory. Results: The LOD of the Abbott Real-time
EBV assay was 1.69 log10 IU/mL. The assay was linear from 1 x 102 to 1 x 105 IU/mL
(r2 = 0.997). The within-run coefficients of variation (CV) ranged from 1.68% to
3.46%, and the between-run CV ranged from 1.73% to 12.83% for samples with
high, medium and low EBV loads. For 96 samples, positive with Abbott Real-time
EBV assay and with Nanogen EBV REAL-TIME Alert Q-PCR assay, were
significantly correlated with both assays (r2 = 0.773, P < 0.0001). The matched 96
samples demonstrated a positive bias of 0.11 log10 copies/mL for the Abbott Realtime EBV assay to the Nanogen EBV REAL-TIME Alert Q-PCR assay.
Conclusions: The Abbott Real-time EBV assay demonstrated reliable performance
for EBV DNA quantification in whole blood and correlated well with to the Nanogen
EBV REAL-TIME Alert Q-PCR assay.
ID73. Concordance of C. difficile Detection by Use of a Multiplex Molecular
Panel with a Singleplex, Diagnostic Assay
T. Hall1 , W. Sinclair1 , G. Hinde1 , M. Ul-Hasan1 , B. Lopansri1 ,2
1Intermountain Healthcare, Murray, UT; 2University of Utah, Salt Lake City, UT.
Introduction: Multiplex, syndromic molecular panels are increasingly being used in
clinical laboratories as an aid in the diagnosis of acute gastroenteritis. This approach
can shorten turnaround time and enhance efficiency; however results can be
confusing without appropriate clinical and epidemiologic context. We conducted
a review of multiplex GI panel (GIP) results to determine concordance of Clostridium
difficile results with a singleplex diagnostic strategy. Methods: At the Intermountain
Healthcare central microbiology laboratory, GIP tests (FilmArray, BioFire
Diagnostics) positive for C. difficile are subsequently processed using C.
difficile GDH antigen/Toxin A and B assay (QuikChek Complete, Alere). We queried
the electronic data warehouse to identify all GIP tests results which had
989
AMP Abstracts
a corresponding GDH antigen/Toxin test performed within a 48 hour period before or
after the index GIP test. The GDH antigen/Toxin test served as the reference method
to determine performance characteristics for the C. difficile component of the GIP.
Results: Between 1 October 2015 and 28 February 2017, 960 GIPs were performed
with 479 (50%) submitted from inpatient, 424 (44%) outpatient and 57 (6%)
emergency room encounters. Of those, 166 (17.2%) were positive for C. difficile.
152/166 had a concomitant GDH/Toxin assay performed with 39/152 (25.7%)
yielding a positive result for both GDH antigen and toxin and 1/166 positive for the
GDH antigen but negative for toxin. Additionally, there were 201 samples which had
QuikChek performed within 48 hours of a GIP test that was negative for C.
difficile. Of these samples, 200/201 were negative for both GDH and toxin and only
1/201 yielded a positive result. Using GDH/Toxin assay as the reference method,
GIP had a sensitivity/specificity/positive predictive value/negative predictive value of
97.6%/64.1%/26.3%/99.5%, respectively. Conclusions: A negative C. difficile result
in a highly multiplexed molecular panel has a high negative predictive value
compared to a non-molecular method. The significance of a positive result by GIP is
unclear and additional information needed.
ID74. Detection of Resistance-Associated Substitutions in the Hepatitis C Viral
Genome Using the Sentosa SQ Hepatitis C Virus Genotyping Next-Generation
Sequencing Assay
M. Campan, J. Pettersson, B. Lee, J. Kahn, L. Dubeau, P. Ward
University of Southern California, Los Angeles, CA.
Introduction: New direct-acting antiviral drugs for hepatitis C result in higher
sustained virologic response rates. However, select patients can fail first line
therapy due to the presence of resistance-associated substitutions in the
nonstructural proteins NS3 (protease), NS5A, and NS5B (polymerase). Early testing
for such substitutions assists in the selection of the most appropriate treatment
regimen and reduces treatment failure and healthcare costs. We aimed to evaluate
the efficacy of the Sentosa SQ HCV Genotyping NGS assay, which is a novel test
detecting resistance-associated substitutions in the hepatitis C viral genome.
Methods: We examined 50 randomly selected samples initially received for hepatitis
C viral load quantification in the USC Molecular Pathology Laboratory. All samples
were processed using the Sentosa automated Next Generation Sequencing
Workflow. This assay detects genotypes 1, 2, 3, 4, 5 and 6, and resistanceassociated substitutions in the 1a and 1b subtypes by targeting the NS3, NS5A, and
NS5B proteins. Results: Of the 50 samples tested, 31 showed genotype 1, 5
showed genotype 2, 12 showed genotype 3, one showed genotype 4, and one
showed genotype 6. A total of 26 substitutions were detected in NS3, 14 in NS5a,
and 2 in NS5b. The most common substitutions in NS3 included 12 at the p.Q80
position (39% of study subjects with genotype 1), 6 at the p.S122 position (19% of
subjects with genotype 1), and 5 at the p.V36 position (16% of subjects with
genotype 1). The most common substitutions in NS5a included 4 at the p.M28 and 4
at the p.Q30 positions (each seen in 13% of all study subjects). Conclusions: There
is a high prevalence of resistance-associated substitutions in the regions of the
encoding proteins targeted by direct-acting antiviral drugs in the treatment of
hepatitis C. Seropositive patients with such substitutions may not respond to first line
therapy, prolonging their infection period and significantly increasing healthcare
costs. This emphasizes the need for early detection of such mutations. The
Combined Automated NGS Sentosa Workflow can provide information predictive of
therapeutic response at the time of diagnosis, facilitating proper early drug selection
and thus reducing the active disease period and minimizing treatment cost for
patients infected with hepatitis C virus.
ID75. Comparative Evaluation of the Omniplex-HPV and RFMP HPV
PapilloTyper for the Detecting of Human Papillomavirus Genotypes in Cervical
Specimen
Y. Yoon, Y. Choi
Soonchunhyang University Hospital, Cheonan, South Korea.
Introduction: Human papillomavirus (HPV) is a major cause of cervical neoplasia
development. HPV screening is very important because early treatment can prevent
cervical cancer. Omniplex-HPV is a luminex type polymerase chain reaction assay
that is designed for detecting 40 HPV genotypes. The aim of this study is to evaluate
the analytical and clinical performance of the Omniplex-HPV comparing with that of
the commercially available restriction fragment mass polymorphism (RFMP) HPV
PapilloTyper. Methods: A total of 2,808 cervical swab specimens was obtained. Of
these, only 1,799 specimens has a cytology result. Direct sequencing test was
performed by reference method in case of discrepancies between 2 test results.
Results: The overall percent agreement (OPA) between Omniplex-HPV versus
RFMP HPV Papillotyper were 97.9% (κ=0.84; 95% CI:0.81-0.88). The positive
percent agreement (PPA) and negative percent agreement (NPA) were 98.0% and
96.2%, respectively. The Omniplex-HPV and RFMP HPV PapilloTyper showed
comparable sensitivities (98.0% and 99.7%, respectively) and specificities (same as
91.8%), while the Omniplex-HPV produced more accurate results and required less
turnaround time and labor. Conclusions: The agreement between 2 methods
showed excellent, not only for HPV genotyping but also clinical performance. In
conclusion, Omniplex-HPV produced results comparable to the RFMP HPV
PapilloTyper.
990
ID76. A Clinical Performance Evaluation of QPLEX STI Detection Kit
S. Cho1, M. Kim1, B. Song1, S. Park2, Y. Yeon1, J. Jang1, M. Seong1, S. Kwon1, S.
Park1
1Seoul National University Hospital, Seoul, South Korea; 2QuantaMatrix Inc., Seoul,
South Korea.
Introduction: Eight STI (Sexually Transmitted Infections) types genotyping kit was
developed based on the QMAP (QuantaMatrix multiplex assay platform) technology
which is suspension bead array system. Methods: This kit’s core technology is a
coded bead system for a multiplex bioassay. Coded beads were fabricated from
lithography process and tetra-ethyl-orthosilicate polymerization. After that the beads
were coated with silica shell and then carboxyl groups were generated chemically on
the surface of the beads. QuantaMatrix Inc. has also developed decoding software
and reader system. To assess the validity of the clinical detection of QPLEX STI
Detection Kit compared to AccuPower STI8A-Plex Real-Time PCR Kit (Bioneer,
Korea). Results: Eight microorganisms (C. trachomatis, M. genitalium, M.
hominis, U. urealyticum, T. vaginalis, N. gonorrhoeae, herpes simplex virus type1
and type2) were evaluated on 737 vaginal swabs and 673 urine samples. Between 2
kits, agreement was excellent, regardless of species, both for the swab samples
(kappa = 0.982 ~ 1) and urine samples (kappa = 0.977 ~ 1). Conclusions: Clinical
detection of sexually transmitted diseases by QPLEX STI Detection Kit is as
accurate as AccuPower STI8A-Plex Real-Time PCR Kit. Our novel coded bead
technology and QMAP system provides multiplex solutions for molecular diagnosis
of STI.
ID77. Comparison of the Hologic Panther Fusion Respiratory Assays to BioFire
FilmArray Respiratory Panel for Detection of Respiratory Viruses in Children
A. Leber, H. Wang, L. Sherman, L Sherman, K. Everhart
Nationwide Children’s Hospital, Columbus, OH.
Introduction: The Panther Fusion Respiratory Assays (Hologic, Investigational Use
Only) was compared to the FilmArray Respiratory Panel (BioFire, v1.7) for the
detection of Influenza A (Flu A), Influenza B (Flu B), Respiratory Syncytial Virus
(RSV), Parainfluenza 1 (Para 1), Parainfluenza 2 (Para 2), Parainfluenza 3 (Para 3),
Parainfluenza 4 (Para 4), Adenovirus (Adeno), human Metapneumovirus (hMPV),
and Rhinovirus (Rhino). Methods: A total of 600 frozen archived pediatric
nasopharyngeal (NP) swab specimens previously characterized as either negative or
positive for at least one of the above-mentioned respiratory viruses by FilmArray RP
were evaluated using the Panther Fusion Respiratory Assays. Results: Overall
agreement between Panther Fusion Respiratory Assays and FilmArray RP assays
for ten respiratory viruses was 98.2% (kappa = 0.96 [95% confidence interval {CI},
97.9 to 98.4). The combined positive percent agreement was 91.8% (95% CI, 89.6 to
94) and the negative percent agreement was 99.0% (95% CI, 98.7 to 99.2). The total
number of discordant results for Flu A was 3, Flu B was 5, RSV was 10, Para 1 was
3, Para 2 was 4, Para 3 was 4, Para 4 was 6, Adeno was 28, hMPV was 14, and
Rhino was 32. Detection of Flu A, Para 2, and Para 3 was comparable between
methods. The FilmArray RP detected more Adeno and Rhino while the Panther
Fusion Assays detected more Para 1, hMPV, Flu B, RSV, and Para 4.
Conclusions: The Panther Fusion assays demonstrated overall comparable results
for detection of respiratory viruses compared to the FilmArray RP. The Panther
Fusion assays for Para 1, hMPV, Flu B, RSV, and Para 4 appear to be more
sensitive than the corresponding assay for the FilmArray RP. The high number of
discordants for Adeno and Rhino may be attributed to differences in inclusivity and
specificity, respectively. The 2 methods may have variable detection for the 57
known human adenovirus serotypes representing gentoypes A – F. Furthermore, in
contrast to the FilmArray Rhino assay which cross reacts with enterovirus(es), the
Panther Fusion assay is designed to be Rhino-specific and does not appear to cross
react with enteroviruses. The Panther Fusion Respiratory assays may provide an
additional option for respiratory testing.
ID78. Stability of Zika Virus and Recombinant Zika Controls
H. Greiss1, S. Ramu1, M. Kaur2, L. Fishel1, C. Rotman2, M. Botros1, A. Fahim1
1FC Lab, Downers Grove, IL; 2Oakbrook Institute of Endoscopy, Oakbrook, IL.
Introduction: Currently the common practice for clinical labs is to test ZIKV in frozen
or refrigerated serum and/or urine. ZIKV detection in semen samples are done on
fresh or frozen specimens. This often poses a problem to peripheral labs and adds to
the testing cost by dry ice shipping. A simple and cost effective collection and
handling method will alleviate problems associated with ZIKV testing. Here, we
evaluated the stability of SeraCare Life Sciences AccuPlex recombinant ZIKV in
Hologic Aptima Urine Specimen Transport Tubes for processing and handling of
semen from suspected patients. Methods: Pooled plasma and semen samples were
tested neat or after being spiked with various copy numbers of ZIKV in Aptima urine
collection kit. Plasma and Semen were tested for Zika on the Panther Hologic
system following manufacturer’s guidelines. The Hologic Aptima Urine Specimen
Transport Tubes were spiked with recombinant ZIKV and tested using the Aptima
Zika assay. For processing, a recombinant ZIKV spiked individual or pooled negative
semen was mixed with Aptima Urine Transport Medium at the ratio of 1:1 (2 mL of
semen was added into Aptima Urine Specimen Transport Tube, which contained 2
mL of urine transport medium). Each Aptima collection kit was spiked with
predetermined quantity of recombinant Zika RNA from SeraCare; Zika Accuplex at
10 viral copies/ml run in duplicates. Blank plasma and semen as well as samples
spiked with negative control were also run. Each sample was tested in duplicate, on
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
day zero, 84 hours, 7 days, 15 days and 21 days. Donor Zika plasma was run neat
and in Aptima RNA preservative as well. Results: Both plasma and semen either
neat or processed in Aptima urine collection kits tested positive up to 15 days after
storage at 2-80 C. 100% of samples tested positive, however, one sample of the
processed plasma tested negative at 21 days after storage. There was no difference
in storage of the donor samples, spiked samples either straight or in Aptima RNA
preservative even at low as 10 viral copies /ml. Conclusion: There is an urgent
need for a reliable and sensitive assay to detect Zika virus RNA from semen and
plasma with reliable storage and handling of samples without the panic time
restraints, freezing requirements and shipping on dry ice. It is recommended to make
Zika testing on plasma and semen tests readily available to labs nationwide to be
shipped refrigerated both straight or processed in Aptima Urine collection kits.
ID79. Evaluation of Cross Reactivity and Inhibitory Effects of Sexually
Transmitted and Mosquito Borne Pathogens on Zika Testing Using Aptima
Zika Virus Assay on the Fully Automated Panther System
H. Greiss1, S. Ramu1, M. Kaur1, L. Fishel1, C. Rotman2, M. Botros1, A. Fahim1
1FC Lab, Downers Grove, IL; 2Oakbrook Institute of Endoscopy, Oakbrook, IL.
Introduction: Multiple arboviruses that can cause febrile illness with rash, myalgia,
or arthralgia coexist and often transmitted in the same geographic areas. The
laboratory testing is more important to confirm the etiology of these diseases.
Therefore, Zika, Chikungunya, and Dengue virus infections should all be considered
for patients with acute fever, rash, myalgia, or arthralgia who have traveled within the
previous 2 weeks to an area with ongoing transmission or are living in an area with
ongoing transmission. In order to rule out interference of these and the sexually
transmitted pathogens on semen samples we tested the Aptima Zika Virus assay on
spiked semen samples at molecular level for cross reactivity. Methods: Semen was
tested for Zika on the Panther Hologic system following manufacturer’s guidelines.
Each Aptima collection kit was spiked with predetermined quantity of AccuPlex
recombinant Zika RNA from SeraCare, at 10 copies /ml of semen or plasma. Each
sample was tested 5 times, first run in single, second and third run in duplicate. On
the same run blood plasma samples previously tested negative for Zika were tested,
and straight control materials at 300 copies per ml each for Chlamydia trachomatis
(CT), Neisseria gonorrhoeae (NG), Trichomonas vaginalis (Trich), Human
papillomavirus (HPV), Chikungunya AccuPlex, Dengue AccuPlex, Dengue
seroconversion, Chagas seroconversion and West Nile Virus seroconversion to
assess for cross reactivity. The specimens with control pathogens above with 300
copies per ml were also spiked with recombinant Zika virus at 10 copies/ml and
tested to assess for inhibition. Results: The pooled semen samples tested was
negative for Zika virus. However, the Aptima assay detected as low as 5 viral copies
per ml when tested on the spiked plasma and semen sample. None of the other
infectious controls were detected positive for Zika. The assay demonstrated 100%
sensitivity and specificity to Zika viral RNA from Semen samples with lower detection
limit of 5 viral copies per ml in semen. None of the tested pathogens demonstrated
inhibition on Zika testing. Conclusions: There is an urgent need for a reliable and
sensitive assay to detect Zika virus RNA from semen samples to help and prevent
transmission of Zika virus through sexual acts or infertility treatment. This is the first
time; a commercial clinical assay was evaluated for detection of Zika virus RNA from
Semen. Our results indicate that, the Zika viral assay on the fully automated Panther
system could be a valuable tool and we strongly recommend screening all semen
samples for Zika prior to using in ART procedures or in sperm banking.
ID80. Evaluation of the Abbott Real-Time RT-PCR EBV Assay for EBV
Detection and Quantification
M. Yoon, J. Lee, S. Park, M. Seong
Seoul National University Hospital, Seoul, South Korea.
Introduction: There has been a need to develop reliable diagnostic assays for
quantifying EBV load. We evaluated the performance of a commercial Abbott Realtime EBV assay, which targets BLLF1 gene. Limit of detection, linearity and
reproducibility were assessed using the Qnostics EBV analytical panel (Qnostics,
Ltd., United Kingdom). Methods: Total 120 whole blood samples were analyzed to
evaluate reproducibility the agreement between the Abbott Real-time EBV assay and
the Nanogen EBV REAL-TIME Alert Q-PCR assay (Nanogen Advanced Diagnostics,
Italy), used in our laboratory. Results: The LOD of the Abbott Real-time EBV assay
was 1.69 log10 IU/mL. The assay was linear from 1 x 102 to 1 x 105 IU/mL (r2 =
0.997). The within-run coefficients of variation (CV) ranged from 1.68% to 3.46%,
and the between-run CV ranged from 1.73% to 12.83% for samples with high,
medium and low EBV loads. For 96 samples, positive with Abbott Real-time EBV
assay and with Nanogen EBV REAL-TIME Alert Q-PCR assay, were significantly
correlated with both assays (r2 = 0.773, P < 0.0001). The matched 96 samples
demonstrated a positive bias of 0.11 log10copies/mL for the Abbott Real-time EBV
assay to the Nanogen EBV REAL-TIME Alert Q-PCR assay. Conclusions: The
Abbott Real-time EBV assay demonstrated reliable performance for EBV DNA
quantification in whole blood and correlated well with to the Nanogen EBV REALTIME Alert Q-PCR assay.
ID81. Performance of the Hologic GBS Assay on the Fully Automated Panther
Fusion System
C. Hentzen, B. Eaton, D. Kolk, B. Grobarczyk, L. Franzil, R. Close, M. Exner
Hologic, San Diego, CA.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
Introduction: Streptococcus agalactiae, or Group B streptococcus (GBS), is a
common bacterium often carried in the intestines or lower genital tract. GBS is
usually harmless in adults, however, in newborns it can cause severe complications.
Thus, GBS testing is a routine part of prenatal care in the United States. The Hologic
Panther Fusion GBS Assay (in development) is a fully-automated in vitro diagnostic
test utilizing real-time PCR for the rapid, qualitative detection of GBS from clinical
specimens. The assay is intended to aid in determining the GBS colonization status
from vaginal/rectal swabs collected from antepartum women. In this study, the
analytical and clinical performance of the assay was evaluated. Methods: Analytical
sensitivity was evaluated by testing serial dilutions of cell lysates in Lim Broth media
for GBS serotypes Ia, III, IV and V. The 95% detection limit of the assay for each
serotype was determined using probit regression analysis. Additionally, the predicted
detection limit was confirmed across 11 GBS serotypes. Comparability across 3
common enrichment broths (Lim Broth, Carrot Broth & Trans Vag Broth) was
demonstrated for 2 GBS serotypes (III & V) by testing dilutions at 0.5x, 2x and 5x the
predicted detection limit. Clinical performance of the assay was evaluated by testing
residual Lim Broth enriched specimens (n = 255) collected from antepartum women
receiving standard of care GBS culture screening. The sensitivity and specificity
versus culture was determined for the Hologic Assay. Specimens with discrepant
results were tested with the BD Max GBS assay. All testing was performed on the
Panther Fusion System. Results: The analytical sensitivity for the Hologic GBS
Assay had predicted 95% detection limits of 27 CFU/mL for type Ia, 31 CFU/mL for
type III, 23 CFU/mL for type IV, and 46 CFU/mL for type V in the test sample. The
detection limit confirmation testing yielded 100% detection for the 11 GBS serotypes
at concentrations ≤ 100 CFU/mL. The enrichment broth comparison testing resulted
in comparable performance for both GBS serotypes across the 3 included broths.
The clinical performance testing yielded a sensitivity of 100% (43 of 43 detected) and
specificity of 98.8% (206 of 209 not detected), when compared to culture. The 3
culture negative, assay positive specimens tested positive with the BD Max GBS
assay. Conclusions: The preliminary sensitivity, specificity, and clinical comparison
studies demonstrated excellent detection of GBS across evaluated serotypes and
enrichment broths. The robust clinical performance of the Hologic Panther Fusion
GBS Assay in combination with the system’s throughput, random access, and full
automation offers a novel approach to GBS testing.
ID82. HPV Genotypes in Precancerous Lesions and Cervical Cancer of Korean
Women
E. Lee
Green Cross Laboratories, Yongin, South Korea.
Introduction: The cervical cancer is the fifth frequent cancer in Korean women.
Infection with human papillomavirus (HPV) is considered to be a pre-requisite for the
development of cervical cancer. The prevalence of human papillomavirus (HPV)
varies between countries according to ethnics and geographical area, and leads to
different effectiveness of prophylactic HPV vaccines We investigated the cervical
cancer and CIN 2/3 rate for 4 leading HPV genotypes, type 16, 18, 52 and 58 in
Korean women. Methods: We reviewed HPV genotype data of liquid based cytology
tests of 1,267 patients diagnosed with cervical intraepithelial neoplasia grade 1 to 3
and invasive cancer by cervical biopsy from Jan 2016 to March 2017. The testing
method of HPV genotypes was a validated HPV genotyping chip kit that can detect
34 different HPV subtypes (PANArray HPV genotyping chip, Daejeon South
Korea). Results: HPV type 16 revealed 46.75 %, 50.68 % and 2.57 % in CIN 1, CIN
2/3 and invasive cancer respectively. The distribution of HPV type 18 in CIN 1,
CIN2/3 and cancer were 79.05 %, 20.95 % and 0 % in sequence. HPV type 52
showed 76.48 %, 22.03 % and 1.49 % in CIN 1, CIN2/3 and cancer respectively.
HPV type 58 positive rate in CIN 1, CIN2/3 and cancer were 22.90 %, 74.81 % and
2.29 % in order. In cancer, HPV type 16 is the most frequent type and followed by
58, 52 and 18. HPV type 16 and 58 were more highly distributed in CIN2/3 and
cancer as 53.25 % and 77.10% compared to 20.95 % of HPV 18 and 23.51 % of
HPV 52. On the other hand, the positive rate of HPV type 18 and 52 were much
higher in CIN 1, mild dysplasia. Conclusions: In Korean women, HPV 16 showed
most highly associated with increased cancer risk, and HPV 16 and 58 showed a
more highly cancer progression rate compared to 2 other types, HPV 18 and 52.
There are 3 kinds of HPV vaccine, Cervarix (GlaxoSmithKline, London, United
Kingdom) for HPV type 16 and 18, Gardasil 4 (Merck and Co, Whitehouse Station,
NY) for type 6,11, 16 and 18, and Gardasil 9 for type 6,11, 16,18, 31, 33, 45, 52 and
58. Based on this study results, Cervarix and Gadasil 4 have the limitation for the
prevention of Korean women’s cervical cancer. Gadasil 9 is the most effective HPV
vaccine for Korean women. For the most effective vaccination program of each
country, the HPV prevalence data in each country should be considered.
Informatics
I01. An End-to-End Bioinformatics Pipeline Optimized for Somatic Variant
Analysis Returns Clinically Actionable Results with a Rapid Turnaround Time
R. Kamal, S. Shekar
Fabric Genomics, Oakland, CA.
Introduction: NGS testing of tumor variants is increasingly performed in many
clinical settings, but is often inefficient, with pipelines cobbled together from several
991
AMP Abstracts
separate pieces of software. Here, we demonstrate an end-to-end bioinformatic
pipeline, from FASTQ files through to a clinically actionable test report, optimized for
the analysis and reporting of somatic variants. Methods: A complete solution was
designed that takes FASTQ files generated from an Illumina next generation
sequencer, uploads them to the cloud, completes secondary and tertiary analysis
and returns molecular and clinical interpretation. Specifically, FASTQ files from the
sequencer were uploaded into a custom somatic secondary analysis pipeline
(Omicia Opal) which aligned and called variants for each sample. Variant calling was
performed using the same mathematics as the Mutect2 somatic variant caller for
both SNP and InDel calling. VCF files obtained from the secondary analysis pipeline
were automatically accepted into the Opal tertiary analysis pipeline, and those
variants annotated. Variant filtering was performed to remove germline variants
using population frequencies given by the 1000 Genomes dataset. Selected variants
were curated against publicly available literature and scored according to AMP
somatic scoring guidelines. Results: The secondary analysis pipeline showed high
sensitivity against known benchmarks. Tertiary analysis provided molecular and
clinical interpretation for each sample, with at least 2 targeted therapy matches and
multiple clinical trial matches reported for each tumor sample analysed. End-to-end
data processing and analysis for this pre-curated panel was completed in under one
hour. Case-specific interpretation using the same pipeline can be provided with nextday turnaround. Conclusions: The Opal Clinical Somatic bioinformatics pipeline
demonstrates a seamless and accurate method to process, annotate and clinically
interpret somatic variants. Rapid turnaround time on the processing and analysis of
NGS data enable faster reporting of clinically actionable results. We expect this to
drive better cancer care.
I02. Informatics to Illuminate Real-World Genetic Test Ordering Practices at a
Large Academic Institution
V.A. Arboleda, R.R. Xian, L. Dardick, A.A. Wu
University of California Los Angeles, Los Angeles, CA.
Introduction: The emergence of large, multi-gene sequencing panels has
revolutionized medicine in the ability to provide rapid diagnostic and prognostic
information, and inform individualized treatment strategies. The rapid explosion in
the number, type and complexity of genetic testing has inundated physicians and
created new informatics challenges in privacy protection, ordering practice, and
storage of clinical genetic test results. The goals of our study are to understand the
scope of genetic testing at the University of California, Los Angeles (UCLA) Health in
terms of ordering provider complement and frequency of orders, and to identify
barriers to ordering and resulting of genetic tests within our EHR. Methods: We
mined the EHR to categorize all genetic-based tests that were not available for direct
order through our Epic-based EHR but ordered through our “miscellaneous
laboratory test” order over a 6-month period. We then conducted a survey of
physicians including primary care, clinical specialists and genetic subspecialists to
understand the true scope of genetic test ordering practices at UCLA. Results: Our
analysis of the “Miscellaneous lab order” option within our EHR suggested that a
significant fraction of genetic tests were ordered directly from 3rd party vendors
(outside of the EHR). We sent our survey evaluating genetic test ordering practices
to 679 physicians spanning the clinical specialties of pediatrics, neurology,
hematology-oncology, obstetrics and gynecology, family medicine and internal
medicine. We had a 6.1% (n=55) response rate. All specialties surveyed were
represented (in training to 20+ years post-training). While 96% stated they ordered
genetic testing in their clinical practice, the majority of physicians ordered less than
10 genetic tests per year. 68% of physicians stated there were significant barriers to
ordering genetic tests, the most prominent or which were insurance authorization
and ease of ordering (e.g., identifying correct paper form or electronic order). Over
half of physicians ordered genetic tests outside of the EHR, which prevents a direct
record of the order and result in the patient’s medical record. Only 70% of physicians
stated they “always” scan a copy of the test result into the
EHR. Conclusions: Understanding ordering practices of physicians is a key
component to ensuring access to all clinically important tests and to documenting
results in the medical record for the patient’s future care. Finally, investigating the
scope of ordering practices factors into laboratory decisions for which tests can be
performed in-house. Our findings highlight the importance of leveraging informatics
tools and directly surveying physicians to better identify and address barriers to
genetic test ordering.
I03. Real-Time Thermodynamics and Local Variant Display for Primer
Selection
Z.L. Dwight, C.T. Wittwer
University of Utah, Salt Lake City, UT.
Introduction: Designing molecular assays for PCR and high-resolution melting often
requires convoluted workflows utilizing a variety of software applications and
websites. Data sources are cumbersome for novice users and advanced
computational expertise is a requirement to consume digital genetic
resources. uVariants is a web-based, streamlined assay design and SNP context
tool that requires user input for: 1) rs# and 2) the number of flanking bases for
investigation. Annotation surrounding the SNP of interest include the location of
neighboring variants, exon/intron casing, minor allele frequencies, and specificity
assessment of possible 3’ primer placement. The software presents visual,
contextual information and allows the user to place their own primers, with real time
992
thermodynamic parameters, and automatically consolidates, saves and sends the
results to other tools and formats. Methods: The rs number is the only required input
but is accompanied by an option to define a specific number of bases flanking each
side of the queried variant. A customized ENSEMBL database of variants is queried
with the rs number provided by the user to determine neighboring variant locations
and characteristics. A second query is required via web services (NCBI E-utils) for
the most up to date minor allele frequency information. Neighboring SNPs are
annotated on a frequency-based color gradient to visualize severity to possible
primer placement and are linked to their respective dbSNP information pages at
NCBI (https://www.ncbi.nlm.nih.gov/projects/SNP/). The user then highlights regions
of the sequence to place a primer on the annotated sequence. Upon placement, the
primer sequence is sent via web service to an internal server script to determine
melting temperature, length, and GC-content, which is then displayed. Results:
Once the user has submitted the rs number and the data retrieval / annotation
algorithm is completed, all information is displayed. The user can place as many
primers as necessary and a running list is maintained. The list of primers is
interactive - the user can select any primer pairing to send to other tools such as
uMelt (product melting behavior) or UCSC In-Silico PCR (specificity). All the
resulting SNP, sequence, and primer information may be exported and saved for
future use in a non-proprietary format (.csv or .txt). Conclusions: uVariants provides
a quick and easy interface to interrogate a SNP and the surrounding genomic region
for the purpose of designing robust and high quality assays. The software is freely
available, without registration in a web application,
at https://www.dna.utah.edu/variants/.
I04. Cloud-Based Somatic Pipeline Development and Validation for Clinical
Somatic Variant Detection, Including Large Indels, from Targeted Panels
A. Bolia1, M. Monroe1, B. Kennedy1, K. Boehme1, J. Weeks1, R. Sunderland1, B.
O'Fallon1, A. Kellogg1, L. Yang1, N. Tirpankar1, K.E. Simmon1, J. Durtschi1, B.L.
Brulotte1, C. Krueger1, D.A. Nix2, J. Schumacher1, P. Rindler1, R.R. Bastien1
1ARUP Laboratories, Salt Lake City, UT; 2Huntsman Cancer Institute, Salt Lake City,
UT.
Introduction: Single gene mutations such as substitutions, and small and large
insertions/deletions, have important diagnostic, prognostic, and therapeutic impact in
a variety of cancers including both leukemias and solid tumors. A number of opensource somatic variant callers detect mutations from aligned paired-end nextgeneration sequencing (NGS) data. However, most somatic variant callers
processing short reads are limited by their inability to call both small and large indels
simultaneously. We show that with thorough experimental vetting using wellcharacterized samples during bioinformatics development, a highly sensitive somatic
pipeline can be developed to successfully call across a range of variant types,
including large deletions. Methods: An extensive literature search on somatic variant
callers identified fifteen callers that were evaluated using in-silico simulations,
synthetic large indels and a dataset of 158 myeloid malignancies samples with wellcharacterized point mutations, complex Multiple Nucleotide Polymorphisms (MNPs),
indels, and orthogonally-confirmed large internal tandem duplications (ITDs). The
bioinformatics component of somatic pipeline was constructed as a Snakemake
workflow here referred to as “Somatic Pipey”. A cloud infrastructure was constructed
in Amazon Web Services (AWS) using a combination of open source tools to support
asynchronous job queuing to run Somatic Pipey in production. Results: Somatic
Pipey is comprised of 3 variant callers to identify a variety of variant types with high
sensitivity: Lofreq for single nucleotide variations (SNVs), Scalpel for indels 2-60 bp,
and Manta for indels > 60 bp. Somatic Pipey detected simulated somatic MNPs up to
5 bp with a 98.2% true positive rate and showed 100% concordance for ~ 4000
common SNPs and > 500 manually reviewed variant calls generated by the
production pipeline at the 5% limit of detection (LOD). Somatic Pipey also detected
large ITDs up to 228 bp and a large synthetic deletion upto 2.8 kb. At 5% LOD
Somatic Pipey had a maximum false discovery rate of 12.5%, which can be
significantly reduced by filtering on quality score. Conclusions: Somatic Pipey is a
fast, robust, and scalable cloud-based NGS solution that deploys a clinical
bioinformatics workflow for tumor-only samples to call multiple somatic variant types
using a software strategy that combines 3 somatic variant callers into a single
pipeline. Performance on simulated data, synthetic mutation standards, and patient
samples confirm the high sensitivity of Somatic Pipey in identifying somatic
mutations, including large indels. Following thorough validation according to
CAP/AMP guidelines, this workflow will be deployed onto the AWS cloud for
production usage in a secure and scalable environment.
I05. A Computational Framework for Large-Scale Analysis of TCRβ Immune
Repertoire Sequencing Data on Cloud-Based Infrastructure
L. Lin1, T. Looney1, G.M. Lowman2, E.A. Linch2, D.S. Topacio-Hall2, L. Miller2, J.
Zheng1, A. Pankov1, J.K. Au-Young1, M. Manivannan1, A. Kamat1, M.R.
Andersen2, F.C. Hyland1
1Thermo Fisher Scientific, South San Francisco, CA; 2Thermo Fisher Scientific,
Carlsbad, CA.
Introduction: TCRβ immune repertoire analysis by next-generation sequencing is
emerging as a valuable tool for research studies of the tumor microenvironment and
potential immune responses to cancer immunotherapy. Generation of insight from
immune repertoire profiling often requires comparative analysis of immune
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
repertoires across research sample cohorts representing immune responses to
defined antigens or immunomodulatory agents. Here we describe the development
of a computational framework enabling large-scale comparative analysis of immune
repertoire data on cloud-based infrastructure. Methods: TCRβ receptors were
amplified from matched peripheral blood and tumor biopsy cDNA using AmpliSeq
multiplex primers targeting the Variable gene Framework 1 and Constant gene to
produce an amplicon spanning all 3 CDR domains. To evaluate assay performance,
we sequenced TCRβ rearrangements from donor peripheral blood leukocyte (PBL)
cDNA that had been spiked with 30 reference rearrangements taken from
literature. Raw data was uploaded to the Ion Reporter data analysis platform for
clonotype annotation and storage to enable rapid downstream comparative analysis
of repertoire features. Results: We demonstrate the ability to rapidly compare
clonotype data across sample cohorts and find that a subset of clones identified in
peripheral blood are also found in matched tumor samples. Peripheral blood-derived
repertoires typically contained 10 to 100 fold more distinct clones than found in
tumor, with correspondingly higher estimates of diversity via the Shannon Index.
Results from sequencing of spike-in reference rearrangements indicate that the
assay is accurate and sensitive over 5 logs of input template amount while showing
minimal amplification bias. Technical replicates showed high concordance (r >.96) in
the frequency of detected clones, indicating that results were reproducible and
samples were sequenced to an appropriate depth. Comparison of AmpliSeq
multiplex PCR-derived data to that produced by 5’ RACE or BIOMED-2 primers
revealed the AmpliSeq solution to provide comprehensive and unbiased coverage of
the human TCRβ repertoire. Conclusions: In summary, we have developed a
computational framework to enable rapid analysis of large immune repertoire
datasets derived from AmpliSeq-based sequencing of human TCRβ receptors via
the Ion Torrent S5. The AmpliSeq procedure, which features the ability to produce
uniform and reliable results in extremely highly multiplexed PCR, is well suited for
immune repertoire sequencing applications.
I06. A New Allele-Centric VCF File for Variants in ClinVar
M.J. Landrum, C. Chen, J. Lee, M. Benson, G. Brown, S. Chitipiralla, B. Gu, J.
Hart, D. Hoffman, W. Jang, K. Katz, C. Liu, Z. Maddipatla, A. Malheiro, K.
McDaniel, M. Ovetsky, G. Riley, R. Tully
National Institutes of Health, Bethesda, MD.
Introduction: ClinVar is a fully public database at NCBI that archives interpretations
of the clinical significance of genetic variants. Interpretations of clinical significance
are submitted to ClinVar by clinical testing laboratories, research laboratories, expert
panels, and others. The data may be browsed on the web or accessed
programmatically with E-utilities. The data are also available for download from the
ftp site: ftp://ftp.ncbi.nlm.nih.gov/pub/clinvar/. The XML files are the archive of the full
ClinVar dataset. However, many users download ClinVar data in VCF format, which
is simpler than XML to parse and which may be used to filter data of interest.
ClinVar’s current VCF files are organized around the dbSNP rs number. This means
that in some cases, data for more than one allele are reported on a single row, which
can make the data harder to parse. It also means that ClinVar variants that are not
yet registered in dbSNP are excluded from the file. Our goal is to generate a new
VCF file for ClinVar data that is allele-centric and that includes all variants in ClinVar
with a precise genomic location. Methods: A proposal for allele-centric VCF files
was generated and distributed to users of the existing ClinVar VCF file. Adjustments
were made based on feedback. The allele-centric VCF files were released in a “beta”
form on the ftp site for GRCh37 and GRCh38:
(ftp://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh37/vcf_2.0/) and
(ftp://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/vcf_2.0/). Results: The new VCF
files are allele-centric, such that each row represents data for a single allele. If
ClinVar has reports for more than one allele at the same location, there is more than
one row for that location in the file with different ALT alleles. The files include all
variants in ClinVar for which there is a precise location. In other words, variants with
“fuzzy” boundaries, such as CNVs detected by microarray, are not included in the
files. The ID for each row is the ClinVar Variation ID, rather than the dbSNP rs
number. The rs number, when available, is included in an INFO tag. The new VCF
files also include several INFO tags for data about “included” variants. These are
defined as variants for which we have an interpretation for a haplotype or genotype
that includes the variant, but no interpretation for the variant itself. INFO tags for
other data types were also modified for easier parsing, including those for allele
origin, molecular consequence, and allele frequency. Conclusions: ClinVar’s allelecentric VCF files are available for download and feedback is welcome. Future
development will focus on providing VCF files for variants with imprecise location,
such that all variants in ClinVar that are localized on the genome are represented in
VCF format.
I07. Custom-Built Heuristic Approach to Variant Calling Tools Development
D. Thakral1, N. Rodic2
1Yale University, New Haven, CT; 2University of Florida, Gainesville, FL.
Introduction: Bioinformatics analysis of next-generation sequencing data is uniquely
challenging. Many genomic data are obtained by outsourcing to marketed software
platforms such that validity of bioinformatics methodology cannot be easily
ascertained. Because the success of clinical genomics is predicated on accuracy of
nucleotide variant identification and interpretation, we developed a systematic
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
custom-built variant calling pipeline. Methods: Using the support of the Yale Center
for Research Computing service as a base for pipeline building and performance we
created an independent workflow solution for resulting NGS-based data sets.
Specifically, by empirically combining read aligner, BWA-MEM, with variant caller,
SAMtools, and annotation toolset, Ensembl Variant Effect Predictor, we created a
customized bioinformatics pipeline to analyze next generation sequencing data from
Illumina2500 platform. The approach entailed a combination of Unix-based operating
system, including command-based tools and bash shell programming. Results: We
resulted whole-exome datasets with favorable performance specifications including:
i), entirely independent pipeline, ii) a turn-around-time of less than 1 day, and iii)
automated software running. The pipeline provides a complete solution starting from
raw FASTQ sequence files to generation of customizable Excel-based worksheets
that contain nucleotide variants shortlists suitable for analysis and interpretation by
medical geneticists and/or molecular genetic pathologists. Conclusions: Our
pipeline provides an alternative to variant calling process. Because it is open source
and free to use, it supports full replicability of results.
I08. Advancing Genomic Knowledge Curation: Piloting the Use of Enhanced
Literature Curation Tools
R.J. Schmidt1, M.J. Kiel2,4, B. Funke3,4
1Harvard Medical School, Boston, MA; 2Genomenon, Inc., Ann Arbor, MI; 3Harvard
Medical School/Massachusetts General Hospital, Cambridge, MA; 4Veritas Genetics,
Danvers, MA.
Introduction: In the era of exome sequencing, establishing and maintaining up-todate clinically valid gene-disease associations has become a crucial bottleneck in
laboratory operations affecting test development and result interpretation. In current
practice, relevant literature is typically identified through PubMed searches, which
return citations by matching terms in the title and abstract. The recent development
of new tools that serve as an alternative or adjunct to PubMed with added features
including full-text searching has the potential to improve the performance of current
literature curation workflows. We conducted a systematic comparison of
conventional (PubMed) and alternative approaches (Genomenon MASTERMIND
[GM], Google, Google Scholar, PubTator), to evaluate their effectiveness in
supporting the curation of gene-disease relationships for hypertrophic
cardiomyopathy (HCM). Following an initial feasibility assessment, we performed a
structured, in-depth analysis comparing PubMed and GM. Methods: We examined
10 genes that are frequently included in diagnostic testing for HCM with a range of
known gene-disease association strengths (MYH7, MYBPC3, TPM1, TNNI3, TNNT2,
ACTN2, CSRP3, TNNC1, NEXN, VCL). Three search strategies were used for each
gene – 1) a PubMed search reflecting “real world” practice, 2) a PubMed search with
medical subject headings (MeSH)-based disease terms which represent a curated
vocabulary that is shared between PubMed and GM, 3) a GM search. Unique
returned references were manually curated blind to their source using the Clinical
Genome Resource (ClinGen) framework to identify relevant genetic or experimental
evidence. Results: Across all genes examined, 1,910, 1,436, and 2,432 PMIDs were
returned for the PubMed, PubMed/MeSH, and GM search strategies, respectively.
GM increased the number of results by 69.4% over a matched PubMed/MeSH
search. Additionally, GM provided an increase of 27.3% over our laboratory’s “real
world” PubMed search. However, 22.0% of PubMed results were not found by GM.
Subsequent improvements to GM reduced this number to 3.9%. By manual curation
of the returned citations, GM was able to identify a greater number of useful
references for 4 of 6 genes examined. Across these 6 genes, the overall percentage
of useful references were 58.2%, 63.3%, and 47.0%, respectively. Conclusions:
GM improves literature curation sensitivity when used in conjunction with traditional
PubMed-based search due to its expanded search capabilities. The number of useful
results returned by GM exceeded our PubMed search strategy but additional
improvements are needed to eliminate “off-target” results.
I09. Repository of Quality Control and Metrics: A Web-Browser Based
Application for Review and Approval of Clinical NGS Quality Metrics
L.M. Peterson, C.K. Keso, E.J. Winter, P.L. Hare, N.R. Mattson, R.L. Gonzalez, R.D.
Gnanaolivu, T. Stroope, D.S. Mussell
Mayo Clinic, Rochester, MN.
Introduction: Next Generation Sequencing (NGS) technologies generate many
quality metrics used to evaluate sample and sequencing run quality. Clinical
laboratories have traditionally relied on manual, spreadsheet based methods to
capture, review and store quality metrics. These methods are time consuming,
difficult to maintain, and prone to error. We have developed the Repository of Quality
Control Metrics (RoQCM) application and database to log, store, review, and
approve quality metrics important to NGS. RoQCM is available for the Clinical
Genome Sequencing Laboratory (CGSL – the clinical NGS core facility) and CGSL
client laboratories within the Department of Laboratory Medicine and
Pathology. Methods: CGSL is able to review quality control (QC) metrics for NGS
instruments and individual samples or controls. RoQCM collects sample and
instrument metrics such as total read count, target read count, undetermined reads,
cluster density, and depth of coverage. The application can display data for a single
run or graphically show instrument trends over time. RoQCM allows a user to
predefine thresholds for any sample or control metric, with threshold violations
prominently displayed in red at the top of the review page. RoQCM is configured to
query a known set of variants associated with a control sample and determines if all
993
AMP Abstracts
variants expected are present, absent, and if there are unexpected variants
associated with the control. A user either passes or fails QC with a sign off feature
and comments can be entered at a run, batch, or sample level. Once CGSL has
completed review, the data and comments are passed to the client laboratory for
their review process. Results: The system accepts metrics from a wide variety of
pipeline sources and displays them for review and approval. The web-browser based
interface provides a single location for review and storing of NGS quality metrics and
is available to both CGSL and client laboratories. RoQCM reduces time spent for QC
review by approximately 50% over manual review methods. Conclusions: A typical
NGS hereditary panel run will generate nearly 700 metrics, with approximately 243 of
these being distinct for any given sample or control allele. Since its launch in fall
2016, RoQCM has collected a total of 21,342 different metrics, at the sample and
instrument run level, allowing CGSL to monitor and track QC information day-to-day
and over time from a centralized location. Quality metrics are automatically
populated and threshold violations are prominently displayed allowing for rapid
review, reducing both manual errors and review time. Users are able to review single
run QC or review more in-depth for trending. QC metric review is available for CGSL
and their customers.
I10. Improving Quality Control of Gene Amplification (GA) Detection in an
Amplicon-Based Next-Generation Sequencing (NGS) Cancer Gene Panel by
Implementing Gene-Level Segment Visualizations
Y. Lo, Z. Walther, M. D'Eletto
Yale School of Medicine, New Haven, CT.
Introduction: Gene amplification (GA) is a common mechanism of oncogene
activation in malignancy. Because GA is often therapeutically targetable, NextGeneration Sequencing (NGS)-based assays that include copy number variation
(CNV) detection capability are increasingly used for clinical tumor profiling. However,
there is wide variability in the degree of GA that may be observed among tumor
samples, and standards for interpreting their potential significance are not
established. In particular, partial gene amplifications, where CNV segments do not
cover all exons of a gene, can be difficult to interpret based on numerical values
alone, such as Phred-scaled confidence scores. We therefore investigated ways to
improve GA detection and interpretation. Methods: We developed a gene-level
visualization method for assessing chromosome segments identified as amplified by
an NGS CNV algorithm. This tool allows one to visualize the putative GA segments
as well as the individual bins (amplicons) that comprise them. By zooming into a
region of interest, pathologists can view individual bins of the GA segment as well
their relation to the introns and exons of all gene transcript isoforms in the region.
Seventeen cases with partial gene amplification were identified from our institutional
Oncomine Comprehensive Assay (ThermoFisher Scientific) clinical database and
were reviewed retrospectively using this viewer to show its utility. Results: This
visualization program provides pathologists with a way to easily differentiate full
versus partial gene amplifications, as well as to identify amplified regions that may
have been incorrectly assigned during the segmentation process. Our 17 cases had
putative GAs detected by the CNV algorithm that had been assigned a confidence
score of >20 (range 21-107) and ploidy ≥5. The involved genes – including
PIK3CA (3 cases), as well as ERBB2, AKT1 and MYC (2 cases each) - are covered
by 11 to 36 amplicons each. Our visualization program enabled these CNV
segments to be identified as partial GAs at a glance. One case with
partial ERBB2 GA was further investigated and found to be negative
for ERBB2 amplification by fluorescent in situ hybridization
(FISH). Conclusions: Our gene-level CNV segment visualization program is a
powerful interpretation tool for pathologists. It provides the pathologist an intuitive
view of CNV data and sufficient relevant genomic information to rapidly recognize
partial gene amplification as well as to identify cases worthy of further investigation
by FISH. This program is now being incorporated into routine practice in our clinical
laboratory. We expect its implementation to significantly increase the efficiency of
our NGS CNV data interpretation while helping to ensure accurate gene amplification
reporting.
I11. Establishing Seamless Electronic Connectivity, an Underestimated
Exercise for Instituting a High Quality Genomic Medicine Service
N. Sidiropoulos
University of Vermont Medical Center, Burlington, VT.
Introduction: Low-cost NGS instruments, an expanding pool of clinically relevant
genes, improved access to high-quality bioinformatics, and increasing access to
targeted therapeutics are key factors driving implementation of genomic testing for
clinical use. UVM Medical Center is developing clinical NGS as "care pathways" as
opposed to assay development solely driven by the laboratory. Seamless electronic
integration of information systems supporting elements of a genomic care pathway is
fundamental to a high quality genomic medicine service and patient safety. We
present lessons learned from building an electronic infrastructure to support a
genomic care pathway for patients with solid tumors. Methods: A descriptive
analysis and report is presented of the effort to institute seamless electronic
connectivity across systems that form the informatics infrastructure supporting
clinical genomics. A draft workflow is created, points of connectivity or lack thereof
are identified, and solutions are proposed. Results: Key information systems that
form the infrastructure for clinical genomics are identified: electronic health record,
laboratory information system for anatomic pathology and the genomics laboratory,
994
and bioinformatics system. Existing electronic interfaces do not support clinical
genomics and require unique solutions to support a clinical genomics workflow.
Conclusions: A genomic care pathway for patients with solid tumors begins with the
clinical suspicion of a tumor, the decision to acquire tissue for diagnosis, accurately
testing the tumor for the genomic profile, using the genomic report at the point of
care, and making clinical decisions with the patient based on the reported findings. In
an effort to establish seamless electronic connectivity across systems, local
practices and logistics should be considered early in the effort to establish a clinical
genomics assay. This facilitates workflow considerations and exposes connectivity
pitfalls that may then be addressed by personnel with pertinent expertise.
Addressing of IT connectivity pitfalls can take longer than expected as the effort and
resources to address connectivity is often underestimated. In our experience, it is
essential to address electronic workflow early in the effort to institute clinical
genomics as the system infrastructure is complex and critical to the perceived quality
of the genomic laboratory service and ultimately patient safety.
I12. Comparison of an Automated Approach to Mining the Genomic Literature
Against COSMIC, a Manually Curated Database
M. Kiel1, L. Chunn2, D. Nefcy1, S. Schwartz1, R. Tarpey3
1Genomenon, Inc., Ann Arbor MI; 2Grand Valley State University, Allendale,
MI; 3University of Michigan, Ann Arbor, MI.
Introduction: Mastermind (MM) is a database of genetic variants produced through
automated data-mining of the medical literature. To determine the validity of using an
automated approach to assembling such a database, we compared the results in
MM with the Catalogue of Somatic Mutations in Cancer (COSMIC) for
comprehensiveness of variant identification and completeness of referenced
sources. Methods: To assess the comprehensiveness of the results in MM, more
than 9,000 protein-coding variants in 407 cancer-associated genes were selected at
random from the COSMIC database. The presence of each variant and the
COSMIC-associated references in MM was assessed. To assess the extent to which
MM identified additional variants not present in COSMIC, 50 of the 407 genes were
selected at random and all variants and associated references in the COSMIC
database were compared with all variants and references for each gene found in
MM. The nature of each of the additional variants found in MM but not in COSMIC
were further assessed according to established variant classification criteria by
manual review. Results: In total, 2,975 articles were identified by COSMIC of which
96.5% were also identified by MM demonstrating an effective automated strategy to
identify high-yield content. Furthermore, MM found a total of 20,645 articles
representing a 5.9-fold increase in total references. Of those articles that were not
found in MM, 38 (1.3%) were foreign-language and 23 (0.8%) were no longer in the
PubMed index and therefore inaccessible to MM. Overall 9,329 COSMIC variants
were examined for concordance in MM. Of these, MM found 96.5% of these variants.
Of the 3.5% that were missed, the variants were described in figures that did not
include a text description (1.0%), tables that presented the variant description across
multiple columns (0.7%) or otherwise text descriptions of the variant that were
difficult to parse automatically (0.8%). Finally, for the 50 random genes, a total of
6,743 variants were identified in COSMIC. For the same genes, MM identified
11,014 that were each manually reviewed for accuracy representing a 63% increase
in variants. These results illustrate the ability of an automated approach to mining
the genomic literature to find more variants than manual methods. Conclusion: An
automated approach to mining the literature identified 96.5% of the articles and
variants identified in COSMIC, a database that has been manually curated for more
than a decade. In addition, our automated approach identified 5.9-fold more articles
and 63% more variants than COSMIC. These results illustrate the utility of
automated text-mining techniques to identify clinically meaningful content and the
superiority of MM results to current gold-standard techniques.
I13. Overlapping Variants Can Lead to Potential for Missed Calls in Custom
Next Generation Sequencing Bioinformatics Pipeline
C. Vanderbilt, D.L. Aisner, D.T. Merrick, M. Yourshaw, K. Davies
University of Colorado, Anschutz Medical Campus, Aurora, CO.
Introduction: Bioinformatics pipelines (BPs) are relied upon heavily in next
generation sequencing (NGS) laboratory testing, as the computational complexity of
the genomic level data sets is beyond what can be performed manually. On the other
hand, manual review and visualization of the data plays a role in assuring the quality
of the BP performance. We present here for consideration a potentially problematic
deficiency in our BP that could be a common clandestine issue in other laboratory
BPs that was identified on manual review of data. Methods: On our amplicon-based,
Illumina TruSight Tumor panel, we have built a custom BP internally validated to
identify single nucleotide variants and small insertion/deletion variants in solid
tumors. Each variant is manually reviewed in Integrated Genomics Browser by the
technologist, pathology resident and attending prior to sign out. Potential deficiencies
are identified on an ongoing basis, with ongoing quality control and
improvements. One potential deficiency was identified through manual case review,
and quality control was applied by manual review of all other cases. Results: The
index case that brought the BP issue to our attention was a colon adenocarcinoma
case in which the BP identified 6 variants in 4 separate genes. On manual review of
a SMAD4 variant p.D537G c.1610A>G at variant allele frequency (VAF) 11.7% that
was identified by BP, we identified a second p.L536Qfs*14
c.1607_1613delTAGACGA variant (VAF 10.8%) in the same region as the called
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
variant. The second variant was not identified by the BP. On further review of our
data, the deficiency was found to theoretically have implications in our BP that could
lead to clinically significant missed calls. Specifically, the BP will only call one variant
at each nucleotide locus, due to specific settings within ANNOVAR. To ensure
adequate clinical performance, we manually reviewed all lung adenocarcinoma
cases from the prior 18 months to ensure that clinically relevant EGFR variants were
not missed due to "suppression" in areas where multiple variants may have been
identified, but only one called. Specifically, benign polymorphisms, which typically
have a VAF of ~50%, and would theoretically mask a low VAF variant were
examined. No additional problematic cases have been identified. A systematic BP
solution is being developed, however vigilance for such issues in manual review
remains crucial. Conclusions: It is important to recognize that even in highly
scrutinized and thoroughly validated BPs, unforeseen deficiencies may exist and the
clinical laboratory team should continue to look for points of failure. Manual review of
variants remains a crucial part of NGS variant identification workflow and ongoing
quality control and quality improvement.
I14. Evaluation of Structural Variant Callers from a Clinical Perspective
C.P. Johnson, S. Raghunath, P.J. Mishra
Intermountain Healthcare, St. George, UT.
Introduction: Structural variants or rearrangements have a long history as
significant therapeutic indicators, such as FLT3 Internal Tandem Duplications (ITDs)
in AML. These variants are hard to target by commonly used somatic variant
callers. Therefore incorporating dedicated structural variant callers can be beneficial
in a clinical pipeline. In this study we evaluated 2 structural variant callers against
clinical samples and standard cell lines. We then develop a pipeline which can
reproducibly call structural variants with a minimum manual intervention, and
evaluate resulting sensitivity, lower limit of detection (LLOD), and
specificity. Methods: As we are interested in calling structural variants in myeloid
malignancies, we focused on mutations in FLT3 and CEBPA. Our laboratory
sequenced 10 clinical samples from 4 patients with AML in duplicate using our newly
developed 63 genes myeloid malignancies panel. Eight of these samples were
reported to contain FLT3 ITD alteration based on previously performed NGS and/or
PCR based tests. Along with patient samples, we utilized commercially available
standard cell line containing a 218bp CEBPA knockout. We analyzed clinical
samples using Pindel and Delly to determine concordance with prior results. Delly
can be configured to detect either germline or somatic mutations. The somatic
parameter in Delly requires a paired normal sample. As our oncology panel uses
tumor only samples for testing, we sequenced HapMap cell line NA18507 in each
run as a control. We then improved variant calling specificity by including read depth
and read length filters, as FLT3-ITD is defined by a minimum duplication
length. Results: In the clinical settings, Pindel successfully detected 7/7 clinical
FLT3-ITDs. Delly only detected 3/7 in germline mode and missed several variants in
somatic mode. This concordance was also observed for Pindel in a duplicate run,
showing the call was reliable. During pre-filtration stage we see an average of 413
variants per sample in FLT3 and CEBPA. This average was decreased to 294
variant after applying a minimum read length filter of 3, and 9 after filtering on a
minimum alternate read depth of 3. We do not lose clinically significant variants and
maintain 100% concordance to previous results after filtration. The expected variant
at lowest AF was used to determine LLOD at a low 4.19%. Conclusions: This study
suggested that our internally developed-proprietary bioinformatics pipeline with the
addition of Pindel can accurately call long structural variants in the newly developed
63 genes myeloid malignancy panel. Our workflow was capable of calling actionable
clinically significant structural variants in myeloid malignancy panel with high
reproducibility, sensitivity, and specificity.
I15. Transmission of North American Association of Central Cancer Registries
(NAACCR) Data Using the Lung Biomarker Template from the College of
American Pathologists (CAP)
K.I. Hulkower1, R.L. Moldwin1, S. Baral1, S.F. Jones2, J.D. Rogers2, J.E. Seiffert2
1College of American Pathologists, Northfield, IL; 2Centers for Disease Control and
Prevention, Atlanta, GA.
Introduction: In 2018, cancer registries will be adding biomarkers and prognostic
factors that are required or recommended for staging in the 8th ed. of AJCC’s Cancer
Staging Manual. NAACCR, CAP, and registry standard setters have recently
validated an interoperable, web-based, data-submission method using CDC’s Web
Plus abstracting application with a breast biomarker template included in CAP’s
electronic Cancer Checklists (eCCs). We now describe the expansion of NAACCR’s
data item fields to include additional biomarker data elements that are clinically
useful for lung cancer patient diagnosis, treatment and prognosis. While the AJCC
currently mandates the collection of only EGFR and ALK mutational analysis for lung
cancer, the CAP lung biomarker template also supports the collection of ROS1, RET,
KRAS, BRAF, ERBB2, and MET mutational analysis as additional data
elements. Methods: We incorporated CAP's lung biomarker template into CDC’s
Web Plus abstracting application. The CAP template was provided as an XML
document in Structured Data Capture (SDC) format. The SDC lung biomarker form
completed by the abstractor is joined with a full NAACCR abstract, and may be
transmitted as an XML file to one or more pre-designated locations. The biomarker
data are mapped to values and locations compatible with NAACCR’s legacy flat file
format, and the lung biomarker XML (in SDC format) is stored in the database as an
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
indexed XML column associated with the registry’s patient and tumor records.
Results: We have successfully integrated the CAP lung biomarker template into
Web Plus to securely capture, transmit, and store CAP biomarker data as SDC XML
along with the associated NAACCR full record abstract, in legacy flat file format, in
the central cancer registry database. The lung biomarker data recorded in SDC XML
format were successfully mapped to existing NAACCR data items and integrated into
the appropriate database fields. Conclusions: We have successfully demonstrated
a method to replace NAACCR’s flat file data transfer system with an SDC solution
that expands biomarker value sets using CAP’s lung biomarker template. Future
work will include additional content and value sets provided by CAP anatomic
pathology and biomarker eCC templates. eCC biomarker and anatomic pathology
data elements in SDC format will be useful for enhancing the depth of data element
coverage in registry datasets and will also be valuable in other contexts, such as
clinical research, workflow quality assessments, and enriching genomic data with
high quality "phenotypic" information.
I16. Homopolymer Compression Improves Reference-Free, Kmer Based Whole
Genome Strain Comparison for Ion Torrent Data
K.E. Simmon, M. Mallory, B.A. Couturier, C. Krueger, E.P. Gee, A.P. Barker, M.A.
Fisher
ARUP Laboratories, Salt Lake City, UT.
Introduction: Epidemiological outbreak investigations often require comparing
bacterial strains. Whole genome sequencing (WGS) has emerged as an alternative
to Pulsed-Field Gel Electrophoresis or Multi-Locus Sequence Typing and can
provide more clinically actionable information. Reference-free comparisons allow
analysis without the need of a reference genome and is an attractive approach given
the incomplete set of suitable references for every organism. Here we compare the
efficiency of kmer-based strain comparisons between the Illumina NextSeq 500 (NS)
and the Ion-Torrent S5 XL (ION) using StrainTypeMer. We show that by applying
homopolymer compression (HC) to the output reads, we can drastically improve
strain comparison performance by reducing the prevalence of sequencing errors
within the data. Methods: Replicates of Staphylococcus aureus USA 300 were
sequenced on the ION (n=8) and the NS (n=7). Low-frequency kmers, with a count
<2.5% of the estimated coverage were removed. Kmers retained were compared to
those found in the reference genome, and kmers not present in the reference
genome were identified as kmer errors. Kmer error rates were compared with and
without HC, which is defined as the reducing homopolymers runs in any read to a
maximum length of 2. Each replicate was then compared pairwise using
StrainTypeMer to determine the effectiveness of HC across instruments.
Results: After removing low-frequency kmers from sequencing runs an average
99.8% of the reference kmers (ASM1346) were covered regardless of instrument or
HC processing. With and without HC, the number of distinct kmers in the reference
was 2.60 and 2.87M, respectively. An average of 2.2% and 1.6% kmer-errors were
produced in each sample by the ION and NS, respectively. With HC the number of
kmer-errors is reduced to 0.3% and 1.4% on the ION and NS, respectively. Pairwise
kmer-identity between replicates data increased from 97.9% to 99.6% and 97.6% to
97.8% with compression on ION and NS samples, respectively. Further analysis on
the non-HC processed NS data shows that 63.5% of the kmer-errors were 1hamming distance away from a high-copy plasmid sequence, as compared to 2.5%
of the kmer errors identified in the non-HC processed ION data. Conclusion: The
ION produces high-quality data suitable to perform epidemiological strain typing
without a reference genome, where performance improves by applying HC to reduce
kmer-errors by ~87%. Unlike the NS, ION errors are more easily identified and can
be reduced by simple preprocessing of the reads using HC. Performing strain typing
on an ION provides faster turn-around times with streamlined sample processing.
Rapid kmer-based methods for strain typing can lead to wider adoption of WGS for
epidemiological outbreak investigation.
I17. Bioinformatics Assay Design for Development of Multiplex PCR-Based
Next Generation Sequencing Panels
D. Wang, P. Chan, H. Gong, S. Meyers, D. Do, J. Qin, J. Brockman, R.
Ramakrishnan, G. Sun
Fluidigm Inc., South San Francisco, CA.
Introduction: Targeted gene sequencing is rapidly advancing clinical research for
screening complex diseases such as cancer. Multiplex PCR-based, targeted
sequencing panels focused on oncogenes and tumor suppressors provides a
powerful tool enabling detection of low-abundance, actionable mutations from
heterogeneous tumor specimens. Assay design using a design pipeline plays a
critical role in developing robust PCR-based panel products. In this study, we
developed a completely automated assay design pipeline to design and pool assays
for selected target regions, including full genes and hot spots for use with the
Juno Targeted DNA Sequencing Library Preparation System. Methods: To
accurately and evenly amplify target regions of a panel, we developed a
bioinformatics assay design pipeline, written in Perl, as an alternative to Primer3,
considering the following assay features: 1) assay thermal parameters; 2) dimer
formation; 3) design coverage; 4) amplicon size; 5) amplicon overlap rate; 6)
amplicon GC content; 7) potential SNPs in primers and their associated allele
frequency; and 8) individual assay specificity and interaction with others in an assay
pool. Each factor is weighted and can be dynamically changed using different design
modes in order to achieve the best balance of target coverage, evenness, and
995
AMP Abstracts
potential mispriming. Results: We developed a solid tumor panel covering 53
oncogenes and tumor suppressors. A total of 1,540 primer sets were designed using
the design pipeline to achieve 100% target coverage, with targets pooled into 38
groups. The preliminary analytical verification results demonstrated that 95.4% of
reads aligned to target regions of interest and 95.7% of assays (1,474/1,540) have
read coverage uniformity within a 5-fold range of average. The sequencing results
from Illumina NextSeq using reference samples of NA12878 and Horizon Tru-Q HDx
further indicated that with average amplicon sequence depth of 3,500x, and 5%
VAF threshold for variant calling, the sensitivity and positive predictive value
are 98.8% and 98.7% respectively for these samples. Conclusions: We have
developed a robust bioinformatics assay design pipeline for development of
PCR-based next generation sequencing panels. It employs complex features for
assay design in addition to regular PCR thermal parameters to balance design ability
and PCR efficiency of assays. The assay performance has been verified with a solid
tumor panel, and the pipeline can be used to develop other panels.
I18. An Open Software Ecosystem for High Throughput Clinical Diagnostics
K.D. Doig1,3, A. Fellowes2, T. Conway1, A. Seleznev1, D. Ma1, C. Love1, J. Ellul1, C.
Welsh1, L. Simpson1, J. Li1, R. Lupat1, A. Kumar1, L. Lara1, G. Reid1, A.T.
Papenfuss2, S.B. Fox1
1Peter MacCallum Cancer Centre, Melbourne, Australia; 2The Walter and Eliza Hall
Institute of Medical Research, Parkville VIC, Australia; 3The University of Melbourne,
Melbourne, Australia.
Introduction: The widespread availability of sequencers has allowed DNA to be
examined at nucleotide resolution. Clinically reporting on biomarkers in a patient’s
DNA requires many steps; wet lab, sequencing, bioinformatics and analysis.
Completing these tasks in volume in a clinically relevant turnaround time needs a
sophisticated ecosystem. Our experiences evolving clinical workflows have informed
the development of engineered systems that are open, standards compliant and
robust. Methods: The Peter MacCallum Cancer Centre has been performing genetic
testing for patients for over for 20 years and high throughput sequencing (HTS) for 6
years. The systems described here have been developed over a period that has
seen sample volumes increase at a rate of 26% per month. Results: The lack of
integrated systems to perform HTS task has prompted in-house development of a
suite of applications. The principles supported by the components include:
·
·
·
·
·
·
·
·
·
·
·
Containerisation – Docker created pre-built images allow rapid deployment
into a wide range of environments without dependency management issues.
Decoupling – Systems can be decomposed into discrete components with
minimum interdependency.
Open source – Systems are available on Github under open source licenses
or are being documented for release.
The ecosystem comprises the following components for a complete clinical
sequencing platform.
Holly: A system for controlling liquid handling robotics and interfacing to the
Illumina Clarity sample management system.
PathOS: A decision support tool to manage, analyse, curate and report on
HTS variants. PathOS is available at github.com/PapenfussLab/PathOS and
can be deployed via Docker.
Mario: A workflow and notification system coordinating inter-task
communication between sequencers, pipelines, data loaders and PathOS.
Muthur: A functional testing system performing functional testing on PathOS
to validate behaviour through a browser interface.
Varys: A sample monitoring dashboard for tracking samples as they progress
through the ecosystem from registration to report.
Canary: An amplicon pipeline utility which creates a VCF file with HGVS
variants and rich annotations. Available as a Docker image
from github.com/PapenfussLab/Canary
Babble: External hospital LMS are integrated by translating HL7 to JSON.
Conclusions: The ubiquity of sequencers has exposed the limited availability of
software to support clinical HTS. A laboratory processing hundreds of patient
samples per week needs a software ecosystem supporting best practice software
engineering. The systems cited above are a suite of open source components
allowing a wide range of pathology laboratories to take advantage of HTS
diagnostics.
I19. NeGeSeI – An All-Purpose Decision Support Tool for the Clinical
Management of Next Generation Sequencing Assays in the Clinical Laboratory
V. Williamson1, A. Kusmirek1, V.S. Gadepalli1, I. Gonzalez1, L. Gonzalez Malerva1, A.
Popa1, V. Borodin1, O. Rafael1, C. Yang1, F. Sábato1, C.N. Vlangos1, C.I. Dumur2, A.
Ferreira-Gonzalez1
1Virginia Commonwealth University Health System, Richmond, VA; 2Aurora
Diagnostics, Palm Beach, FL.
Introduction: Routine clinical use of next generation sequencing (NGS) has
generated the need for a comprehensive tool that collects data throughout the
course of sample processing and employs all information collected as a component
of the variant interpretation process. The tool should standardize data collection
regardless of sequencing platform and allow for seamless integration of NGS results
996
into existing hospital electronic record systems. Methods: Here we
describe NeGeSeI, a tool designed in our lab to meet all of these tasks and to score
somatic variants according to current AMP-CAP-ASCO guidelines. Written in PHP
and MySQL, this tool uses a multi-study population database, e.g. ExAC and
gnomAD, and prediction software to assess variant actionability. The tool also
provides information on relevant clinical trials and FDA approved therapies. A
hierarchical review design protects against accidental data change through a multilayer approval process. All information regarding variant quality, functionality, and
relevance to the patient’s case is deposited into a searchable database as a set of
tissue-specific rules that document the variant’s respective level of pathogenicity as
well as its tier within the AMP-CAP-ASCO system. Unique features of the application
include on-demand assessment of assay quality via Levy Jennings plots and a
PubMed search mechanism that ranks article relevance using an algorithm based on
the author’s publication history, number of article views and journal impact factor.
Administrative capabilities of the tool include a complete security audit system that
tracks user activity and the enforcement of unique single user logins. NeGeSeI can
be run on a standard Intel desktop with a minimum of 16 GB RAM. Results: A total
of 542,693 variants were reviewed (GRCh37/hg19) using NeGeSeI in 159 solid
tumor and hematological patients since its validation in January 2017. Of that
number, 3260 variants met lab-defined metrics for strand bias, frequency and
sequencing quality. Missense single nucleotide polymorphisms were enriched in this
dataset (Fisher’s, p = 0.003) above all other variant types. Approximately, 60% of
patient cases reviewed by the tool generated multiple clinical trial recommendations
(5 or more). FDA-regulated therapies were recommended for 14.4% of the cases as
indicated and overwhelmingly, 75% of the other therapies recommended addressed
both sensitivity and resistance to drugs on a variant and tumor type basis.
Conclusion: A novel tool for the clinical interpretation of NGS results for somatic
variants has been developed. NeGeSeI facilitates greater efficiency of the variant
review process by combining variant quality, along with information on available
clinical trials and FDA-based therapies.
I20. Redesigning the Molecular Pathology Clinical Report for the NextGeneration Genomic Era: The MSKCC Experience with the MSK-IMPACT
Assay
A. Syed, S.H. Madur, J. Birnbaum, A.R. Balakrishnan, D. Chakravarty, J. Gao, H.
Zhang, S.M. Phillips, N.D. Schultz, J. Rudolph, M.E. Arcila, M. Ladanyi, M.F.
Berger, A. Zehir
Memorial Sloan Kettering Cancer Center, New York, NY.
Introduction: The management of advanced cancer patients is increasingly
dependent on precise identification of somatic or germline mutations. Nextgeneration sequencing (NGS) based assays are increasingly used in the clinic to
identify multiple types of genomic alterations in multiple genes which need to be
reported clinically in a clear, concise, and readily interpretable fashion. Further,
traditional pathology laboratory information systems are often poorly suited for
storing and reporting the results of complex, multi-gene clinical genomic assays.
Here, we describe a systematic redesign of the clinical genomic report to address
these challenges. Methods: MSK-IMPACT (Zehir et al., Nat. Med., 2017) is an NGS
based assay offered in the Molecular Diagnostics Service at Memorial Sloan
Kettering Cancer Center, interrogating 468 cancer genes to identify point mutations,
small insertions and deletions, copy number variants (CNVs) and select structural
variants (SVs). We designed a portable document format (pdf) based clinical report
that communicates this information along with levels of actionability of somatic and
germline variants based on OncoKB (Chakravarty et al., JCO PO, 2017), a precision
oncology knowledgebase. Python-based ReportLab is used to generate the reports
from MPath, an in-house NGS results database, automatically and the reports are
transmitted to the patient’s electronic medical record using standard HL7 messaging.
Results: The MSK-IMPACT report is divided into four sections: i) administrative data
points ii) genomic alterations iii) assay performance of the sample and iv) general
regulatory disclaimers. Genomic results begin with a concise summary of results,
followed by sample level notes the molecular pathologist may want to relate to the
treating physician. Each genomic alteration is associated with its annotation, variant
allele fraction for mutations and fold change for CNVs, whether it is recurrent and/or
oncogenic and finally levels of evidence for actionability. Table footnotes are used for
variant level notes. List of alterations are followed by detailed variant annotations
and treatment implications in that indication. A technical summary section contains
assay performance information followed by general regulatory disclaimers. A
glossary of terms is included at the end of the report that explains more technical
terms used in the reports for clinicians who may not be familiar with genomics.
Currently, 800 MSK-IMPACT reports are generated per month. Conclusions: Large
panel NGS based molecular testing is becoming an integral part of the standard of
care for cancer patients. A redesign of existing molecular pathology reports is
required to clearly and concisely communicate the often extensive genomic
information to clinicians.
I21. In-Silico Framework for Detection and Evaluation of Contamination in
Clinical Diagnostic Next-Generation Sequencing
M. Sarmady1,2, M.A. Gonzalez1,2, K. Cao1, B. Krock1
1The Children's Hospital of Philadelphia, Philadelphia, PA; 2The Perelman School of
Medicine at the University of Pennsylvania, Philadelphia, PA.
Introduction: Next-generation sequencing (NGS) has rapidly become a primary
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
methodology used in clinical molecular diagnostics. In part due to the high sensitivity
of NGS assays, contaminating DNA from a different sample could result in a false
positive variant call and possibly an erroneous clinical report. The implementation of
good laboratory practices can reduce the possibility of pre-analytical sample
contamination, but cannot eliminate this risk. Accordingly, it is critical to integrate
methods in clinical bioinformatics pipelines to quantify and screen data for various
levels of contamination. To better understand the influence of cross sample DNA
contamination in NGS assays, we developed an in-silico methodology to model and
evaluate contamination. Here we present our evaluation of a computational tool,
VerifyBamId, to measure contamination in clinical exome sequencing data (CES). In
addition, we share our clinically adopted contamination quality control metrics for use
in clinical bioinformatics NGS pipelines. Methods: We constructed insilico contamination by merging BAM files from different CES datasets at predetermined mixture levels ranging from 0.5-5%. Predicted levels of contamination
were measured using VerifyBamId and compared to the amount of insilico contamination. To investigate the influence of contamination on variant calling,
we constructed in-silico contamination of 6 inter and intra-run CES sequencing trials
of the Genome in a Bottle (GIAB) sample NA12878. First, we used our clinical
pipeline to call variants (GATK HaplotypeCaller) on the in-silico contaminated BAM
files. Next, we filtered to retain variants that were concordant with GIAB gold
standard variants. Variants from the original runs were compared against the insilico mixtures to assess genotype concordance. Results: The VerifyBamId
estimated contamination level conformed well to expected contamination level from
the in-silico mixtures (r2 = 0.9599). At 1% contamination, genotype concordance
begins to deviate from 100%, with a systematic bias towards changes in variant
zygosity from homozygous to heterozygous. A positive correlation was observed
between contamination level and number of discordant calls. Conclusions:
VerifyBamId is an appropriate tool for detection of contamination in clinical pipelines.
Our results indicate that VerifyBamId can reliably detect cross sample DNA
contamination in CES data. We have shown that in high coverage (>100x) CES data,
5% cross contamination begins to significantly affect genotyping and downstream
variant calling. Our proposed validation strategy may be used by clinical labs to
evaluate and incorporate contamination detection in CES. This will lead to improved
data quality and accuracy of clinical NGS diagnostic tests.
I22. HLA on FHIR in the Cloud to Facilitate Entry in Electronic Medical Records
Y.S. Wang1, S.J. Mack2, A. Diallo3, J.J. Sninsky1, M. Machrus1, D.A. Ross1
1CareDx, Brisbane, CA; 2CHORI UCSF, Oakland, CA; 3DNAnexus, Mountain View,
CA.
Introduction: Integrating genomic information into electronic medical records
(EMRs) is currently not efficient due to data complexity. Appropriate structuring and
standardization of data is crucial. HLA is a key component of the immune system
that plays central roles in transplantation, autoimmune disease, infectious disease,
and cancer. While genomic data for the HLA region are among the most critical and
complex data used for patient management, dense HLA sequence data are routinely
reduced to canonical HLA allele nomenclature for transmission to EMRs. Enabling
entry of HLA sequence and structural data as well as associated interpretations to
EMRs would greatly increase the clinical value of the HLA genomic
data. Methods: For more than 2 decades HL7 has facilitated the exchange of
healthcare data and the standardization of information modeling. Fast Healthcare
Interoperability Resources (FHIR) is a developing interoperability standard that
facilitates secure electronic sharing of healthcare data. We used HLA sequence data
generated with 10X Genomics methodology to demonstrate the potential of FHIR to
encompass rich genomic content. Megabase HLA haplotypes were assembled using
bar-coded long molecules sequenced with short reads. Results: We developed HL7FHIR HLA data structures using 10X Genomics haplotypes phased by chromosome
on a scalable bioinformatics and data management platform. The platform is
equipped with security protocols to ensure compliance with CLIA regulations. In
addition, we demonstrated that de novo read assembly surpassed reference-based
analysis in deriving long-range HLA haplotypes by reducing the potential
misinterpretation or loss of information. Individual elements of HML can be
encapsulated in an HL7 message and serves as the current standard for clinical HLA
data exchange, but HL7-FHIR has proved to be more extensible and adaptable for
long-range NGS HLA haplotype data. HL7-FHIR structures facilitated i) retention of
standard HLA allele nomenclature, ii) inclusion of structural variation data, iii)
incorporation of additional genetic diversity across the HLA region, and iv)
transformation of sequence into antigenic determinants
(epitopes). Conclusions: We describe HL7-FHIR structures using 10X Genomics
haplotypes phased by chromosome, on a cloud-based scalable bioinformatics and
data management platform equipped with security protocols, ensuring compliance
with CLIA regulations. The HL7-FHIR format is uniquely positioned to facilitate the
transition of content-rich HLA sequence data into clinical records, permitting new
standards and dimension expansion while maintaining historic serological
discriminations.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
I23. High-Throughput, Low-Pass Whole-Genome Sequencing (LP-WGS)
Method for Single-Cell Copy-Number (CN) Profiling on Ampli1 Whole-Genome
Amplification (WGA) Products for Illumina Platform
N. Manaresi, V. del Monaco, G. Buson, A. Ferrarini, C. Forcato, P. Tononi, F.
Fontana, E. Petrini, C. Bolognesi, C. Mangano, M. Teracciano, G. Medoro, G.
Giorgini
Menarini Silicon Biosystems Spa, Castel Maggiore BO, Italy.
Introduction: LP-WGS is an unequalled method to study copy-number alterations
(CNAs) in single-cells. Here, we describe a single-tube protocol for detecting CNAs
by LP-WGS on Illumina platforms exploiting the deterministic nature of Ampli1 WGA
technology. This method, implemented in the Ampli1 LowPass kit, is designed to
generate dual-indexed sequencing-ready libraries. The method has been validated
on a variety of input samples including single circulating tumor cells (CTCs) and
even challenging single-FFPE cells. Methods: Four single cells from NCI-H23, NCIH441, NCI-H1563 and NCI-H1650 cell lines, were isolated with the DEPArray. For
comparison, these cell lines were also analyzed by array CGH using Agilent
SurePrint 4x180k arrays. Four CTCs and 2 single white blood cells (WBCs) from
peripheral blood (PB) of a lung adenocarcinoma patient were enriched with the
CellSearch System and isolated with the DEPArray. In addition, cell pools (range=6213 cells) and single-cells from pure tumor and stromal populations were recovered
with the DEPArray starting from an FFPE specimen of pancreatic adenocarcinoma.
All sorted samples were amplified using Ampli1 WGA Kit. Starting from
purified Ampli1 WGA products, a novel streamlined, fragmentation-free and size
selection-free method was developed for the generation of Illumina compatible
libraries (Ampli1 LP-WGS). The method is based on 2 single extension reactions and
a final library amplification. Pooled libraries were sequenced on MiSeq generating
≈500,000 single-end reads per library. CNA analysis was performed using ControlFREEC for LP-WGS data. Results: The profiles generated by Ampli1 LP-WGS
method were compared with those obtained by aCGH assay for the 4 cell lines
tested. High agreement in gains and losses was achieved for all single tumor cells
analyzed. As expected, WBC and sorted FFPE stromal pool analysis resulted in
normal flat profiles while several gain and loss events were identified for single tumor
cells from different sample types (CTCs and FFPE). The method produced welldefined profiles showing low noise as measured by Derivative Log Ratio Spread
(0.2<DLRS<0.3) and Median Absolute Deviation (MAD<0.3). Conclusions: Ampli1
LP-WGS allows to accurately characterize the CNA profiles of single cells isolated
with by the DEPArray starting from multiple sample types. The method is streamlined
with a hands-on time of only 4 hours and produces Illumina-compatible libraries
suitable for high-throughput liquid handling automation. The whole workflow meets
the need for rapid results and affordable cost required for massively parallel singlecell analysis such as multiple tumor cells or CTCs, pre-implantation genetic
diagnosis (PGD) and screening (PGS).
I24. Breaking the Turnaround Time Barrier in Next Generation SequencingBased Clinical Mutation Profiling Using an Integrated Workflow and
Informatics Approach
R. Ruiz-Cordero, W. Chen, K.C. Floyd, S. Rodriguez, J. Galbincea, B. Barkoh, M.
Routbort, R. Luthra, R. Singh, D. Hatfield, R. Kanagal-Shamanna, C.C. Yin, Z.
Zuo, S. Loghavi, C.Y. Ok, J. Medeiros, K.P. Patel
University of Texas M.D. Anderson Cancer Center, Houston, TX.
Introduction: Next generation sequencing (NGS) allows simultaneous interrogation
of multiple genes using small amounts of DNA with higher throughput and lower
overall cost. Since the ability to provide a faster turnaround time (TAT) for specific
clinically relevant genes is usually compromised by this multigene testing approach,
we developed an integrated solution comprised of workflow optimization and
bioinformatics pipeline to provide targeted NGS results as early as 48 hours after
sample receipt. Methods: After comprehensive review of existing workflow from
sample collection to final results, we identified opportunities for TAT improvement in
both physical and informatics workflows, including schedule modification for DNA
extraction, OD measurement, library preparation and loading, and triggering the data
analysis steps. Specific protocols for identification and handling of rush specimens
were integrated in existing pathologists triage and wet bench workflows. A custom
informatics solution was developed for targeted interrogation of bi-directional MiSeq
sequencing data, to allow genotyping on specific mutation hotspots on a 28- or an
81-gene panel at least 24 hours before our regular pipeline. Results: Workflow
optimization alone allowed TAT reduction of 2 business days from a 5-day baseline.
The bioinformatics solution (raw data to vcf) that allowed genotyping of 7 (IDH1,
IDH2, JAK2, KRAS, MPL, NPM1, NRAS) out of 28 genes on the panel at least 24
hours earlier was validated against the existing pipeline through an extensive
retrospective correlation of variant calls for 50 existing retrospective runs
encompassing a total of 367 samples (311 clinical samples and 56 controls). At 5%
mutant allele percentage cut-off and >250x coverage, 100% concordance was noted
for all positive and negative calls between the 2 pipelines. Additional parallel review
comprising 72 non-control prospective samples showed similar concordance. Three
additional genes (CALR, CSF3R, SF3B1) were validated on an updated 81-gene
panel with similar concordance. The optimized workflow enables reporting of 10
selected genes as early as 48 hours of the specimen collection for immediate
diagnosis and treatment decisions, followed by final reporting of the entire panel in
the next 1 to 2 business days. Ongoing clinical reporting shows 100% concordance
between focused and entire panel results reported 24 to 48 hours
997
AMP Abstracts
apart. Conclusion: An integrated workflow and informatics optimization approach
allowed reporting of 10 genes with immediate diagnostic and/or therapeutic
importance as early as 48 hours, representing a 60% TAT reduction from the 5-day
baseline. This approach allows faster reporting of selected genes ahead of the entire
panel without the need for separate testing.
available panels and included several genes with a clear association to AML with
potential clinical significance that are otherwise not present on any such panel.
Conclusions: We have used MM as a tool to rapidly and comprehensively
interrogate, organize and display disease-gene-variant association data to inform
rational gene panel design.
I25. Vetting Targeted Capture Probe Design with a Computational Strategy
Combining KmerSniper and BLAT
A.E. Kellogg, A. Bolia, K.E. Simmon, R. Margraf, P.M. Rindler, R.R. Bastien, J.A.
Raney, A.A. Hall, L.V. Furtado, B. Kennedy, K. Gligorich, E.P. Gee
ARUP Laboratories, Salt Lake City, UT.
Introduction: Probe design of targeted-capture next-generation sequencing (NGS)
assays is challenging due to non-unique sequences within the genome, a property of
pseudogenes, repetitive elements, and low complexity regions. In developing a new
assay targeting solid tumor hotspot mutations (SNVs, indels, and DNA translocations
in ALK, RET, ROS, and NTRK1), we developed tools and an analysis strategy that
aided in optimizing target capture. Non-unique sequences lead to promiscuous
capture of the target, creating downstream bioinformatics challenges in alignment, by
reducing on-target coverage/mapping quality, and in variant calling. We developed a
combined approach using BLAT and an internally developed kmer tool (KmerSniper)
to address alignment and uniqueness mappability issues. Methods: Probes were
obtained from Integrated DNA Technologies. Sequencing was performed on a
NextSeq 500 and reads were aligned to the human genome using BWA-MEM.
Detrimental probe sequences were determined using a combined approach to
address alignment and uniqueness with BLAT and KmerSniper, respectively. BLAT
identified alternate mappings for >40 bp sequences, allowing mismatches.
KmerSniper, alternatively, uses a 30 bp sliding window and reports the frequency of
each kmer in the human genome and alternate mapping locations. KmerSniper
leverages kmer and alignment (Bowtie) data to annotate these intervals. The
combined results were manually reviewed during iterative probe design to improve
on-target efficiency. Results: KmerSniper provides higher resolution output as
compared to the UCSC Mappability Track (Duke Uniqueness, 35bp) by indicating the
exact frequency of each kmer in the genome and provides the locations of the
alternate mapping sites. In total 45 of 950 probes were removed and 45 new probes
were manually designed to avoid the most problematic regions and generate read
depth over difficult targeted regions. For example, probes were designed and
empirically tested for difficult translocation areas (including introns with repeats) to
improve read depth over the regions with the least impact in creating off-target
reads. The probe redesign improved on-target reads from 10-60% in the initial
design to ~80% in the final probe design. Conclusions: This approach provided
more resolution than using standard mappability tracks alone and led to faster
evaluation and redesign of probes. KmerSniper with BLAT analysis allowed probes
to be efficiently designed by eliminating sequences that would cause hybridization
and alignment challenges. This will aid in designing probes around problematic
regions, in order to enable the inclusion of medically-relevant genes with known
pseudogenes by focusing only on targeting subsequences with high capture and
alignment performance.
I27. Creating Custom Gene Panels for Next-Generation Sequencing:
Optimization of 5000 Gene Assays, Selection by Disease Research Area and
Integrated Analysis for Variant Prioritization
F. Hyland1, M. Manivannan1, B. Krishnaswami1, N. Vissa1, C. VanLoy2, A.
Broomer2, T. Biorac2, E. Williams2, Y. Zhu2, Y. Tian2, M. Andersen2, F. Hernandez
Guzman2, A. Kothandaraman1, S. Roman2
1Thermo Fisher, South San Francisco, CA; 2Thermo Fisher, Carlsbad, CA.
Introduction: Next-generation sequencing gene panels enable the examination of
multiple genes, identifying previously described variants and discovering novel
variants, to elucidate genetic disease. The challenges are substantial, including:
identification of all genes of interest; assay optimization to create robust,
reproducible, multiplex panels; and developing accurate, comprehensive,
reproducible analysis pipelines. Methods: We describe a comprehensive set of tools
to simplify targeted gene selection, sequencing and analysis. First we developed the
Content Selection Engine to describe the relationship between genes and diseases,
enabling browsing through a disease hierarchy and selection of genes by disease
categories, along with a scoring algorithm (rank-weighted sum score, RWSS) which
scores every gene by clinical relevance for all diseases across the disease
hierarchy. We optimized assays for each of 5000 disease research genes, creating a
database of empirical amplicon performance and pairing this with improved assay
design tools. A cloud-based web interface enables easy selection of any diseases,
shows all associated genes, and enables selection of any associated genes and
addition of more genes. We incorporate IGV to facilitate visualization of empirical
base and exon coverage. A custom Ion AmpliSeq On-Demand panel is created
using the performance-verified assays from these genes. Smaller gene panels which
are narrowly targeted to specific diseases, or broader gene panels with hundreds of
genes for extended disease categories, can be created. Genetic disease research
areas include neonatal phenotypes such as metabolic disorders, Severe Combined
Immunodeficiency, musculoskeletal disorders, heme disorders; and late onset
phenotypes such as cardiovascular disorders and cancer predisposition.
Results: We tested multiple panels developed on various types of input material,
including frozen blood and dried blood spots. Uniformity, coverage, on target
mapping, and reproducibility were high in all cases and above established quality
criteria (>90% or > 95%). Finally, a coordinated analysis solution imports information
about the custom panel and provides an integrated analysis pipeline with a simple
and powerful visual interface, including variant calling, functional annotation,
population MAF, predicted protein effect, and annotations including ClinVar, OMIM,
COSMIC, etc. Filtering tools utilizing this information facilitate variant prioritization.
Informatics workflows provide support for single sample, trios, and tumor-normal
paired sample analysis. Conclusions: This integrated suite of analysis tools from
disease area to gene assay design to variant analysis, together with optimized
assays, enables robust and fast development of custom panels.
I26. A Novel Automated Approach to Identifying Disease-Gene-Variant
Associations from the Medical Literature to Inform Gene Panel Design
M. Kiel1, M. Schu2, S. Schwartz1, V. Weigman2
1Genomenon, Inc., Ann Arbor, MI; 2Q2 Solutions, Morrisville, NC.
Introduction: Rational approaches to the design of gene panels for next-generation
molecular diagnostic sequencing assays are complicated by the amount and
complexity of information that must be manually collected and assessed from the
medical literature. Mastermind (MM) is a novel text-mining infrastructure that
automatically identifies disease-gene associations from the titles and abstracts of
more than 30M scientific publications in PubMed and disease-gene-variant
associations from the full-text of 4.5M prioritized articles and can be used for rational
gene panel design for diagnostic purposes. To test the capability of MM, we have
chosen the disease Acute Myeloid Leukemia (AML) as a proof-of-concept for this
approach. Methods: MM search algorithms utilize curated lists of diseases and
genes as initial search parameters to identify disease-gene associations. Customdesigned algorithms scan full-text articles using a comprehensive variant list for
every possible variant. This information is organized according to the strength of the
association based on the total number and quality of citations and the position of key
terms within the text. This data is further annotated for the mention of additional
clinically relevant key terms according to both ClinGen and ACMG classification
schema to prioritize content useful for gene panel design. Results: As a proof-ofconcept that automated identification of disease-gene and disease-gene-variant
associations can be useful to provide a rational approach to gene panel design, we
sought to produce a list of all genes associated with AML and their associated
variants and compare this to lists used by commercial gene panel providers. In total,
11K unique variants in 151 genes were associated with AML and ranked according
to the number of citations for each. Each variant was cited at least once from 3,865
scientific publications. These variants were then classified by the journals in which
they appeared and the resulting data manually inspected for accuracy. Additionally,
custom search algorithms identified recurrent gene amplifications, gene deletions
and gene-fusions resulting from translocations and were also catalogued and
incorporated in the final design. The final panel was compared to commercially-
998
I28. An Interlaboratory Assessment of Complex Variant Detection Using
Multiplexed Positive Controls
S. Lincoln1, J. Zook2, R. Truty1, S. Chowdhury3, A. Fellowes4, S. Mahamdallie5, M.
Ferber6, M. Cleveland2, C. Huang7, F. Tomson7, E. Klee6, W. DeSilva4, S. Seal5, S.
Kingsmore3, N. Rahman5, S. Aradhya1, R. Nussbaum1, M. Salit8
1Invitae, San Francisco, CA; 2National Institute of Standards and Technology,
Gaithersburg, MD; 3Rady Children’s Hospital, San Diego, CA; 4Peter MacCallum
Cancer Centre, Melbourne, Australia; 5The Institute of Cancer Research, London,
United Kingdom; 6Mayo Clinic, Rochester, MN; 7SeraCare Life Sciences,
Gaithersburg, MD; 8National Institute of Standards and Technology, Palo Alto, CA.
Introduction: Next-generation sequencing (NGS) is a capable and cost effective
technique for detecting SNVs and small indels in relatively tractable parts of the
genome. However NGS can have limitations for other variant types. We examined
one clinical laboratory cohort of over 80,000 patients, and found that pathogenic,
medically important variants of these other, challenging types are prevalent. They
comprise over 9% of positive findings in hereditary cancer genes, 10% in cardiology,
12% in neurology, and 19% in metabolic syndromes. The technical challenges
presented by these variants are diverse: 32% are single or sub-exon CNVs, 18% are
large indels or structural variants, 45% are in homopolymer, low complexity or
unmappable regions, and 15% are in a region poorly captured by standard pull-down
kits. Despite their prevalence, developing and validating tests that detect such
mutations is difficult, in part because of the scarcity of positive controls. Methods:
We selected 23 of these challenging variants in 7 commonly tested genes, and
followed a methodology (previously demonstrated for simpler cases) where large
synthetic constructs containing these variants were spiked into a known genomic
background, creating a single control sample with all 23 alterations. This DNA was
provided to 7 laboratories who sequenced it using a total of 9 NGS workflows,
including 5 validated clinical tests with custom bioinformatics and 2 vendor (Illumina,
Ion Torrent) default pipelines. Multiple target enrichment technologies were used, as
well as whole genome sequencing. Results: Twelve of 23 variants were detected by
all 9 workflows, but just 2 of the workflows detected all 23. Many but not all of these
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
limitations were previously known. Importantly, evidence of each variant was present
in the raw data files, suggesting that this control strategy is compatible with diverse
biochemical methods. Raw data for the synthetic variants mimicked that of the
endogenous ones (including similar technical artifacts) demonstrating that controls
such as this may be useful in the development and validation of tests with improved
sensitivity. The vendor-supplied bioinformatics pipelines fared the worst, reinforcing
the importance of carefully selecting algorithms and parameters for use in laboratory
developed tests. Conclusions: Medically important but technically challenging
pathogenic variants are prevalent across a range of inherited conditions. Although
actual patient specimens are also critical, multiplexed synthetic controls can help
efficiently assess the analytic range of a test and can help laboratories develop and
validate methods for these challenging variants.
I29. Estimating Mutation Load from Tumor Research Sample Using Targeted
Next-Generation Sequencing Assay at ≥5% Allelic Frequency
R. Chaudhary1 , D. Cyanam2 , W. Tom1 , J. Au-Young1 , S. Sadis2 , F. Hyland1
1Thermo Fisher Scientific, South San Francisco, CA; 2Thermo Fisher Scientific, Ann
Arbor, MI.
Introduction: Immunotherapies have shown anti-cancer effects in melanoma,
NSCLC, and bladder cancer. High tumor mutation load is associated with positive
responses from immune checkpoint inhibitors. However, current methods to estimate
tumor mutation load often have high infrastructure needs, and require large amounts
of DNA. Herein, we characterize the ability of a targeted panel to estimate mutation
load from tumor research samples using low input. Methods: We developed a single
sample analysis workflow for estimating mutation load per megabase (Mb) from
FFPE and fresh frozen tumor research samples. The assay utilizes a PCR-based
target enrichment panel that interrogates 409 key cancer genes covering ~1.7 Mb of
genomic space. Our workflow requires only 10 ng of input DNA, and enables a 2.5day turn-around time from sample to the final report. The simplicity of the workflow
enables less than 60 minutes of hands-on time for automated library preparation and
templating on a batch of 8 samples. Next-generation Sequencing is performed using
high throughput semiconductor sequencing platform to achieve sufficient depth
(~500x coverage) and accuracy. Our custom analysis pipeline calls variants with
optimized parameters on the tumor sample only, with no matched normal sample
required, and applies filters to remove germ-line variants and background noise.
Results: An in-silico analysis demonstrated the panel has sufficient genomic
coverage to provide high sensitivity (> 90%) and specificity (> 95%) to separate high
and low mutation load samples. Through allele ratio distribution of variants, we
consistently notice that somatic and germ line variants were distributed in Gaussian
peaks of predictable height and allele ratio. Matched tumor-normal analyses on lung
and colorectal samples suggested that the single sample analysis on tumor samples
detects mutation load with strong correlation (r=0.9266) with tumor-normal analysis.
To assess reproducibility, we compared the mutation load in library replicates for a
cohort of 9 samples (FFPE and fresh frozen tumors, engineered control from
Acrometrix, and NIST cell-lines) and observed high correlation (r=0.9967). Our filters
consistently eliminate an average greater than 98% of germ line variants from the set
of all variants called in a single sample analysis workflow. Our pipeline identifies
mutation signatures consistent with specific mechanisms such as UV damage and
spontaneous deamination of 5-methyl-cytosine, as well as base-damage from FFPE
processing. Conclusions: A simple workflow has been developed on the Ion Torrent
sequencing platform with an AmpliSeq panel to estimate per Mb somatic mutation
burden from FFPE and fresh frozen tumor research samples. This solution will
advance research in immuno-oncology.
I30. Improvement of Indel Detection Power by Revising Default Parameter
Settings in Vendor Supplied Next Generation Sequencing Analysis Software
W. Zhang, S. Rapp, A. Cushman-Vokoun, T. Greiner
University of Nebraska Medical Center, Omaha, NE.
Introduction: Clinical labs often start with default parameter settings in
bioinformatics tools. It is straightforward and reduces the complexity of analysis. As
clinical labs continue to collect more data, developing customized parameter settings
could maximize the detection power of indels. The verification of performance of
software version changes by the vendor is highly essential for quality assurance. By
revising the default parameter settings, we were able to 1) detect a previously
missed indel and 2) rescue the identification of an indel missed by an upgraded
software version. Methods: A Gastrointestinal Stromal tumor was extracted and
sequenced using the Ion Torrent 50 gene Ion AmpliSeq Cancer Hotspot Panel v2,
the Ion Chef with Chef Package software version IC.5.2.1 and Ion Torrent Personal
Genome Machine. Sequence data were analyzed using the Torrent Suite v5.2.2 and
variantCaller v.5.2.2.41 with the parameter “allow_complex (AC)” toggled on (value
1) and off (value 0). The second case was a previously sequenced colorectal tumor
that analyzed similarly except that we revised the “Torrent Mapping Alignment
Program (TMAP)” parameter default setting with “context” and reanalyzed the data
without the use of “context”. We then applied the settings to an earlier version of
VariantCaller (v.5.0.4), to verify the consistency of performance. Results: We
compared the variants detected from different settings and verified that the
customized settings identified all expected mutations. We were able to detect
additional indels and confirmed them by Sanger sequencing. In case one, we
discovered an 18bp KIT deletion that was originally missed by the default setting of
AC (value 0). Turning on the setting activates the FreeBayes hypothesis generator
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
leading to the generation of complex candidate variants, therefore increasing the
detection power. All the variants results were consistent in variantCaller v.5.0.4. In
the second case, the default TMAP setting (with context turned on) missed a
3bp FGFR1 insertion due to the right alignment caused by the context parameter,
which confused the VariantCaller. It is a bug in version 5.2.2.41 and we have
reported it to Thermo Fisher Scientific, who was not aware of the change in the
parameter. Conclusions: We identified 2 different bioinformatics issues that could
result in the failure to detect indels depending on the settings of default parameters.
Verification of default parameter performance in software updates is extremely
important to ensure equivalent variant detection.
I31. Dynamic Levels of Evidence Tiering to Support Evolving Guidelines in
Variant Assessment
X.S. Li1, R. Duren1, N. Tackes1, R. Smith1, S. Neeley1, S. Shiller2
1MolecularMatch, Inc., Houston, TX; 2Baylor University Medical Center, Dallas, TX.
Introduction: Depending on locale, molecular diagnostic laboratories find
compliance with evolving guidelines a challenge in the environment of multiple
regulatory and guideline agencies with no historical uniform consensus.
Standardization and consensus are emerging, with recently published ClinGen
Minimum Variant Level Data (MVLD) and AMP/CAP/ASCO guidelines for somatic
variant reporting. Molecular Assertions (MA) addresses at scale comparisons of
recommendations, management of changes as new data emerges, and maintenance
of consistency with legacy reports. MA eliminate repetitive manual curation by
decoupling levels of evidence tiering templates from underlying data model for
evidence attributes defined by MVLD and AMP/CAP/ASCO guidelines. Methods:
Drawing upon previous advances in publications, clinical trials and associative
therapeutics search, MA establish levels of evidence for a given condition, molecular
alteration, clinical significance (prognostic, diagnostic, predictive), therapeutic if
predictive, and citation or source. In defining a cancer specific targeted gene panel,
genes and alterations of interest are identified based on given cancer versus
population prevalence. Publications are then associated based on therapeutic,
condition, alteration/gene, and extracted evidence attributes including clinical trial,
preclinical cell line, etc. Manual oversight defines biomarker class, direction and
clinical significance and type of evidence per alteration. Algorithms then address offlabel (Tier 2C) applications as well as distinguishing between well powered studies
(Phase 2 and above) versus small studies. Based on structured data underpinnings
human readable narratives are generated to populate reports and provide rationale
for recommendations. Results: At time of writing 2,080 MA spanning lung, breast
and colorectal cancers have been curated covering 59 genes, 626 alterations, and
3,899 pieces of evidence. In 3 weeks MA will number over 4,000 with completion of
lung panel. Tiering templates applied to date include MVLD, AMP/CAP/ASCO, and
customer specific 8 tier template. Compatibility with external collaborators is
ongoing; beginning with definition of tiering template for CIViC to facilitate export
through API for comparisons at scale. Conclusions: Applications of MA are broad,
and adoption has been rapid for a variety of steps in a diagnostic laboratory’s
workflow including assay panel design, reporting consistency, physician portals for
clinical decision support, human readable assertions, and rationale for assigned
tiers. Collaborating across institutions remains an essential activity in defining
consensus in accordance with guidelines.
I32. Evaluation of the Open-Source Variant Caller Platypus in the Clinical
Laboratory for Detecting Somatic Variants in Tumors
J. Reuther1, E. Fang-Tam2, V. Kumar2, H. Voicu1, A. Roy1
1Baylor College of Medicine, Houston, TX; 2Texas Children's Hospital, Houston, TX.
Introduction: Benchmarking variant caller (VC) performance for the analysis of nextgeneration DNA sequencing (DNA-seq) data in clinical pipelines requires an
assessment of computational requirements, turnaround times, as well as accuracy
and sensitivity for detecting variants of different classes. Platypus is a relatively new
open-source VC designed that uses local de novo assembly and haplotype phasing
for variant calling; here we assess its performance on unmatched tumor-only
sequencing data from clinical specimens in comparison to 2 other widely used VCs,
MuTect2 and NextGENe. Methods: Variant calling was performed on sequencing
data generated on a ~1 MB capture region covering 124 cancer genes on a set of
formalin-fixed paraffin-embedded tumors (n=11) previously characterized by wholeexome sequencing and/or targeted panels and confirmed by Sanger sequencing to
harbor 14 “true-positive” variants, including single nucleotide variants (SNVs, n=9),
multinucleotide variants (MNVs, n=1) and indels (n=4). Hybridization-capture
libraries underwent paired-end sequencing on an Illumina MiSeq (2x150 bp), yielding
~2.4 million reads per case. Sequences were aligned to the hg19 reference by
Burrows-Wheeler Aligner and NextGENe followed by variant calling using Platypus
(University of Oxford), MuTect2 (Broad Institute), and NextGENe (SoftGenetics LLC).
Results: Analysis of tumor-only sequencing data by Platypus, MuTect2 and
NextGENe generated an average of 14, 13 and 25 variants per case, respectively,
post-removal of highly prevalent single nucleotide polymorphisms and filtering for
functional coding variants with adequate coverage and quality. Platypus, as well as
the 2 other callers detected all “true-positive” SNVs (9/9; sensitivity=1.0). For indels,
Platypus correctly called all indels (4/4; sensitivity=1.0), similar to MuTect2, including
a 12bp TSC2 deletion and an 18bp TP53 deletion, the former of which was not
detected by NextGENe. For the MNV, a c.180_181delTCinsCA dinucleotide
alteration in KRAS, Platypus alone made the correct phase-aware variant call,
999
AMP Abstracts
whereas both MuTect2 and NextGENe detected it as 2 separate adjacent SNVs.
Run-time for Platypus was shortest (~ 5 minutes) followed by NextGENe and
MuTect2. Conclusions: On a representative set of clinical tumor samples, Platypus
demonstrated high sensitivity and accuracy in calling variants of 3 different classes
from tumor-only sequencing data, notably outperforming the other callers in phaseaware variant calling on the MNV. Further evaluation of Platypus on gold standard
reference cell lines and tumor specimens with low allelic fraction variants would
provide valuable insight regarding its suitability for clinical cancer genomics
pipelines.
I33. Monitoring Germline SNPs to Control for Sample Cross-Contamination in
the Ion AmpliSeq Cancer Hotspot Panel Next-Generation Sequencing Assay
P.A. Kenny
Gundersen Medical Foundation, La Crosse, WI.
Introduction: The advent of cancer gene next-generation sequencing panels is
rapidly changing practice in oncology, as patients’ treatments are increasingly
matched to specific mutations in their tumors. Typically, a sample from an individual
is tagged with a short, unique (but usually reused) DNA barcode, a pool of samples
from several individuals is sequenced, and informatics is used to deconvolute the
barcode-tagged sequences to allow assignment of sequence reads to the correct
individuals. Intra-run or inter-run cross-contamination or misidentification of DNA
samples may lead to incorrect diagnoses and ineffective patient treatments.
Methods: We analyzed data from 60,706 exomes (Exome Aggregation Consortium)
to determine the overall population allele frequency of all SNPs covered by the 207
amplicons of the Ion AmpliSeq Cancer Hotspot Panel v2 50 gene sequencing panel,
and also the SNP frequencies found in different ethnic groups (European [NonFinnish], European [Finnish], African, Latino, East Asian and South Asian). We
selected 7 germline SNPs and evaluated the ability of genotyping these loci to
discriminate between samples and identify samples of common origin. Results: We
evaluated the approach using 234 bam files from 170 unique individuals (NCBI SRA
ERP012966). Our “germline barcode” consisted of the genotype of 7 SNPs (minor
allele frequency range 7.9-47.7%) in the following genes: EGFR (rs1050171), APC
(rs41115), FLT3 (rs2491231), RET (rs1800861), PDGFRA (rs2228230) and
KIT (rs3822214). Thirty-nine of the 40 individuals represented by more than one
BAM file (range 2-4) had the same genotype in all samples, confirming the utility of
this approach for identifying samples from the same individual. The single individual
with discordant samples (2 identical and 1 different genotype) had robust coverage
at these loci, raising the likelihood of sample misidentification in this case. The 7
SNP combinations yielded a total of 90 genotypes from this 170 patient population
which was insufficient for unique sample identification, yet provides a very
substantial number of germline barcodes with which to monitor for contamination in a
low- to-medium throughput NGS laboratory, especially when also coupled with
knowledge of somatic mutations in each sample. Conclusions: Monitoring these 7
SNPs provides a straightforward method for flagging potentially cross-contaminated
or misidentified samples when analyzed using the Ion AmpliSeq Cancer Hotspot V2
panel and should be straightforward to implement using other NGS panels of a
similar size. Additionally, the analysis of germline SNPs in the regions covered by
this NGS assay in over 60,000 ethnically diverse individuals may provide guidance in
cases where it is unclear if a rare variant is germline or somatic.
I34. Breaking the NGS Noise Barrier to Accurately Detect Variants Below 1%
Allele Frequency
S.K. Sandhu, A.M. Wood, V. Kelchner, J. RoseFigura, J. Lenhart, L. Kurihara, V.
Makarov, T. Harkins
Swift Biosciences Inc., Ann Arbor, MI.
Introduction: Reliably identifying and monitoring of low frequency variants (<5%) by
next generation sequencing (NGS) is often confounded by the non-specific
background noise from amplification and sequencing errors. The ability to track
original DNA molecules through various steps of NGS library preparation using
molecular barcodes called Molecular Identifier (MIDs) prior to sequencing can help
mitigate this problem. MIDs result in elimination of amplification and sequencing
errors thereby improving sensitivity and enabling detection of variants below 1%.
This is beneficial for liquid biopsy assays which offer non-invasive monitoring of early
onset of disease and recurrence but can be challenging due to low yields,
fragmentation and variable admixture of normal DNA. Methods: Blood samples from
patients with ovarian, colon, liver, stomach, and kidney cancers were collected in
Streck Cell-Free DNA BCT vials and circulating, cell-free DNA (cfDNA) was isolated
using the Qiagen QIAamp Circulating Nucleic Acid Kit. About 20ng cfDNA was used
to make NGS libraries with Accel-NGS 2S MID Hyb kit to tag each molecule followed
by hybridization capture with targeted cancer panels. Captured libraries were
sequenced on the Illumina HiSeq to greater than 8,000x depth. Molecules containing
the same MID were grouped to generate consensus sequences, facilitating removal
of false positives due to PCR and sequencing errors and increased data retention.
Variant calling was performed using Vardict and Lofreq enabling highly sensitive and
precise detection of variants down to 0.5-1% allele frequency. Results: A total of
20ng to 500ng DNA was obtained from 10ml blood samples with a nucleosomal
peak at ~170bp and a mean Alu repeat qPCR integrity score of 0.22, characteristic
of high quality cfDNA lacking cellular DNA content. We observed a 90% library
conversion rate, and obtained high complexity libraries with uniform target coverage.
This method was used to validate variants at 0.5% to 1.0% from 10ng of cfDNA and
1000
tumor DNA increasing both the sensitivity and specificity of variant detection.
Additionally, we identified variants in cfDNA that were also present in corresponding
tumor samples. Conclusions: Due to the limited material available from liquid biopsy
asaays, the requisite PCR introduces a significant source of noise followed by errors
which occur during sequencing. Hence, it becomes even more important to use an
NGS method that enables accurate capture of the most relevant information free
from any errors to guide variant discovery and monitoring. Using MID technology in
targeted hybridization capture or amplicon enrichment permits variant detection
down to 0.5-1.0% by reducing false positives, enabling high confidence in lowfrequency variants.
I35. Genome in a Bottle: You’ve Sequenced a Genome, How Well Did You Do?
J.M. Zook1, J. McDaniel1, M. Salit2
1National Institute of Standards and Technology, Gaithersburg, MD; 2National
Institute of Standards and Technology, Palo Alto, CA.
Introduction: As genome sequencing is increasingly used clinically, wellcharacterized genomes are important to benchmark variant call accuracy. The
Genome in a Bottle Consortium (GIAB) was formed as a community effort for
authoritative characterization of benchmark genomes. Previously, we
integrated multiple technologies to form high-confidence SNP, indel, and reference
genotype calls for 78% of the pilot GIAB genome, NIST Reference Material (RM)
8398 (aka HG001/NA12878). These benchmark calls have been used in analytical
validation of clinical sequencing assays, comparisons of bioinformatics tools, and
optimization, development, and demonstration of new technologies. However,
characterization of more difficult variants and regions in HG001 and additional
genomes is needed to benchmark performance for challenging variants, which can
be clinically important. Methods: Here, we refine and simplify the methods we use to
integrate multiple sequencing technologies, with the intention of deploying a robust,
reproducible, cloud-based pipeline to analyze arbitrary human genomes. We form
high-confidence calls with respect to GRCh37 and GRCh38 for HG001 and four
additional genomes of Ashkenazi Jewish and Chinese ancestry that are now
available as NIST RMs 8391, 8392, and 8393. Results: Our new methods produce
17% more SNPs and 176% more indels than our previously published calls for
HG001. We call about 90% of the reference genome with high-confidence, increased
from 78%. Our calls only contain 108 differences from pedigree-based Platinum
Genomes calls, only 14 of which are ambiguous or likely to be errors in our calls. By
comparing example callsets to different benchmarks, we elucidate challenges in
interpreting performance metrics when benchmarking against imperfect highconfidence calls. We have worked with the Global Alliance for Genomics and Health
to develop standardized performance metrics and sophisticated tools to compare
variant calls with differing representations. These tools were used in the
PrecisionFDA variant calling challenges. They enable stratifying performance metrics
by variant type and genome context to elucidate strengths and weaknesses of a
method. We also explore differences between comparing to high-confidence calls
for the 5 GIAB genomes, and show that performance metrics are largely similar but
not identical across genomes. Conclusions: Reference samples developed by
GIAB serve as enduring scientific assets for development of genome-sequencing
measurement science, acting as benchmarks. These well-characterized genomes
and GA4GH benchmarking tools enable clinical and research labs to gain a detailed
understanding of variant calling performance. This work has enabled us to establish
best practices for benchmarking variant calls.
I36. Rapid RNASeq: Rapid and Hugely Scalable Fusion Gene Detection in the
Cloud
S. Newman, Y. Li, X. Zhou, C. McLeod, M. Rusch, J. Easton, S.V. Rice, S.A.
Shurtleff, J. Nakitandwe, E.M. Azzato, K.E. Nichols, J.R. Downing, D.W. Ellison, J.
Zhang
St. Jude Children's Research Hospital, Memphis, TN.
Introduction: For precision medicine applications, tumors are evaluated by an evergrowing battery of time-sensitive molecular and cytogenetic tests to detect gene
fusions and internal tandem duplications (ITDs). While many labs are using targeted
next generation sequencing (NGS) panels to detect these events, we considered
whole transcriptome sequencing (RNA-Seq) to be the ultimate solution, as it can
evaluate all expressed loci, including those only relevant to rare pediatric tumors.
Clinical RNA-Seq presents several unique challenges, including how to: i) avoid an
informatics bottleneck within a time-sensitive workflow; ii) minimally burden those
reviewing the data; and iii) standardize and share complex computational analysis
methods. Here we present a pilot study of clinical RNA-Seq analyzed in the cloud to
support oncology diagnostics. Methods: Using a highly secure platform, we
developed our Rapid RNA-Seq pipeline centered on an extensively validated de
novo-assembly and fusion gene detection algorithm “CICERO.” As an initial method
validation, we evaluated 78 diverse but well-characterized tumor samples and
showed that Rapid RNA-Seq detected all previously identified fusion and ITD
abnormalities. To aid in clinical review, we developed a rich graphical interface
based on our Protein Paint software (https://pecan.stjude.org/#/home) that allows
pathologists to easily assess supporting evidence and functional impact. We
performed a clinical validation of Rapid RNA-Seq, running it in parallel on
fresh/frozen samples from patients consented to G4K, a clinical feasibility study of 3platform NGS for molecular diagnostics. Gold-standard molecular and cytogenetics
testing was also performed. The Illumina TruSeq Stranded Total RNA LT Kit was
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
used to generate libraries, which were sequenced using a paired end 2 x 125 bp
cycle protocol and SBS technology on Illumina HiSeq or NextSeq instruments, with
>20% of bases at 30X depth of coverage. At least 2 reads supporting a fusion
junction were required to call a fusion. Results: Eighty-eight solid and liquid tumor
cases were included in the clinical validation; 74 had confirmed structural variants,
including 55 gene fusions, 14 enhancer proximity rearrangements and 5
ITDs. Rapid RNASeq had 97% clinical sensitivity and 100% clinical specificity, with
a 0.001% analytical sensitivity at 40% tumor burden. Run times averaged 4 hours
from raw data acquisition to reporting and were not affected by caseload. Time from
sample receipt to pathologist report was consistently under 15 days. Conclusions:
Rapid RNASeq provides a flexible, scalable and clinically valid solution to timesensitive diagnostic, prognostic and targetable fusion detection. Our tools and
workflows are securely available through www.stjude.cloud.
variants/hotspots as well as assay noise. Variants which are frequently detected
across many samples and clustered around a certain allelic fraction may be
attributed to PCR or sequencing error. Variants which have rarely been observed
previously are most likely not assay noise and are therefore potentially reportable
variants. Using the allelic fraction histogram, potential germline variants may be
distinguished from somatic variants. In addition, the same descriptive analytics
visualization may identify a reportable variant relative to a clustered distribution of
assay noise. The CNV log2ratio histogram helps identify outlier cases with potential
gene amplifications and copy losses. Conclusion: Storing structured NGS variant
data in a database combined with a user interface for retrieval and visualization
enables descriptive analytics for decision support in clinical genomics.
I37. Building the Enterprise Omics Repository for an Integrated Healthcare
System
G.B. Christensen, N.K. Jenkins, G.M. Wood, D.K. Crockett
Intermountain Healthcare, Murray, UT.
Introduction: DNA sequencing is a valuable technology for disease diagnostics,
tumor profiling, and other clinical and research applications. The final clinical
interpretation of a genomic test is typically included in a patient’s electronic medical
record (EMR), but the underlying sequence data files are rarely integrated with the
EMR. These raw data files can also be difficult to store and access. In contrast to
many traditional lab test values, the DNA sequence data may be useful throughout
the entire patient lifetime. These files may be necessary in order to ascertain data
quality, review secondary findings, and follow-up on variants of uncertain
significance using new evidence or alternative analytical pipelines. A centralized and
aggregated collection of sequence data also has tremendous research potential
when linked to the longitudinal patient record. Methods: Intermountain Healthcare
has developed the Enterprise Omics Repository as a central archive for DNA
sequence data. The repository is designed to store sequence data from both clinical
and research activities, generated by either internal or external laboratories. All data
is indexed with a patient identifier to allow linkage with other clinical data. This
electronic archive has 2 primary components: a bulk file storage component and a
metadata database. DNA sequence data (including FASTQ, BAM and VCF formats)
and related files are written to the bulk storage. The metadata database indexes
those files, capturing patient demographics, assay details, test results, file size
statistics and other pertinent data. A RESTful API controls all interaction with the
contents in a secured manner. A web browser application allows authenticated users
to search the contents of the database, upload new data, and launch analysis
workflows. Somatic and germline DNA sequencing applications are both supported.
Results: The Enterprise Omics Repository was launched in December 2016. Data
feeds have been established from the internal sequencing facility at Intermountain
Precision Genomics and from one external reference laboratory. Additional
connections are expected to be established in 2017. Conclusions: Development
work is continuing. Automated systems are being developed to populate the
repository with data from internal sources and external reference laboratories. We
anticipate the repository will expand to accommodate data from multiple “omics”
technologies, potentially including proteomics, transcriptomics and direct-toconsumer sequence data. The Enterprise Omics Repository, together with the
Electronic Health Record and Enterprise Data Warehouse, comprise a unique and
powerful informatics resource to support clinical and research genomics applications.
I39. Analysis of Therapy and Trial Recommendations Based on Gene Panel
Size
O.G. Miller, C.J. Miller, J. Coleman, M. Stachowiak, B. Parikh, M. Glynias
GenomOncology, Cleveland, OH.
Introduction: Panels of various sizes exist for NGS analysis of tumor samples.
However, not much is known regarding the panel size that represents the best
choice for specific cancer types. Therapeutic goal of NGS sequencing is to facilitate
therapy and clinical trial identification based on a patient’s specific mutation profile.
GenomOncology’s proprietary Match software enables us to generate therapy and
trial recommendations from molecular marker data. We set out to discover whether
the use of large gene panel leads to the generation of a greater number of
recommendations for certain tumor types. Methods: We analyzed therapy and trials
recommendations for the patients in the GENIE consortium. Patients were grouped
by size of NGS panels used in the study: large gene panels (275 genes or greater),
and small gene panels (74 genes or fewer). Therapy and trial recommendations
were analyzed separately. We calculated and compared the number of unique
recommendations, percent of patients receiving recommendations per tumor type,
and median number of unique recommendations per patient per tumor type.
GenomOncology's therapy database is based on FDA, NCCN, ASCO, and My
Cancer Genome, and includes database of all cancer-related clinical trials
from http://clinicaltrials.gov. Results: We did not find an observable difference
between proportion of patients receiving trial recommendations between institutions
using large and small gene panels. A significant increase in the median number of
trial recommendations was found for 22 tumor types in institutions using large gene
panels. Predictably, we found an appreciable (>25%) increase in percent of patients
receiving therapy recommendations in institutions using large panels for ovarian
cancer, esophagogastric cancer, pancreatic cancer, glioma, bladder cancer, breast
cancer and prostate cancer. A significant increase in the median number of therapy
recommendations for institutions using large gene panels was found for breast
cancer. Increase in recommendations for larger panels was driven by the presence
of BRCA1 and BRCA2, as well as gene fusions, on the large gene panels. ARID1A,
BRCA2, FLT4, ROS1 and TSC2 alterations generated the greatest number of breast
cancer trial recommendations in large panel institutions. Conclusions: The use of
large gene panels may be advantageous when gathering therapy and trial
recommendations for breast cancer patients. This occurred due to detection of gene
fusions and amplifications, as well as BRCA1, ARID1A, BRCA2, FLT4, ROS1 and
TSC2 mutations in large gene panels. Increasing panel size significantly increases
the number of trial recommendations for a number of tumor types, whereas it was
only found to significantly increase the number of therapy recommendations in
breast cancer.
I38. Descriptive Analytics Decision Support for Clinical Genomics
E. Dominguez Meneses1, S. Duraisamy1, A. Macleay1, M. Pacula1, S. Al Turki1, M.
Rivera1, V. Nardi2, D. Dias-Santagata2, J. Lennerz2, A.J. Iafrate2, L. Le1
1Massachusetts General Hospital, Charlestown, MA; 2Massachusetts General
Hospital, Boston, MA.
Introduction: Next-generation sequencing (NGS) has empowered the practice of
clinical genomics. The vast number of targets covered in NGS and the breadth of
variant calling significantly increases the complexity of the clinical sign out
process. The structured results of variant calling from bioinformatics pipelines may
be utilized to gain additional knowledge about a variant observed in one patient
relative to its detection across a testing population. We describe our software
solutions which take advantage of the structured data from variant calling to drive
descriptive analytics decision support for NGS reporting. Methods: We use the
ArcherDx anchored multiplex PCR (AMP) methodology for a 91-gene solid tumor
DNA genotyping panel and a 53-gene solid tumor RNA fusion panel. The structured
details of SNVs, indels, CNVs, and gene fusions are parsed and stored in a
MongoDB database backend. We developed a web-based user interface called
VarVetter to retrieve, display, filter, and sort all DNA or RNA variants. The interface
also allows user input to mark variants for reporting or to annotate with technical
tags. Each variant is uniquely tracked across samples for how many times it has
been detected and how many times it has been marked with various tags. Two
descriptive analytics visualizations are available in VarVetter: 1) a histogram
distribution of allelic fractions across all cases for any variant of interest and 2) a
histogram distribution of CNV log2ratio across all samples. Results: We have tested
almost 2,500 cases on the DNA panel and over 2,000 cases on the RNA fusion
panel. The tracking of each individual variant for detection prevalence and prior
reporting across our testing population identifies frequently reported
I40. Automated Cancer Risk Scoring Using FHIR Genomics Profiles and Secure
Web Services
M. Harney1, G.M. Wood1, D. Atnoor2, D.K. Crockett1
Intermountain Healthcare, Murray, UT; 2SysBioChem, Lexington, MA.
Introduction: The development of SMART (Substitutable Medical Applications,
Reusable Technologies) on FHIR (Fast Healthcare Interoperability Resources)
applications is quickly becoming the preferred platform for electronic messaging of
genomics and family history data. This study explores the feasibility of using FHIR
resources and secure web services to identify and manage women at high risk for
hereditary breast and ovarian cancer. The goal is to provide automated and real-time
risk assessment back to clinicians for timely clinical decision support. Methods: A
FHIR-enabled web service was provided through an Office of the National
Coordinator (ONC) funded pilot program. The risk algorithm (Hughes Risk App)
calculates the allele carrier probability of genetic conditions. This is done by
retrieving the fields of a FHIR message that describe the medical conditions of family
members of the patient. Test data was generated from an artificial patient application
that draws random SNOMED codes along with randomized birthdates and gender.
Both single patient messages and bundled messages (3,000 patients) were
submitted to the secure web service for load testing. The risk algorithm calculation
was also automated during the phase of receiving the FHIR message. Server
bandwidth and computational capability were also benchmarked. The initial phase of
the study required successful single patient submission. The follow-on phases of the
pilot program approach a real-time, interactive service with multiple clients. Results:
Individual patient messages formatted to a standard FHIR resource specification that
contain medical information about the patient and family member were successfully
submitted. Submission was processed via sFTP server where the web service can
retrieve and process asynchronously. Risk score calculation was also successfully
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
1001
AMP Abstracts
automated as part of the electronic submission and return. Conclusions: The
results of this study demonstrate that as the FHIR standard is used to transmit
bundles of more complex genomic information, load testing and requirements for the
optimal network design may be used to increase traffic flow. This would allow a realtime web service to process multiple FHIR messages from many clients
simultaneously and report back automated risk assessments based on the family
history information provided in each message.
I41. Engraftment Assessment by Next Generation Sequencing Using Single
Nucleotide Polymorphism (SNP) Fingerprinting
A. Mohanty, J. Gomez-Gelvez, K. Petriva-Drus, M. Zaidinski, W. Wang, J. Yao, C.
Ho, A. Zehir, M. Arcila
Memorial Sloan Kettering Cancer Center, New York, NY.
Introduction: Stem cell transplant (SCT) engraftment assessment (EA) and
molecular testing for relevant genetic alterations are integral components of the
monitoring of leukemia patients in the post-transplant (PT) setting. Testing is often
performed using separate assays which may be laborious and expensive depending
on the assays used. In this study, we describe the use of next-generation
sequencing (NGS) for concurrent mutation assessment and engraftment analysis
through the use of single nucleotide polymorphism (SNP)
fingerprinting. Methods: Fifty-five hematologic samples submitted for routine
mutation status and engraftment analyses were selected for the study. This included
blood and bone marrow samples from 20 patients, encompassing diagnostic
baseline host and unrelated (de-identified) donor, and follow up PT samples at
various monitoring time points. Somatic mutations were evaluated using an in-house
developed NGS based myeloid panel interrogating 28 genes (total of 847 amplicons
designed with the RainDance DeepSeq system). Seventy-seven highly polymorphic
SNP loci were identified in the regions interrogated by the NGS assay by using
population frequency information from 1000 Genomes. All informative SNP loci in
donor and host were used to calculate percentage of donor DNA contribution based
on the genotype counts in the PT sample. Routine EA was performed using standard
short tandem repeat (STR) analysis (Promega GammaSTR and CTTv kits).
Results: The samples were sequenced with an average coverage > 800X with
overall median coverage >1,500X. All tested post-transplant samples showed highly
correlated donor percentage values compared to the STR analysis (R2 = 0.99, p <
0.001). The number of informative loci per case ranged from 4 to 17. Two patients
assessed as 100% donor by STR analysis, showed minor host component by the
NGS assay (<5%). Low level mutations, including silent mutations previously
detected in the baseline diagnostic samples were also detected, confirming that our
assay was more sensitive. Compared to STR analysis, calculations based on
informative SNPs circumvented confounding issues related to overlapping peaks,
stutters due to polymerase slippage, and preferential allele amplification commonly
observed in clinical practice. Conclusions: SNP fingerprinting and differential
analysis constitutes a robust method for engraftment monitoring. Compared to
standard methods, NGS based assays consolidate both mutation and EA in a single
assay and circumvent confounding associated with polymerase slippage and
preferential amplification of allelic loci.
I42. ClonoTracker: A Computational Framework and Clinical Tool for NGSBased Clonality and MRD Analysis
J. Nakitandwe, M. Chen, J. Gu, P. Stow, W. Mathews, S. Shurtleff, E.M. Azzato
St. Jude Children's Research Hospital, Memphis, TN.
Introduction: Next-generation sequencing (NGS) technology offers an opportunity
for clonality and minimal residual disease (MRD) testing at a higher sensitivity,
throughput and precision, with ease of application to a wide range of patients.
However, although efforts have been made to develop bioinformatics tools for NGS
of the immune repertoire, there are limited, validated options for the clinical setting.
Available software may analyze a set of samples, but provides no mechanism for
easily following subsequent patient samples and monitoring changing clones over
time. The objectives of this study were to: 1) Develop NGS-based bioinformatics
tools for analyzing, visualizing, tracking and reporting of clonality and MRD results
and 2) perform a clinical validation of this computational framework using the
Lymphotrack assay in pediatric patients with acute lymphoblastic
leukemia. Methods: We analyzed the immunoglobulin Heavy Chain gene (IGH) and
T-cell receptor γ-chain gene (TRG) rearrangements in a cohort of 86 bone marrow
and peripheral blood specimens from pediatric patients diagnosed with B-cell (BALL) or T-cell (T-ALL) acute lymphoblastic leukemia using Invivoscribes’
Lymphotrack IGH and TRG assays and compared data with parallel results from a
clinically validated flow-cytometric analysis. DNA extracted from the samples was
used to generate libraries and sequenced on the Illumina MiSeq. To analyze the
NGS data, FASTQ data and sample information was uploaded into an in-house
developed software, ClonoTracker, a computational framework consisting of: 1) a
pipeline for NGS-based clonality reporting and 2) a web application for MRD
monitoring that extracts CDR3, identifies V, D and J gene segments, and infers
clonotypes, taking into account PCR and sequencing errors. Results: B- and T-cell
clones were detected in 87% of B-ALL and 92% of T-ALL diagnostic samples,
respectively. Using ClonoTracker, the average analytical precision (100%),
sensitivity (95%) and accuracy (95%) for the lymphotrack assay at diagnosis,
compared to flow cytometry, were determined using a threshold of 5%. For MRD,
variable thresholds of 0.0005% to 5% were used. Overall, there was good correlation
1002
between Lymphotrack and flow-MRD values, with few
outliers. Conclusions: ClonoTracker is a new computational framework for clinical
laboratory grade NGS-based clonality/MRD analysis and reporting. By detecting
clonotypes, as opposed to unique sequences which can vary in a patient sample due
to sequencing errors and mutations, allowing user-definable thresholds for clone
tracking (e.g. >5%), providing quality control metrics and collating data for clinical
reports and data mining, ClonoTracker can improve and ease pathologist NGS
clonality review and reporting.
I43. Clinical Next Generation Sequencing Leveraging Unique Molecular
Barcodes in Somatic Mutation Calling Absent a Matched-Normal
A. Bigdeli1, K. Ganapathy1, R. Sussman1, J.D. Morrisette2, R.B. Faryabi2
1Hospital of the University of Pennsylvania, Philadelphia, PA; 2University of
Pennsylvania, Philadelphia, PA.
Introduction: Achieving high sensitivity and specificity in detection of somatic
mutations in the absence of a matched-normal tumor sample is a common challenge
in clinical cancer diagnostics. To increase the power of such detection, we
developed a combined experimental-computational approach based on molecular
barcoding (MBC) technology provided by Agilent’s HaloPlex assay for use in The
Center for Personalized Diagnostics (CPD) latest 153 gene solid tumor panel.
Tagging next-generation sequencing (NGS) reads with unique MBC’s is used as a
mechanism to reduce the observation of PCR amplification errors which can lead to
false positive variant calls downstream. Post amplicon-based NGS and reference
alignment the highest quality read per amplicon is selected for further variant calling.
Using these consensus reads we developed a novel ensemble variant caller
approach capable of producing sensitive calls for single nucleotide (SNV),
insertion/deletion (Indel), and somatic copy number alterations (SCNA) while
reducing the number of erroneously called mutations typically seen in non-matched
normal informatics analysis. Methods: Data for analysis includes 45 clinical samples
run on both internal and externally validated bioinformatics pipelines. Additionally,
well characterized HAPMAP samples NA12878, NA19240 and Horizon DX HD701
were tested. Samples were assessed for sensitivity and specificity against their given
truth set using individual variant callers prior to the development of an ensemble
method. Truth sets for the 45 clinical samples were generated through manual
review to account for confounding results commonly seen when comparing unreviewed informatics output. Results: We were able to assess variant call accuracy
and understand biases introduced by each algorithm to optimize the detection of
somatic variants leading to the development of a novel ensemble method. This
method was then evaluated using HAPMAP and Horizion DX samples for correction
of biases that may be introduced through data derived from targeted panels. Using
HAPMAP samples, the method was assessed over 441,588 possible base calls and
achieved an average sensitivity of 99.34% and an average specificity of 99.99%.
Analysis of the MBC technology using Horizon DX showed a variance of 1.83% in
known versus experimental allele frequencies. Conclusions: Utilizing manually
reviewed data from allows for a more appropriate benchmarking of informatics tools
for calling somatic mutations in the absence of a match-normal allowing for the
development of a method that accurately corrects inherent biases in assay specific
technologies. This coupled with MBC technology provides a refined call set that
shows reduced amplifications errors and a higher accuracy in variant allele
frequencies.
I44. Evaluation of Copy Number Variation Detection Methods for Amplicon
Sequencing Assays
A. Bigdeli1, R. Sussman2, A. Schrank-Hacker1, V. Aikawa2, J. Rosenbaum1, J.D.
Morrissette1, R.B. Faryabi1
1University of Pennsylvania, Philadelphia, PA; 2Hospital of University of
Pennsylvania, Philadelphia, PA.
Introduction: Somatic copy number alterations (SCNA) exhibit the largest breath of
changes in cancer genomes compared to any other type of somatic alteration. These
somatic lesions play critical roles in activating oncogenes and inactivating tumor
suppressors, and have been associated with significant clinical outcomes. However,
the existing limitations of amplicon-based sequencing technologies have hindered
the ability of identifying SCNAs in clinical setting. These challenges stem from interassay variability, and are further exacerbated by the lack of complimentary normal
tissue in many clinical labs. Therefore there is an urgent and unmet need to establish
a rigorous baseline for detection of SCNAs using amplicon-based assays to achieve
clinically acceptable reproducibility. Here we performed a comparative study of four
different methods for SCNA detection using amplicon-based technologies and
validated our findings using orthogonal Fluorescence in situ Hybridization (FISH)
analysis. Methods: We performed comparative study of 4 methodologies for SCNA
detection, CNVKit, CODEX and DNAcopy, and internal normalization. Our
comparative study is based on 14 samples encompassing 8 clinically relevant
SCNAs. In addition, 21 normal samples were used to establish assay baselines in
the absence of clinically significant SCNAs. We examined the ability of each
methodology to accurately detect SCNAs, and benchmarked the results of the
comparative analysis using output of FISH experiments on the same samples with
positive and negative control probes. Results: We observed a wide range of
additional detections with each method and through orthogonal validation were able
to determine the limitations of each methodology for amplicon-based sequencing.
Methods incorporating data from a pool of normal samples for baseline
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
determination outperformed the ones that attempted to correct for variation using
internal, or flat metrics. Sensitivity measures were established based on a detection
of greater than 2 and total loss of copies. Conclusions: Our results demonstrate
that an ample pool of normal samples run across multiple sequencing experiments
are necessary to determine an accurate baseline for SCNAs detection using
amplicon-based technologies. Using these methods, we are able to accurately detect
SCNAs from measurements obtained from amplicon-based sequencing assay in a
targeted panel with as few as 153 genes.
I45. GIMP: Genomic In-silico Mutator Program for Bioinformatics Validation of
Clinical Next Generation Sequencing Assays
I. Mujacic, S. Kadri, S.A. Patil, L.L. Ritterhouse, J.P. Segal
University of Chicago, Chicago, IL.
Introduction: The last few years have seen a tremendous push to expand clinical
genetic testing for cancer patients to increasingly large sets of genes and associated
cancer-relevant loci. The reasons for this upsurge in oncology testing are manifold,
including steep reductions in sequencing cost due to advances in next generation
sequencing (NGS), increased numbers of relevant targets and marketplace
competition. However, as assay size grows it can become impossible to adequately
test mutation detection at all loci across an assay’s target region using existing
reference samples. This is an area where a bioinformatics validation can be a
critically helpful adjunct to the process, via the in silico creation of datasets harboring
specifically designed anomalies that can be passed through the assay’s analytical
pipeline. We created a Python/Pysam-based Genomic In-silico Mutator Program
(GIMP) that accepts an aligned BAM file along with a target region and will produce
paired-end fastq files containing a specified type of mutation (SNV, deletion,
insertion or duplication) within all desired targets via targeted modification of original
sample data. These fastq files may then be passed through the bioinformatics
pipeline of the assay to determine the detection efficacy of the pipeline.
Methods: With a library of processed BAM files from our CLIA validated hybrid
capture (UCM-OncoPlus) and amplicon based assays we developed GIMP to
introduce SNVs and structural variants (insertion, deletion, or duplication) into clinical
NGS data. GIMP works by accepting a BAM file and a target region bed file. The
features of the software allow the user to create mutations with varying mutant allele
frequencies (MAF) and indel lengths/types to test the pipelines variant calling
capabilities assay-wide. Assay types can be specified to accurately simulate the
effects of deletions in amplicon vs. capture assay data. Results: To assess the
performance of GIMP we conducted a series of simulation experiments with fixed
positions and at variable MAF. At the midpoint of each exon in our 1,213 gene UCMOncoPlus assay, a concordance of 100% was observed at varying MAF between 5%
and 100%. Additionally for the amplicon based OncoscreenST2.0 assay a
concordance of 100% was observed. We also saw a high level of concordance when
simulating variants at variable positions at a fixed MAF of 20%. Conclusions: Here
we present The Genomic In-silico Mutator Program or GIMP, a utility for simulating
SNV, insertions, deletions, and duplications in clinical NGS data. GIMP can serve as
a critical tool for establishing the bioinformatics performance of large-scale NGS
assays during the clinical validation.
I46. Discrepancies between the Human Reference Genome (GRCh37) and
Transcriptome (RefSeq) Complicate Variant Detection and Interpretation for
Clinical Exome and Genome Sequencing
B. Yoo, N. Miller, M. Cadieux-Dion, E. Farrow, E. Repnikova, C.J. Saunders, S.
Vemula, M.S. Farooqi
Children's Mercy Hospital, Kansas City, MO.
Introduction: The human reference genome (GRCh37) and RefSeq gene
annotation are used for clinical exome/genome sequencing (CEGS) at our institution.
Though both GRCh37 and RefSeq are considered gold-standard, primary sequence
discrepancies between the 2 exist for clinically-relevant genes. These mismatches
are rarely discussed, though they can have a significant impact on variant detection
and interpretation. Here, we systematically compare corresponding GRCh37 and
RefSeq sequences to quantify discrepant locations. We also explore whether this
issue is improved by using GRCh38 or other genome annotations, such as UCSC or
Ensembl (GENCODE). Methods: GRCh37 and GRCh38 were used for analysis;
primary chromosomes 1-22, X, and Y, were examined. Genome annotations used
were RefSeq (version 1-13-2017) for GRCh37 and annotation release 108 for
GRCh38, release 88 of Ensembl, and UCSC refGene (downloaded 5/31/2017).
Genes listed with a phenotype in the Online Mendelian Inheritance in Man (OMIM)
database were categorized as disease-associated (i.e. OMIM genes (Og)).
Results: We found that 3,083 RefSeq transcripts out of 45,006 total (6.9%) had
either a sequence or length mismatch in the coding DNA sequence (CDS) when
compared to GRCh37; 346 of these mismatched transcripts (11.2%) represented
Og. For Ensembl and GRCh37, 320 of 57,532 transcripts mismatched (TrMm)
(0.56%); 14 were Og. For UCSC and GRCh37, 3,224 of 45,036 TrMm (7.2%),
involving 353 Og. For GRCh38, the numbers were: RefSeq, 1,774 of 45,065 TrMm
(3.9%), 227 Og; Ensembl, 299 of 55,620 TrMm (0.54%), 12 Og; and UCSC, 1,944 of
45,090 TrMm (4.3%), 240 Og. Further investigation found pathogenic variants
present in the reference genome at some mismatched locations. For example,
GRCh37 contained the Factor V Leiden (FVL) variant (c.1601A); the RefSeq
transcript had the major allele (c.1601G). A preliminary query of our internal variant
database found at least 30 potentially pathogenic variants in Og present in
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
GRCh37. Conclusions: We characterized primary sequence discrepancies between
GRCh37 and transcripts in RefSeq. This issue, albeit improved, persists even when
GRCh38 or other annotations are used. In addition, the issue of mismatched
transcripts becomes irrelevant for genes where the reference genome carries a
pathogenic variant. For example, CEGS assays based on RefSeq and GRCh37
would miss individuals homozygous for the FVL variant since they are not variant
from the reference genome; non-carriers of FVL would result as having the reverse
variant, c.1601A>G (p.Gln534Arg). In such cases, having a transcript that matches
the reference genome just reinforces the incorrect variant call. We suggest
cataloging such positions and performing a second mapping directly to RefSeq to
help solve this problem.
I47. The Quality Sequencing Metric (QSM) a Concise, Transparent Notation of
NGS Data Quality for Clinical Testing
S. Yost, S. Mahamdallie, E. Ruark, M. Munz, A. Rimmer, A. Strydom, N. Rahman
Institute of Cancer Research, Sutton, London, England, United Kingdom.
Introduction: Conventional next generation sequencing (NGS) data quality
descriptions give only basic summary information e.g. average coverage across the
target. This is inadequate for clinical applications as comprehensive description of
the minimum sequence quality requirements for accurate genotyping are needed.
We developed and validated a Quality Sequencing Metric (QSM) to provide a
standardised, concise, transparent description of NGS data quality. Methods: The
QSM describes information about sequencing coverage (C), base quality (B) and
mapping quality (M) in the format QSM C_B_M, using data present in the BAM
file. C: the number of sequencing reads that map to the reference genome at any
given base position. B: Sequencing platforms automatically generate base calls from
clusters to produce raw sequencing reads. A base quality (Q) score of a base call
predicts sequencer platform base calling error. M: Sequence aligners automatically
output for each read mapping to a reference base position a mapping quality
(MAPQ) score. A MAPQ score reflects the probability that a read is mapped to the
wrong position. We used the QSM to provide standard quality thresholds for clinical
testing of cancer predipostion genes using the Illumina TruSight Cancer Panel
(TSCP). The QSM selected was QSM: C50_B10(85)_M20(95). This requires that
100% of the constituent bases of the test region meet the minimum sequencing
quality requirements of coverage of ≥50 reads with a base quality score of ≥10 in at
least 85% of reads, and mapping quality score of ≥20 in at least 95% of
reads. Results: We applied the QSM in TSCP data from 960 samples. For each
QSM parameter we determined the number of samples that failed to meet the
required thresholds and assigned regions as PASS when thresholds were met in at
least 99% of samples. 93% of test regions fulfilled QSM C50_B10(85)_M20(95). On
average per sample, 97% of test regions met the QSM. Across the 960 samples,
97% of test regions fulfilled C50, 99% of test regions fulfilled B10(85) and 96% of
test regions fulfilled M20(95) in at least 99% of samples. Conclusions: The Quality
Sequencing Metric (QSM) is a concise, standardised, comprehensive, transparent
description of NGS data quality, which is easy-to-use and easy-to-understand. The
QSM describes the minimum requirements a test region must fulfil for coverage (C),
base quality (B) and mapping quality (M). This is given per base, but can be provided
per region, per exon, or per gene, as appropriate. Fulfilment of the QSM can be
assessed from BAM files allowing integration within a NGS-based pipeline. Stating
that a test region has fulfilled the predefined QSM provides transparency that the
requirements for accurate genotyping have been met.
I48. +STAR-SEQR: Accurate Detection and Quantification of RNA Fusions
Using NGS Data
J.S. Jasper, J.G. Powers, V.J. Weigman
EA Genomics | Q Squared Solutions, Morrisville, NC.
Introduction: Genomic structural variation and associated RNA fusions are a
common clinical feature known to be involved in the initiation and pathogenesis of
cancer. This complex class of variants also has significant implications on
therapeutic decisions and has emerging roles in evidence-based clinical
applications. Despite recent advancements, there is a critical need for a tool to
accurately and precisely identify fusions from diagnostic applications in the NGS
space. Method: Here we present STAR-SEQR, an algorithm used to detect and
quantify fusions from RNA-Seq data. STAR-SEQR identifies fusions using split
junction and discordant paired reads, read assembly, and kmer approaches. Several
features are used in the probabilistic module which include the number of unique
reads, diversity of junction reads, read quality metrics, assembly metrics, and TPM
values. A network graph approach is used to mitigate false positive events arising
from misalignment and further pruned based on pairwise homology scores within
each cluster. Finally, PCR primers are produced for orthogonal validation and fusion
events are annotated and then classified into one of 5 categories. Moreover, STARSEQR is also computationally efficient (runs in short minutes), produces
standardized outputs (for interpretation of results), and saves supporting read info for
subsequent visualization. Results: For evaluation, we first utilized truth sets in the
Ge et al. (Bioinformatics, 2011) dataset of 50 synthetic fusions. STAR-SEQR had
the highest accuracy of the all tools, calling 49/50 events without false positives.
STAR-SEQR also achieved the highest F1 score in the ICGC-TCGA DREAM SMCRNA Challenge, reaching an overall average of 0.96. Analytical lab testing has also
been performed using the Seracare Fusion RNA Mix V2 sample containing 15
clinically significant fusions with known concentrations in the low range of
1003
AMP Abstracts
detectability. STAR-SEQR accurately called 13/15 events with the other events
having no identifiable read alignments.With FFPE material, we have also
benchmarked methods against varying read lengths and library methods and
demonstrated excellent tradeoff between sensitivity and precision while generally
outperforming all other widely-used tools. Notably, STAR-SEQR has been used on
several large clinical datasets and the fusion events identified have displayed strong
overlap with identified DNA structural variants. Conclusions: In summary, STARSEQR has excellent analytical performance that surpasses existing fusion detection
tools. It is broadly applicable to RNA-Seq methods, is computationally efficient, and
robust across sample characteristics allowing for use in clinical applications.
I49. Pediatric Gut Microbiome Characterization as a Companion Diagnostic in
the Clinical Evaluation of Gastrointestinal Symptoms
R. Luna1,2, M. Balderas1,2, J. Runge1, E. Welebob1, A. Venkatachalam1, M.
Dahdouli1, E. Hollister1, J. Versalovic1
1Baylor College of Medicine, Houston, TX; 2Texas Children's Hospital, Houston, TX.
Introduction: The critical role of the microbiome in human health and disease has
expanded the diagnostic potential of microbiome characterization. This emerging
clinical application is currently intended as a companion diagnostic to routine
laboratory testing and the complete clinical picture. With utility ranging from a single
time point evaluation as part of a differential diagnosis to longitudinal monitoring of a
chronic disease, here we present the initial validation of gut microbiome
characterization in pediatric patients with gastrointestinal (GI) symptoms.
Methods: Pediatric gastroenterologists identified patients with unresolved GI
symptoms. Stool specimens were collected at home or within the hospital or clinic
setting. Bacterial DNA was extracted from stool aliquots homogenized in sterile
saline. Next-generation sequencing targeting the V4 region of the 16S rRNA gene
was performed on the Illumina MiSeq. With a minimum of 15,000 reads per sample,
sequences were quality filtered, clustered into operational taxonomic units, and
taxonomic assignments were made. Healthy control populations were amassed from
publicly available datasets as well as healthy children identified in our clinical
research cohorts, allowing for age and sex-matched controls for each patient.
Smaller internal pediatric cohorts of GI-based conditions (ex. irritable bowel
syndrome) were also utilized for comparison. Results: Comparisons of single
pediatric gut microbiomes to subsets of healthy children yielded clinically useful
information. In a child with no organic cause of GI symptoms and the absence of any
identified pathogen via routine methods, a significant deviation (i.e., complete
absence of Bacteroides spp.) from the typical microbiome in age and sex-matched
healthy controls was observed, suggesting that a GI-related issue persisted in this
child. Longitudinal monitoring of individual samples pre- and post-fecal microbiota
transplantation (FMT) for the treatment of recurrent C. difficile infection also provided
insight into shifts in microbial composition associated with resolution of symptoms
related to FMT. Additionally, suspected GI pain in a child with autism spectrum
disorder was found to correlate with clear shifts in the microbiome, along with the
identification of a specific organism of interest that increased in abundance during
episodes of perceived pain. Conclusions: Pediatric gut microbiome characterization
has demonstrated clinical utility ranging from evaluation of difficult to treat cases to
monitoring during treatment. With mounting evidence regarding the ability to predict
treatment success based on pre-intervention profiles as well as predict prognosis
based on longitudinal changes, this strategy will continue to evolve in the clinical
realm.
I50. Identification of Distinctive Cell Signaling Patterns in Renal Cell Carcinoma
Gene Expression TCGA Data Sets
K. Volyanskyy1, M. Zhong2, J.T. Fallon2, E. Vail2, R. Lucito3, N. Dimitrova1
1Philips, Valhalla, NY; 2Westchester Medical Center, Valhalla, NY; 3Hofstra
University, Hempstead, NY.
Introduction: Cell signaling profile characterization may provide significant clinical
impact in revealing driving events in complex diseases such as cancer. Due to
inherent heterogeneity and highly complex interactions of biological pathways this is
by far a non-trivial task. Using novel approaches in supervised and un-supervised
machine learning, we demonstrate the ability to identify and quantitatively describe
cell signaling patterns from genomic data. Methods: We developed a computational
framework based on a combination of unsupervised gene selection and supervised
predictive modeling of histological and molecular subtypes. This framework is
multilevel, involves many steps of data transformations on the observed distribution
patterns with fuzzy clustering and Reactome pathway assessment. We applied our
method on Level III RNASeq gene expression data from 20,531 genes in 889 tumor
samples of renal cell carcinoma (RCC) across 3 subtypes – clear cell renal cell
carcinoma (ccRCC), papillary renal cell carcinoma (pRCC), chromophobe carcinoma
(chRCC), and 129 normal samples from The Cancer Genome Atlas (TCGA).
Results: Each histological subtype is uniquely characterized by the top predictive
gene sets identified in our framework. In particular, in chRCC NOTCH was identified
as the most altered signaling pathway when compared to normal samples (P value <
0.01) characterized by significant down-regulation of DTX1 and CNTN1 in tumor. In
prRCC compared to normal, the most distinct signaling pathways involve insulin
receptor recycling, prostanoid ligand receptors (PTGER3), G alpha (q) signaling
events (GNA14, CASR, PTGER1, GCGR, HCRTR2, and KNG1), GPCR ligand
binding, and FGFR3b ligand binding and activation (FGF9) (all P value < 0.01). In
ccRCC compared to normal, characteristic pathways involve response to elevated
1004
platelet cytosolic Ca2+ (EGF, HRG, SERPINA5), and tight junction interactions
(CLDN8, CLDN19, CLDN16) (P value < 0.01). When all 3 subtypes compared with
each other the most distinctive cell signaling pathway is VEGF and associated
neurophilin interactions, specifically,with distinct levels of FLT1, KDR, VEGFA
expression (P value < 0.01). Conclusions: Using our method we identified genes
and associated cell signaling pathways which may potentially be used in targeted
therapy for renal cell carcinoma. This will require additional testing and validation
using new datasets as well as other genomic data modalities.
I51. Proficiency Testing for Next-Generation Sequencing: Multi-Institutional insilico FASTQ File Exchange Ensures Robust and Reproducible Bioinformatics
Workflows for Reporting Complex Mutations
T. Schneider1, O. Rouhi1, K. Davies2, M. Gruidl3, J. Woolworth-Hirschhorn4, W.
Glen4, A. Kuraishy5, D. Agius6, E. Allen5, C. Coldren7, T. Hong8, N. Patel9, G.
Smith1, C. Hill1, D. Qin3, D. Aisner2, M. Rossi1
1Emory University, Atlanta, GA; 2University of Colorado, Boulder, CO; 3Moffitt Cancer
Center, Tampa, FL; 4Medical University of South Carolina, Charleston, SC; 5Illumina,
San Diego, CA; 6Illumina, Seattle, WA; 7Pathgroup, Nashville, TN; 8Hackensack
University Medical Center, Hackensack, NJ; 9University of North Carolina, Chapel
Hill, NC.
Introduction: Proficiency testing (PT) is essential to assess the accuracy and
reproducibility of clinical reporting. In 2015, a consortium of 6 academic clinical
laboratories was formed to test their respective bioinformatics workflows. Each lab,
which performs clinical laboratory developed tests for solid tumors using the same
commercial library preparation kit, exchanged 24 FASTQ datasets. In that study, the
discordance rate for calling insertion/deletion (indel) mutations was as high as 60%.
We re-challenged bioinformatics PT last year, focusing on indels, and demonstrated
improvement in bioinformatic workflows. This year we implemented generation of insilico FASTQ files to provide challenging mutations from the COSMIC database as
means of improving FASTQ PT and providing an accessible archive.
Methods: Using NIH’s ART read simulator tool a set of 5 FASTQ files each
containing 3 complex in-dels previously reported in the COSMIC database were
designed to simulate an Illumina TruSight Tumor 26 run. Two duplications near the
limit of single read detection for this assay (36 and 42 base pairs) were also
generated. Data analysis was performed independently by participating sites and
clinically significant mutations were reported. The results were reported back to each
site, and an assessment of the performance of the bioinformatics workflows were
discussed. Results: Preliminary analysis of our series of 5 FASTQ datasets showed
while there is significant improvement, issues with complex indels still persists.
Concordance was not achieved among all institutions and the large duplications
eluded detection by the majority of institutions. Conclusions: Despite disparate
clinical bioinformatics workflows, PT is required to achieve consistency in detecting
mutations that push the limits of available technologies. An in-silico FASTQ analysis
provides a resource to laboratories to fully explore the range of detection for their
respective workflows. These files are freely available through the Illumina
BaseSpace and we hope to expand this archive to include a broad range of
sequence complexity, mutational heterogeneity and variant allele frequencies for a
variety of commercially available library preparation methods used a laboratory
developed tests.
I52. Analysis of Individual Genes Identifies the Impact of Physiological
Functions on AlloMap Gene Expression Profiling
R.N. Woodward, M.J. Nelles, J.P. Yee, M. Machrus, L. Wilson, S. Scott, J.J. Sninsky
CareDx, Brisbane, CA.
Introduction: Gene expression profiling (GEP) scores, such as the AlloMap test for
heart transplant rejection, are enhanced by an understanding of the contribution of
the individual genes. The components of the test are reflective of various
physiological pathways and a study of their relative contribution improves the
biological understanding of the pathology. This analysis compares the relative
contributions of individual genes and metagenes to AlloMap scores for commercially
provided results from 2005 to 2015. Methods: The results for each AlloMap score
are computed from the expression levels of 11 genes, which were grouped by
functional contribution into 7 groups (4 individual genes and 3 metagene
groups). The results from 85,244 AlloMap scores from 16,785 patients were
compared to the expression levels of the 7 functional groups. For each
gene/metagene a linear trend model was computed given the GEP raw score and
using Oracle Business Intelligence and Tableau, both designed to handle large
datasets. The inputs for the genes/metagenes were weighted using the originally
published algorithm. The relative contribution of genes/metagenes to clinically
impactful diagnostic scores were determined. Results: Genes/metagenes with
relationship to steroid responsiveness, T cell activation/migration, and inflammation
had the most significant contribution to the magnitude of the combined GEP
score. FLT3/IL1R2/ITGAM, R2=0.361; MARCH8/WDR40A, R2=0.250; PD-1,
R2=0.232; ITGA4, R2=0.221. Platelet genes had the least impact on the score
(R2=0.039 for PF4/G6b1). Although the platelet gene group had a low correlation
with the overall GEP score, there were individual patient GEP scores for which these
genes played a large role in the overall score. The results for other genes were
intermediate in contribution and included: SEMA7A, R2=0.151 and RHOU,
R2=0.040. Conclusions: The most significant contributors to the magnitude of
AlloMap scores are genes associated with steroid responsiveness, T-cell activation
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
and migration, and inflammation. These functions are consistent with the biology of
transplant rejection. Despite less impact over the full set of patient results, each of
the other genes did have a significant impact on a subset of the AlloMap
scores. This study highlights the value of analyzing large real-world datasets to
confirm and extend the findings from prior biomarker development and validation
studies. Further analysis is needed to determine if these subsets correlate with
clinical outcomes.
maintain high performance below 0.1%. This represents a significant performance
improvement with respect to published molecular barcoding strategies and can be
deployed using existing molecular workflows.
I53. Genotype Matching of Serially Collected Clinical Samples Using Next
Generation Sequencing Can Identify Sample Handling Errors
M. Grskovic, J. Collins, Y.S. Wang, D. Ross, K. Thompson, C. Marchis, J.
Beausang, R.N. Woodward
CareDx, Brisbane, CA.
Introduction: Clinical studies including large numbers of samples processed at
multiple sites carry a risk of sample handling and labeling errors. The errors may
occur at the initial sample labeling and processing at the draw site; over the transfer,
labeling and handling in the processing lab; and during data processing and
analysis. Studies that include serial sampling of genetic material provide an
opportunity to determine the correct patient attribution using data generated from
multiple specimens from the same patient. Next generation sequencing (NGS)
technologies can identify differences in genetic sequence as part of the
measurement. We developed a tool to QC patient-sample assignment by matching
genotypes of samples within studies of donor-derived cell-free DNA in
transplantation. Methods: NGS data generated to quantify donor-derived cell-free
DNA was adapted for use as a genotype of the recipient. The recipient genotype of
each sample was correlated to the recipient genotype of all other samples within 3
different clinical studies in heart and kidney transplantation. The genotype matching
tool was developed in Python to create visual and numeric outputs for quick
assessment of specimen assignment to the correct patient. Results: Specimens
from multicenter studies in heart transplant (n=274 from 68 patients) and kidney
transplant (n=1272 from 384 patients) and a single-center study in kidney transplant
(n=186 from 36 patients) were assessed using the genotype matching tool. Twentytwo, 0, and 2 mismatches were identified in each study, respectively, where data
labeled from a specimen assigned to a given patient did not match the other
specimens from that patient. Several of these mismatches could be ascribed to
specific step(s) in the sample handling process, including one that was determined to
have occurred at the time of blood collection, one at the time of sample processing at
draw site, and another during a sub-step of the library preparation
process. Conclusions: Identification of specimen collection and handling errors can
identify the contribution to the overall variance in the study and reduce the impact of
mis-labeled samples. In some cases, identification of the steps where the error
occurred can lead to process improvements to avoid future errors. The genotype
matching may be of great value in clinical testing laboratories where serial testing is
performed.
OTH01. Optimizing Somatic Genomic Reporting and Physician Interpretation
with Web-Based, Interactive Reports
S.W. Gray1, A.M. Cronin2, J. Gagan3, L.M. Sholl3, E. Cerami2, H. Uno2, N. Oliver2, C.
Lowenstein2, R. Lederman2, A. Revette2, A. Suarez4, C. Lee5, J. Bryan6, E.M. Van
Allen2
1City of Hope, Duarte, CA; 2Dana-Farber Cancer Institute, Boston, MA; 3Brigham and
Women's Hospital, Boston, MA; 4Massachusetts Institute of Technology, Cambridge,
MA; 5Harvard Medical School, Boston, MA; 6Broad Institute, Cambridge, MA.
Introduction: Molecular pathology has seen an increased availability and decreased
cost of tumor genomic profiling, but underutilization and misinterpretation of complex
genomic data at the point of care remains a major limitation toward effective clinical
implementation. Current reporting mechanisms of genomic data are typically static
documents that lack context to aid interpretation. We hypothesized that enhanced
interactive genomic reporting mechanisms may improve physician satisfaction and
interpretation of individualized molecular profiling data. Methods: We created webbased, interactive reports with enhanced data visualization elements and embedded
decision support for OncoPanel, a 300+ gene panel. We conducted a randomized
vignette-based survey study to determine whether exposure to the interactive
reports, as compared to static reports, improves physicians’ genomic understanding
and report-based satisfaction. Overall comprehension and satisfaction scores were
calculated across 3 vignettes (possible range 0-18 and 1-4 respectively, higher score
correspond to improved endpoints). Results: One hundred 5 physicians at
DFCI participated (29% participation rate); 67% medical, 20% pediatric, 7% radiation
and 7% surgical oncology; 37% female. Prior to viewing the case-based vignette
reports, 34% of physicians reported that they found it difficult to make treatment
recommendations based on the standard report in their routine practice. After
viewing the case-based vignettes, physicians’ overall comprehension scores did not
differ significantly by report type (mean score interactive 11.6 vs. static 10.5,
difference=1.1, 95% CI -0.3, 2.5, p=0.13). However, physicians who viewed the
interactive report were more likely to correctly assess sequencing quality (p<0.001)
and understand when reports needed to be interpreted with caution (e.g., if low
tumor purity, p=0.02). Overall satisfaction scores were significantly higher in the
interactive group than the static group (mean score 2.5 versus 2.1, difference=0.4,
95% CI 0.2, 0.7, p=0.001). Of the 92 physicians who endorsed the need for
additional genomic support for providers, 66% reported that interactive genomic
reports would be helpful. Conclusions: Continuing the growth of molecular
pathology will require support from clinical colleagues, so it is important to maximize
their ability to utilize molecular data. We show interactive, genomic reports may
improve physicians’ ability to accurately assess genomic data and increase
physician satisfaction. To advance the field, further research in representative
provider populations is warranted and efforts to integrate interactive genomic reports
into electronic health records are needed.
I54. Using Replication to Break the NGS Noise Floor for Liquid Biopsy Variant
Detection
C. Ionescu-Zanetti, N. Kamps-Hughes
Fluxion Biosciences, South San Francisco, CA.
Introduction: The ability to accurately identify ultralow frequency mutations has
been problematic due to errors introduced by both library preparation and
sequencing chemistries. As liquid biopsy samples harbor somatic mutations in the
0.01-1% allele frequency (AF) range, they require specialized molecular protocols
and algorithms designed to reduce error rates. Molecular barcoding has shown
promise in lowering the variant limit of detection. However, practical attempts to
apply this technology to clinically relevant sequence spaces have retained loss of
accuracy in the 0.1-1% range. Methods combining molecular barcoding with
background-aware callers that provide good sensitivity but high specificity is
sacrificed and false positive counts are kept low by confining variant tests to a small
number of possible variants. Methods: We validated ERASE-Seq (Elimination of
Recurrent Artifacts and Stochastic Errors), a method for the detection of ultralow
frequency variants. The basis of our method is elimination of 2 categories of false
positive variant calls: Recurrent artifacts occur at error-prone loci that are
predisposed to base misincorporation either during library prep or sequencing. Wildtype control DNA technical replicates are used to quantify the error background for
each variant across the multiplexed amplicon panel. Stochastic errors also occur
during the library preparation and sequencing processes, especially below 0.5%
allele frequency. These occur randomly throughout the sequencing
space. Stochastic errors can only be eliminated using technical replication,
decreasing false positive rates with increasing replicate number. We employ a
statistical test comparing replicate count values in sample and reference replicates.
Results: ERASE-Seq was tested using spiked DNA mixtures with clinically realistic
input amounts to detect alterations between 0.05% and 1%. Alterations were
detected with greater than 90% sensitivity and a false positive rate below 0.3 calls
per 10,000 possible alterations. We analyzed the composition of false positive calls
using standard methodologies in terms of recurrent versus stochastic errors and
show the ability of ERASE-Seq to eliminate both types. Conclusions: We
demonstrate the accurate detection of low frequency SNVs and indels on 3 different
highly multiplexed oncology amplicon panels: 56G, Trusight Tumor 15, and Spotlight
59. Our data show perfect sensitivity and specificity to 0.3% allele frequency and
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
Other (Education, etc.)
OTH02. Good or Bad Sequencing Data? Setting a Benchmark for the Quality of
Diagnostic NGS in the Lab
W. Gutowska-Ding1, J. Ahn2, K. Brugger3, J. Coxhead4, K. Thomson5, C.
Boustred6, S. Abbs7, E. Souche8, P. Westwood9, B. Ylstra10, S. Stegmann11, G.
Taylor12, F. Khawaja13, Z. Deans13, S. Patton14
1Central Manchester University Hospitals NHS Foundation Trust, Greater
Manchester, United Kingdom, 2Guy's and St. Thomas' NHS Foundation Trust
London, United Kingdom; 3University of Cambridge, Cambridge, United
Kingdom; 4Institute of Genetic Medicine, Newcastle University, Newcastle, United
Kingdom; 5Oxford University Hospitals NHS Foundation Trust The Churchill, Oxford,
United Kingdom; 6Hospital for Children NHS Foundation Trust, London, United
Kingdom; 7Cambridge University Hospitals NHS Foundation Trust, Cambridge,
United Kingdom; 8UZ Leuven, Leuven, Belgium; 9NHS Greater Glasgow & Clyde,
Glasgow, United Kingdom; 10VU University Medical Center (VUmc), Cancer Center
Amsterdam, The Netherlands; 11Maastricht University, Maastricht, The Netherlands;
12Viapath Genetics Labs, London, United Kingdom; 13UK NEQAS, Edinburgh, United
Kingdom; 14Central Manchester University Hospitals NHS Foundation Trust,
Manchester, United Kingdom.
Introduction: Next Generation Sequencing (NGS) is increasingly being introduced
into clinical genetics laboratories worldwide. The huge amount of data generated by
NGS cannot be duplicated by alternative methods for laboratories to internally
validate all results, therefore external assessment of data quality and consistency is
required. The European Molecular Genetics Quality Network (EMQN) and the UK
National External Quality Assessment Service (UKNEQAS) for Molecular Genetics
have developed a joint external quality assessment (EQA) scheme for NGS, with the
aims to: a) assess and improve quality; b) enable laboratories to benchmark their
NGS service against others and against best practice; c) work towards consistency
of reporting clinical results generated by NGS; and d) contribute towards defining
best practice. Methods: EMQN and UKNEQAS offer numerous disease-specific
1005
AMP Abstracts
EQA schemes, and the objectives for developing an NGS EQA were to make it
generic (independent of genes, diseases, and platforms used), applicable to as
many users as possible and to provide them with as a broad range of quality
indicators on their NGS data. So far, 4 pilot schemes have been run. Laboratories
were sent a genomic DNA or FFPE sample and asked to sequence it using their
usual approach, submit technical details, and genotypes of known variants. The
sequence of the DNA provided was validated in 3 clinical laboratories using different
NGS platforms. The two latest EQA runs (2015, 2016) were divided into Germline
and Somatic NGS, in order to address the different challenges of NGS in these
settings (e.g. depth of coverage). Results: The number of participant laboratories
has grown from 30 in 2013 to 303 in 2016. With this number of laboratories, we have
established a reliable way of developing a consensus genome based on the variants
identified by the participants, reducing the need to rely on time-consuming and
expensive prior validation. This has required the development of sophisticated tools
for the integration and benchmarking of NGS data. These NGS germline and
somatic EQA schemes are technology and testing context independent. They
seamlessly fit with clinical laboratories’ wet lab and bioinformatics
processes. Conclusions: Disease-specific EQA has improved the quality of results
and consistency in diagnostic test reports. The results obtained in our NGS EQA
schemes enables clinical diagnostic labs to start to address the quality of their NGS
testing. These technology-specific NGS EQAs will play an important role in enabling
labs to benchmark this new technology, assess the accuracy of data and facilitate
high quality reporting for patient benefit.
were then biotinylated with linker lengths ranging in size from 13.5 Å to 56 Å to
determine optimal biotin distance from the polymer backbone for detection by the
streptavidin-HRP conjugate. Next, improved PED was applied to microtiter plate- and
thin film biosensor chip-based detection of Candida glabrata and Candida
krusei grown in BD Bactec blood culture bottles. In addition, improved PED was
applied to after-market modification of BioLegend Legend Max and BD OptEIA
ELISA kits for human EPO and IFN-γ. Results: Improved PED demonstrated
statistically significant assay performance with a 2,000 kDa dextran polymer
conjugated with ~1,100 biotins and possessing a linker length of 22.4 Å. A polymer
with these features enhances assay sensitivity with Candida glabrata and krusei
detection at levels as low as 105 to 220 CFU in blood culture using a microtiter platebased assay and levels as low as 20 CFU in a thin film biosensor chip-based assay.
The same polymer configuration demonstrates 100 to 400 fold increased sensitivity
of EPO and IFN-γ detection. Conclusions: The present work further optimizes the
polymer chemistry of our previously published simple, inexpensive and rapid signal
amplification method based on a polybiotinylated dextran polymer. Along with the
ability to enhance staphylococcal genome detection and work in both microtiter
plate- and chip-based assays, improved PED now has the sensitivity relevant for
detection of Candida 28 S rRNA in blood culture and demonstrated ability to
enhance protein detection. Niche applications for this technology may include
speciation for which rRNA along with protein targets are needed for species unable
to be differentiated by MALDI-TOF, in addition to low income settings where
amplification and mass spectrometry are limited.
OTH03. Liquid Biopsy Based Monitoring of PD-L1 Expression in Non-Small
Cell Lung Cancer (NSCLC) Patients for Immunotherapy
G. Singh1, M. Leong1, S. Sylvester1, Y. Lee1, C. Johnson2, A. Wu1
1Clearbridge Biomedics Pte Ltd, Singapore; 2Clearbridge Biomedics, Inc., Carlsbad,
CA.
Introduction: Development of therapeutics targeting PD-1 and PD-L1 has put
forward immunotherapy as a promising cancer treatment. Immunophenotyping of the
tumor is crucial for the selection of patients whom may benefit clinically from
treatment. However, the dynamic nature of PD-L1 as a biomarker for patient
stratification and considerations such as, safety concerns in re-biopsy limits the
practicality of evaluating PD-L1 expression on tissue, and necessitate a “minimally
invasive” alternative. Liquid biopsy approach, with circulating tumor cells (CTC) has
the potential to circumvent these limitations. Here, we showcase a minimally invasive
approach using commercially available CTC enrichment platform- ClearCell FX and
BOND RX autostainer to effectuate an automated immunocytochemistry workflow for
the evaluation of PD-L1 status on liquid biopsy samples. Methods: Peripheral blood
samples and pleural effusion from metastatic NSCLC patients were analyzed in this
study. Patient samples were first enriched for tumor cells on the ClearCell FX
system. Enriched samples were concentrated and immobilized on glass slide, and
subjected to multi-color PD-L1 immunofluorescence staining implemented on BOND
RX system. Control samples were included in each staining cycle. Empirical
threshold of the assay was determined using positive (MDA-MB-231 cells) and
negative controls (white blood cells from healthy individuals), with at least 75 cells
scored in each control sample. Results: Automated workflow on BOND RX yield
good staining quality, with high signal to noise ratio and minimal cell losses (<5%) on
control samples. Variable PD-L1 expression level was observed among tumor cells
in clinical samples. Among the two pleural effusion samples under consideration,
tumor cells expressed strong level of epithelial markers. However, PD-L1 positivity
were variable, and detected on 7.6% and 71.8 % of tumor cells respectively. Isolated
CTCs expressed variable level of both epithelial markers (EpCAM/Cytokeratin) and
PD-L1. PD-L1 positivity were 40% (2/5 CTCs) and 100% (1/1 CTCs) among the two
peripheral blood samples analyzed. Interestingly, a CTC “doublet” isolated from one
of the peripheral blood samples had a cell with strong membranous staining and
other with weak diffused signal, reiterating the heterogeneity of PD-L1 expression
among tumor cells. Conclusions: We present here an automated workflow, which
facilitates the detection of PD-L1 expression on liquid biopsy specimens. Results
suggest that the workflow is robust and reproducible for the detection of PD-L1
positivity in enriched tumor cells population in blood and pleural fluid obtained from
NSCLC patients.
OTH05. A NGS Library Preparation Training Module Facilitating Rapid
Orientation and Productivity of New Employees in a Clinical NGS Core
Laboratory
S. Henke, C.K. Keso, L.M. Baudhuin, M.J. Ferber
Mayo Clinic, Rochester, MN.
Introduction: The fast-paced workflow in a clinical Next Gen Sequencing (NGS)
laboratory is not conducive to the intensive training required to introduce new
employees to complex NGS workflows. The short turnaround time (TAT)
requirements for NGS testing and defined endpoints for each shift preclude the
flexibility to fully explain processes to a trainee. In addition, errors made by a trainee
performing the process on clinical samples would require repeat testing and
significantly increase TAT or worse, create a risk for patient care. Therefore, an
alternate training method was developed to streamline the training time and avert the
risk of trainee mistakes causing sample failures with clinical samples. Methods: Our
training program comprises 3 major components: general lab/safety training, team
shadowing, and hands-on training. Trainees shadow an experienced employee
through the library preparation process. A dedicated trainer is present to describe
each step and answer questions without interfering with the production technologist.
Next, a recently developed pre-training quiz is used to document current knowledge.
With this baseline background in place, required equipment training occurs. The
principal component of the new module follows: trainee and trainer work together
through the entire process, simulating a clinical run. The pace is controlled by
comfort level of the trainee, and the process is repeated as many times as required
for the trainee to demonstrate mastery by generating libraries that meet clinical
quality control (QC) requirements and by scoring 100% on the post-training
assessment. The training is completed and initial competency documented by
performance of an observed clinical run, with results passing the QC review process.
Retrospective surveys about training experiences were completed by 5 employees
12 to 18 months after their hire date. Two of these employees had completed this
training module. Results: The 2 employees who completed this training module did
not require any additional time before demonstrating competency to perform library
preparation independently. Similarly, all 5 employees demonstrated competency in
performing additional processes independently within 6 months of this training. The
confidence in initial training providing a sufficient foundation was higher for the 2
employees who completed this training module. Conclusions: This independent,
one-on-one, library preparation training module helps new employees in the
laboratory master the process more thoroughly than simply observing clinical
production bench work. In addition, the risk to patient samples from exposure to
handling by a trainee under observation is eliminated. We plan to extend this model
to more methods in the laboratory.
OTH04. Improved Polymer Enhanced Detection of Nucleic and Amino Acid
Targets
J. Klonoski1, I. Gorrell-Brown2, J. Jensen2, W. Ao2, R. Jenison2
1University of Utah, Salt Lake City, UT; 2Great Basin Scientific, Salt Lake City, UT.
Introduction: Molecular diagnostic platforms capable of both protein and nucleic
acid target detection are currently available, however, there remains a need for
simple and inexpensive assays with improved sensitivity. Previous work by our group
demonstrated that polymeric enzyme enhanced detection (PED) was capable of
rapid detection of nucleic acid targets at low femtomolar levels. Herein, we describe
recent efforts to optimize dextran polymer size and biotin linker length followed by
parallel development of improved PED for clinically relevant Candida speciation and
after-market modification of commercially available ELISA kits for protein targets.
Methods: Dextran polymers ranging in size from 70 to 2,000 kDa with corresponding
biotinylation of 3 to 2,800 biotins were selected for initial studies on Corning DNABIND plates using synthetic oligonucleotide targets to interrogate the relationship
between polymer size/biotinylation and assay sensitivity. Lead polymer candidates
1006
OTH06. Interactive Online Lymphoma Unknown Conference: An Instructive
Platform for Ordering Flow Cytometry and Molecular Studies
S.E. Harley, B. Baskovich
University of South Alabama Medical Center, Mobile, AL.
Introduction: Traditional anatomic pathology unknown slide conferences are carried
out either with microscopic glass slides, slide images, or virtual slides, which are
available several days in advance, allowing the resident to formulate a differential
diagnosis based on the history and histology. The cases and their respective
ancillary studies, such as immunostains and molecular studies, which are not
provided beforehand, are discussed during the conference to reach a diagnosis. In
this model residents do not learn how to select appropriate additional studies on their
own or to carry the case from differential to final diagnosis. We utilized a previously
developed interactive web system to allow the ordering of flow cytometry,
immunostains, and molecular and/or cytogenetic studies in the context of
lymphomas. Methods: Using HTML, PHP and Javascript the online unknown slide
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
conference system allows the resident to view H&E images of a case and order and
view ancillary studies pertinent to each case. The entry form is AJAX-enabled for
autocompletion from all possible stains and other studies recognized by the system,
including FISH probe sets. For the case-appropriate FISH study, an image of FISH
from an actual case is shown. If an irrelevant probe set is chosen, an image of
normal FISH for that set is shown. Flow cytometry options include a basic panel as
well as B- and T-cell focused panels. Karyotypes, and occasionally other studies
such as T-cell receptor gene rearrangement studies, are also included as available.
Two function keywords are included: “cheat,” which guides the resident on what
relevant studies to order for each case, and “presentation,” which displays the final
diagnosis and key pathologic information including the molecular aspects of the
disease. Results: The system was presented to residents during a neoplastic
hematopathology conference at our department. Ten cases were available for
review, including diffuse large B cell lymphoma, Burkitt lymphoma, anaplastic large
cell lymphoma, and double-hit lymphoma. The system was highly received by
residents as it allowed them not only to develop an initial differential diagnosis, but
also to take the case to completion by ordering pertinent ancillary studies and
receiving immediate results. Sample cases are available:
http://drdoubleb.com/lymphomaunknowns/. Conclusions: The interactive online
unknown conference system with special studies was well-received and should help
to build habits for lifelong learning and improve diagnostic skills. We plan on
expanding the platform with more cases. Widespread use of this system could make
for an effective learning tool and could be applied for other hematolymphoid
malignancies as well as additional solid tumors.
OTH07. Long QT Syndrome: Integrating Genetic Testing into a Diagnostic
Work Flow: A Process to Identify Opportunities and Gaps
E.R. Lockhart1, J.E. Frank2, C. Greene2, V.Y. See2, S. Vashist2, I.M. Lubin1
1Centers for Disease Control and Prevention, Atlanta, GA; 2University of Maryland
School of Medicine, Baltimore, MD.
Introduction: There has been limited attention in the literature with regard to how a
genetic test is integrated into a broader diagnostic workflow. Analysis of the
diagnostic workflow provides the opportunity to assess concordance with
professional guidance, setting-specific practices, and availability and use of
specialized expertise. Such an analysis can identify practice gaps that can inform
interventions to optimize the use of genetic tests. To test this approach, the
diagnosis of Long QT Syndrome (LQTS) was used as a model in the context of a
hypothetical patient entering the University of Maryland Medical System. Methods:
Centers for Disease Control and Prevention in collaboration with the University of
Maryland school of Medicine developed a diagnostic algorithm that describes how a
teen with loss of consciousness during an athletic activity with a suggestive family
history for a heritable cardiovascular condition would be managed. The scenario
covers events that occur in the emergency room, the cardiologist’s office, and the
clinical laboratory. The workflow is described and evaluated in terms of the
diagnostic/treatment goals for each patient encounter, use of professional guidance,
familial issues, use and timing of genetic testing, and how laboratory consultation is
accessed to assist in test selection and understanding the test results. Results:
While family history has a role in guiding diagnosis within the emergency room,
genetic testing is not used. The decision for when to offer genetic testing in the
cardiology office setting is predicated on the likelihood for informing a diagnosis,
identifying familial risk, and treatment choice. The availability of different tests
relevant to LQTS makes test selection an integral component of the diagnostic
workup. Interpretation of less established variants that are detected typically
includes consultation with laboratory, genetic, and cardiovascular professionals to
understand the clinical implications and inform patient management decisions.
Conclusions: Guidance for a more structured approach for collecting family history
in the emergency department may be advantageous. In the cardiologist’s office,
consultation with laboratory expertise relevant to test selection and interpretation of
detected variants can be important in making a diagnosis. Such consultations should
be institutionally supported and compensated as a component of a diagnostic
management team as advocated within the 2015 Institute of Medicine
report, Improving Diagnosis in Healthcare. Development and visualization of the
diagnostic algorithm has utility in identifying practice gaps and is proposed to be
useful for cost analyses and modeling to improve laboratory and clinical processes.
OTH08. The Northern New England Genomics Consortium
N. Sidiropoulos1, M. Babcock2, F.B. de Abreu3, S. Dufresne2, K. Hampel1, K.
Lebel4, E.Y. Loo3, J.D. Peterson3, C. Scott1, D. Seward1, M. Skacel2, H.B.
Steinmetz3, L.J. Tafe3, G.J. Tsongalis3, J.A. Lefferts3
1University of Vermont Medical Center, Burlington, VT; 2Dahl-Chase Diagnostic
Services, Bangor, ME; 3Dartmouth-Hitchcock Medical Center, Lebanon,
NH; 4Baystate Health System, Springfield, MA.
Introduction: The rapid evolution and growth of clinical genomic services has
resulted in part due to next generation sequencing (NGS). This technology is being
used clinically to interrogate targeted panels of genes, exomes, and genomes. The
technical complexity is rivaled by that of informatics and an expanding medical
knowledge base that must be navigated in a way that permits scalability for clinical
laboratory practice. Additional challenges include the current reimbursement
landscape and intelligent design of health services research to investigate the value
of genomically informed healthcare. Implementation and maintenance of NGS-
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
based services is enhanced by collaborative approaches. We have created a
regional consortium of clinical molecular diagnostic laboratories to address the
myriad of challenges and implementation considerations associated with NGS in a
clinical setting. Methods: In 2016, a meeting of laboratory personnel from 4
institutions was held to discuss the practices and needs of molecular diagnostic
laboratories practicing in relatively rural settings. Topics included: sharing samples
for validation and proficiency testing, sharing experiences with different platforms
and assays, sharing experiences with reimbursement challenges and solutions, and
participating in collaborative research projects. It was agreed that mandating
changes in clinical testing practices at each institution, exchange of clinical testing
services and referral testing would be beyond the scope of this consortium.
Results: Clinical molecular laboratories from the Baystate Health System
(Springfield, MA), Dahl-Chase Diagnostic Services (Bangor, ME), the DartmouthHitchcock Medical Center (Lebanon, NH) and the University of Vermont Medical
Center (Burlington, VT) created the Northern New England Genomics Consortium.
Within the first year, two face to face meetings and sample exchanges to expedite
NGS assay validations occurred. In addition, a collaborative project to assess NGS
assay and platform reproducibility was completed and published.
Conclusions: Creation of a non-competitive collaborative regional consortium has
the advantages of expediting the implementation of new initiatives through sharing of
experiences, samples and new ideas. The collaborative nature of this consortium
promotes successes of each participating laboratory which would have otherwise
been beyond the scope of each laboratory individually.
OTH09. Time-Resource Analysis for Right-Sizing an NGS Laboratory:
Exercising Restraint, Building Responsibly
J.R. Milano1, D. Lieberman1, J.J. Morrissette2
1Penn Medicine, Philadelphia, PA; 2University of Pennsylvania, Philadelphia, PA
Introduction: Establishing a next-generation sequencing (NGS) lab requires
significant capital and operational resources. With an evolving and volatile
reimbursement environment, it becomes necessary to limit expenses to ensure labs
can subsist on their operations as opposed to the vagaries of philanthropy. Penn’s
Center for Personalized Diagnostics has been able to meet the evolving needs of its
clinicians while remaining financially viable through disciplined expenditures. A
critical component of this financial restraint is devising a method to determine
adequate staffing and prioritize deficiencies to achieve a clinically acceptable TAT
while minimizing personnel overutilization. Methods: A time-resource study was
performed to determine the number of man-hours required to perform testing for
anticipated volume, assay schedule, time commitments, and non-productive time.
Roles for analysis included total staffing (exempting faculty), total technical staffing,
wet bench, bioinformatics, and variant analysis. Non-clinical personnel were
excluded. In order to determine which role was most closely associated with TAT
variations, TAT was plotted against tests-per-FTE, a productivity measure. Results:
Time analysis revealed that an anticipated volume of 3,749 clinical tests over 5
clinical assays with extractions would require 5.73 FTEs of wet lab personnel staffed
at a desired schedule and at current automation capability (0.73 FTE from current
budget). Variant analysis staffed to allow timely review of data within two days of
availability requires 2.47 FTEs (0.97 FTE from current budget). Analysis of FTE
productivity against TAT showed the following correlations: Total staffing (R=0.7815),
variant analysis (R=0.7544, the specific role with closest correlation), total technical
staffing (R=.6798), bioinformatics (R=0.6181), and wet lab technicians with
supervisor (R=0.4423). Based on this analysis, the laboratory will be adding 2 FTEs,
prioritizing one variant analyst to mitigate rising and variable TAT and one wet lab
technician to meet desired assay schedule. Conclusions: While the results of
analysis are specific to our lab, the method is applicable for other labs seeking to
model projections for appropriate resource allocation. Our approach suggests that
starting with a desired state while considering realistic constraints and prioritizing
personnel allocation that is expected to have the biggest impact on TAT yields viable
results. Of note, this analysis is focused on clinical needs; R&D and administrative
personnel will impact the test menu, operations, and margins. We anticipate that our
proposed approach and description of our experience will be useful to other
institutions encountering challenges with cost mitigation in an evolving environment.
OTH10. Characterization of BCR-ABL Laboratory Ordering for Quality
Improvement
W. Zheng, M.D. Lucas, Y. Zhou, R.A. Allen
University of Oklahoma Health Sciences Center, Oklahoma City, OK.
Introduction: The detection and monitoring of BCR-ABL can be performed by
different assays, leading to confusion amongst ordering clinicians. Prior to March
2017, there were 5 tests available at our institution: a multiplexed PCR-based
Hemavision, BCR-ABL qualitative assay (part of the Hemavision), and as send-out
tests, BCR-ABL qualitative with reflex to quantitation, BCR-ABL major quantitative,
and BCR-ABL minor quantitative. To improve patient care, we sought to
characterize BCR-ABL orders. Methods: We retrospectively reviewed all BCRABL1+ cases in our laboratory from January 2011 - February 2017. Patients’
electronic medical records were reviewed for diagnoses and related laboratory tests.
Unnecessary orders were classified as duplicate (e.g., same orders), incorrect
(e.g., BCR-ABL major quantitative on patients with the minor transcript), or
redundant (e.g., in-house and send-out qualitative assays ordered, orders on
multiple specimen types). Results: From 64 patients, we found 57 errors (30
1007
AMP Abstracts
incorrect, 15 redundant, 12 duplicates). Patients with BCR-ABL minor had B-ALL
(n=16) and AML (n=1) and accounted for 30 errors. The inappropriate ordering of the
major quantitative assay for these patients (n=11) could be misleading for disease
monitoring. In two instances, both the major and minor quantitation were ordered
upfront. Duplicate orders included concurrent ordering of the full Hemavision panel
and the BCR-ABL qualitative assay (n=4). All 30 patients who tested positive for the
major fusion by our BCR-ABL qualitative assay had CML. They received the major
quantitative assay prematurely (n=10), minor quantitative incorrectly (n=3), and
redundant orders (n=11). Duplicate orders include the Hemavision and qualitative
tested together (n=2) or the same test ordered twice (n=4); these were caught by the
laboratory, with the extraneous orders cancelled. Our 17 patients identified with the
full Hemavision panel more frequently presented with acute leukemia (5 with B-ALL,
2 with CML blast crisis) and appeared to have fewer errors (n=7), which could reflect
their poor outcome and short follow-up period. Conclusions: Our results likely
reflect the clinical urgency during acute leukemia presentations and ordering
clinicians’ limited understanding of available molecular tests. While redundant and
duplicate orders waste medical resources, incorrect orders, such as a BCRABL major quantitation for a BCR-ABL minor patient, could be dangerous. As
molecular pathology continues to advance, it becomes increasingly necessary to
assist our clinical colleagues in the selection of appropriate laboratory orders. Our
molecular pathology laboratory has since validated BCR-ABL major quantitation,
which is allowing us to provide better guidance to our clinical colleagues.
Solid Tumors
ST01. Improvement in Diagnostic Laboratory Performance by Participation in
External Quality Assessment for Molecular Pathology: Lessons Learned and
the Need for Continued Quality Improvement
M.H. Cheetham1, D.E. Barton2, N.L. Wolstenholme1, S.J. Patton1
1European Molecular Genetics Quality Network, Manchester, United Kingdom; 2Our
Lady’s Hospital for Sick Children, Dublin, Ireland.
Introduction: The promise of precision medicine (PM) for all has been a major driver
in the rapid evolution of the field of molecular pathology. This need has been met by
laboratories from different diagnostic environments, predominantly in the fields of
pathology and genetics. Labs have been working to set up testing for tumour
markers and commercial manufacturers have responded by developing new
diagnostic kits and end-to-end diagnostic solutions to meet this need. However,
accuracy of genotyping and clinical interpretation thereof is essential for the
realization of the promise of PM. External Quality Assessment (EQA) schemes are
needed to ensure that the quality of testing delivers the right result, for the right
patient, at the right time. Methods: The European Molecular Genetics Quality
Network (EMQN) is an organization promoting quality in molecular testing worldwide
by providing EQA schemes to its 1800 members. These include EQA schemes for
molecular pathology, monogenic disorders and technical approaches to diagnostic
testing. The main objective is to establish inter-laboratory consistency. The process
establishes, harmonises and standardises best practice in correctly identifying
sequence variants, interpretation of the genotype and clerical accuracy. EMQN sent
identical clinically relevant samples to each EQA participant which they tested using
their routine methodologies. The anonymised results were peer reviewed and a
summary report was made available to all participants in order to enable
comparisons between laboratories and assess individual laboratory performance.
Longitudinal analyses of participant performance (genotyping and interpretation)
have been performed for the Lung, Colorectal Cancer and Melanoma EQA schemes.
Results: The 2016 schemes have recently concluded and our presentation will
summarise the data from 6 years of EQA showing that the rate of serious diagnostic
error remains stubbornly high with a mean of 3.65%. Errors are made by labs using
a broad range of methodologies. Analysis of the error patterns indicates that poor
validation of new tests contributes significantly, especially when implementing an
NGS strategy, or using a “white box” commercial diagnostic solution. Over the 6
years, there has been significant improvement in the clinical reporting of results
with far less 'over interpretation' of the genotyping results and improved practice with
respect to the nomenclature used for the reporting of sequence variants.
Improvement in the quality of testing is evident but there remains more to do.
Conclusions: We conclude that annual participation in EQA can improve the quality
of diagnostic testing in molecular pathology and significantly contribute to the
achievement of the promise of precision medicine.
ST02. APC and KRAS Genetic Variants Associated with Colorectal Cancer
Histology Grade and Tumor Staging
W. Zhang1, K. Gettler2, A. Waluszko1, T. Sidorenko1, G. Katava1, D. Zhang1, F. Ye1
1Mount Sinai Health System, New York, NY; 2Yale University School of Medicine,
New Haven, CT.
Introduction: Colorectal cancer (CRC) is the fourth leading cause of cancer deaths
worldwide. Next generation sequencing (NGS) has been widely used for detecting
the somatic mutation in clinical CRC management. This study is to characterize the
association of CRC-specific somatic mutation with tumor grade and staging.
Methods: We reviewed all CRC cases that have next-generation sequencing results
at Mount Sinai Health System. To evaluate the staging status, total 111 cases with
1008
surgical resection specimen and adequate clinical information are included in this
study. The demographic information, specimen, tumor site, tumor grade, TNM and
AJCC staging are obtained from the Pathology report and medical record. NGS was
done on The Ion Personal Genome Machine (PGM) sequencer by using Ion
Torrent AmpliSeq Cancer Hotspot Panel v2. The mean depth across various
amplicons is 2,000×, and 200× nucleic acid coverage and 5% of somatic gene
mutation are used as the cutoff to make the final variant call. All somatic mutations
with a frequency higher than 2% were included in the final analysis. For statistical
analysis, we initially run the logistic regression on the full dataset and then use
stepwise regression to predict the best model. Results: The age of these 111 cases
is ranging from 36 to 90 years old with a mean age at 62.5 years. Total 144 different
mutations are detected in 30 genes, which include APC, BRAF, CTNNB1, FBXW7,
IDH1, KRAS, PIK3CA, SMAD4, STK11, SMARCB1, TP53, and etc. A stepwise
analysis reveals that the 2 variants, APC_R1450* and KRAS_G12C, appeared to be
significantly associated with tumor histological grade (p<0.05); the variants,
APC_R1450*, and TP53_R248W appeared to be significantly associated with Tumor
T stage (p<0.05). Stepwise regression analysis results are shown in Table 1.
Conclusions: APC, KRAS, and TP53 are well-known important genes in the CRC
etiology pathway. This study confirmed that certain mutations in these genes are
associated with tumor histology grade and staging. Sequencing a larger CRC cohort
may aid us to develop a risk model that can predict tumor histology grade and
staging.
ST03. WITHDRAWN
ST04. Development and Validation of ColoScape – A New Colorectal Cancer
Mutation Detection Assay
M.J. Powell1, E. Peletskaya1, Q. Sun1, L. Pastor1, A. Zhang1, K. Koprowska2, S.
Bodmer2
1DiaCarta, Inc., Richmond, CA; 2John Radcliffe Hospital, Oxford, England, United
Kingdom.
Introduction: Colorectal cancer is a highly preventable disease as early detection
increases rates of patient survival to near 100%. Herein we report the development
and validation of ColoScape, a highly sensitive test powered by XNA technology to
detect novel multigene mutation biomarkers in CRC by real-time PCR. The assay
allows the sensitive detection of the presence or absence of mutations in the
targeted regions of the genes interrogated in tissue biopsy (FFPE) and plasma
samples. Methods: The high sensitivity of this multigene biomarker assay is
achieved due to xenonucleic acid (XNA) probe technology. XNA probes are novel
backbone modified oligomers with natural nucleoside bases (A, T, C and G) that
hybridize by Watson-Crick base pairing to natural DNA and RNA with much higher
binding affinity. XNA probes are designed that bind to the selected wild-type
sequences at the respective genetic loci in the target genes. These XNA probes
cannot be extended by DNA polymerase thus suppress amplification of WT DNA
templates and only allow amplification of the target mutant DNA templates in the
sample. For each of the selected mutation sites, primers and FAM-labeled TaqMan
probes were designed and tested together with the selected XNA oligomers. An
internal PCR control selected in the Human -Actin (ACTB) gene was employed
utilizing an HEX-labeled TaqMan probe. Results: The ColoScape kit was
demonstrated to have robust analytical performance and clinical accuracy for FFPE,
stool and plasma samples. The rapid, precise and sensitive molecular assay for
mutation detection in colon cancer has key benefits listed below. The assays showed
a sensitivity of as low as 0.1% mutation in 5-10 ng of WT DNA/well. No crossreactivity was observed with wild-type up to 320ng purified gDNA and up to 20ng
FFPE DNA per reaction demonstrating high specificity of the ColoScape assay.
Intra-assay, inter-assay, lot-to-lot and operator variation comparison showed CV%
between 3% and 8%. Excluding pre-cancer samples, the assay clinical specificity
and sensitivity were 95% and 100%, respectively. Pre-cancer detection sensitivity
was 60% and 62.5% for stool samples. For tested FFPE clinical samples, the assay
specificity and sensitivity were 95% and 91% respectively while the assay clinical
specificity and sensitivity were both 100% for plasma samples.
Conclusion: The ColoScape Colorectal Cancer Mutation Detection qPCR assay is
shown to be a sensitive tool intended to aide in the early detection of colorectal
cancer, disease monitoring and therapeutic interventions.
ST05. Detection of Microsatellite Instability in Circulating Cell-Free DNA of
Patients with Colorectal Carcinoma
J. Pettersson, M. Campan, L. Du, T. Long, A. Barzi, L. Dubeau, P. Ward
University of Southern California, Los Angeles, CA.
Introduction: Checkpoint inhibitors targeting programmed cell death pathways and
immunotherapy are especially active against colorectal cancers with microsatellite
instability. We sought to test the hypothesis that tracking biomarkers of such
instability in circulating cell-free DNA is feasible and might be an effective means of
monitoring treatment responsiveness in these patients. Methods: Circulating cellfree DNA was isolated from blood samples collected in Streck DNA tubes from 3
patients with stage IV colorectal carcinoma associated with high microsatellite
instability and 3 patients with tumors showing no evidence of such instability
(controls). Cell-free DNA isolation and testing for microsatellite instability were
performed using kits obtained from Promega Corporation. Microsatellite alterations
were scored by comparing the electrophoretic profile of each biomarker to that
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
present in genomic DNA extracted from buffy coats from the same
patients. Results: One patient experiencing a deep response to immunotherapy
showed no evidence of microsatellite instability in his circulating cell-free DNA. The 2
other patients showed high microsatellite instability in their liquid biopsies. Two liquid
biopsies were obtained 2 weeks apart from one of those patients while showing a
clinical response to immunotherapy. There was a 3-fold decrease in the relative
amounts of abnormal versus normal microsatellite alleles between the first and
second biopsies from this patient. The electrophoretic profiles of the microsatellite
alleles in both liquid biopsies were identical to those previously seen in this patient’s
primary tumor. The 3 patients with no evidence of microsatellite instability in their
colorectal tumors likewise showed no such instability in their circulating cell-free
DNA. Conclusions: Circulating cell-free DNA is a source of cancer-derived
genomic DNA in patients undergoing treatment that can be readily obtained using
minimally invasive procedures. Clinical responsiveness to therapy coincided with
either decreased or absence of abnormal microsatellite alleles in liquid biopsy
samples from 2 patients showing clinical responsiveness to immunotherapy. Our
results clearly demonstrate the feasibility of detecting microsatellite instability in
circulating cell-free DNA from patients with metastatic colorectal cancer and raise the
possibility of using this approach to monitor treatment response and disease
progression in patients with carcinomas harboring such instability.
ESR1 and RAF1. 40ng of gDNA was used as template. In addition, 2 cell lines NA12878 and HD200 (Horizon) with known mutational status were analyzed.
Secondary data analysis was performed using the QIAGEN Clinical Insight Analyze
(QCIA) software and the QIAGEN Clinical Insight Interpret (QCII) software. Results:
Forty-six of 50 (92%) specimens had sufficient DNA quantity and quality for analysis.
Thirty-six of 50 (72%) specimens showed mutations in the BRAF, EGFR, KRAS or
NRAS gene using Pyrosequencing or Real-Time PCR. All mutations were positively
identified by the GeneReader. The GeneReader additionally detected PIK3CA and
ERBB2 gene mutations in 4 specimens which were later confirmed by
Pyrosequencing. Ten of 20 (50%) of the liquid biopsies showed pathogenic
mutations. In 7 cases with matched primary tumor samples, 3 (42.9%) harbored
EGFR, KRAS or NRAS mutations not detected in the primary tumor. Replicate
analysis (n=20) of cell line mutations showed a 100% agreement in all
cases. Conclusions: We observed a 100% agreement between the GeneReader
NGS system and previous Pyrosequencing and RT-PCR detection methods. The
GeneReader NGS system provides a solution with low hands-on time, high
performance and accuracy in the detection of relevant mutations. The QCIA and
QCII software provide an easy tool for data analysis, variant interpretation and
quality control. In summary, the GeneReader NGS system is easy to implement in
daily routine and provides an accessible technology for molecular laboratories.
ST06. Lung and Colon Adenocarcinoma Mutational Landscape in a Tertiary
Academic Healthcare Center
O.C. Rafael-Rosca, V.S. Williamson, A. Kusmirek, V.S. Borodin, L. GonzalezMalerva, A. Popa, C. Yang, F. Sábato, C. Vlangos, A. Ferreira-Gonzalez
Virginia Commonwealth University Health System, Richmond VA.
Introduction: Next Generation Sequencing (NGS) has started to dominate the
clinical practice for precision medicine, driving therapeutical decisions and advancing
the knowledge of genetic alterations landscape. This allows a fast, informed disease
management, especially in solid tumors, known for their complexity and
heterogeneity. Understanding mutations pathways and coexisting genetic alterations
advances optimal patient care. We report on NGS results on lung adenocarcinoma
(LADC) and colorectal adenocarcinoma (CRC) in our tertiary health center
population. Methods: A 1.5-year retrospective inquiry of our NGS database
identified 354 malignancies with 51 CRC and 188 LADC unequivocal histology. Of
these, 16 CRC and 32 LADC had at least 2 genomic alterations. Patients’
demographics, smoking behavior and tumor mutational status, with emphasis on ≥2
genetic alterations/case were analyzed using Pearson’s
correlation. Results: Average age was 61-year in the CRC and 66 in the LADC
selected patients. 37.5% of the CRC were African-Americans (AA) and 62.5%
Caucasians (EA). 41% AA and 53% EA had LADC (6% unknown).APC was the most
commonly mutated gene in CRC (81%), followed by KRAS (50%), PIK3CA (44%),
and TP53 (44%). Four CRC patients had 2 variants for each of the TP53, PIK3CA,
and KRAS genes. The genomic profile for LADC was predominated
by TP53 (72%), KRAS (50%), and EGFR (25%) gene alterations. Two different
variants for EGFR and TP53 were present in 2 LADC patients. Ten genes in CRC
and 13 in LADC showed various alterations. 56% of CRC had 2 concomitant
mutations, 31% had 3, 13% had 4. Similarly, 2 coexisting variants were identified in
84.4% and 3 in 15.6% of LADC. Analysis revealed significant interactions of CRC
mutations: APC and TP53 in female (F) smokers, APC and PIK3CA mutations in
EA males (M) and F smokers, andPIK3CA and TP53 in
F. PIK3CA and TP53 simultaneous alterations were observed in the general
population (p=0.0449; r2=-0.51), as well as APC and PIK3CA (p=0.0017; r2=0.72), APC and TP53 (p=0.028; r2=0.55), APC and FBXW7 (p=0.977; r2=0.43). In
LADC, a significant interaction was seen for PTPN11 and TP53 variants (p=0.0099;
r2=0.45), PTEN and TP53 (p=0.0285; r2=-0.39), EGFR and KRAS (p=0.0153;
r2=0.43), and CDKN2A and KRAS (p=0.0739; r2=-0.32). PTPN11 and TP53 coalterations were observed in LADC of M AA smokers, PTEN and TP53 variants were
specific to F smokers, EGFR and KRAS to F AA, while CDKN2A and KRAS gene
alterations were more prevalent in EA smokers. Conclusions: Our study indicates
that coexisting variants might be population specific and the mutational profile can be
shaped by individual smoking status. This might have clinical relevance in the future
precision medicine landscape for the effectiveness of CRC and LADC therapies.
ST08. Detection of MLH1 Promoter Methylation by MassARRAY MALDI-TOF
A.A. Hall, J.A. Raney, R.L. Margraf, J. Nelson, K.M. Gligorich, A.P. Matynia
ARUP Institute for Clinical and Experimental Pathology, Salt Lake City, UT.
Introduction: Lynch syndrome, also known as hereditary nonpolyposis colorectal
cancer syndrome (HPNCC), is an inherited condition that predisposes individuals to
development of colorectal, endometrial, and other types of cancer. It is characterized
by germline mutations in DNA mismatch repair (MMR) genes such as MLH1. Tumors
arising in this setting show microsatellite instability and loss of MMR expression by
immunohistochemistry. Since somatic events can also lead to abnormal MMR
expression, the ability to distinguish between Lynch syndrome-associated and
sporadic tumors becomes clinically relevant, as follow-up differs. MLH1 promoter
hypermethylation is a common cause of abnormal MMR expression in microsatellite
unstable sporadic tumors, and is often evaluated using methylation-specific qPCR of
bisulfite-treated DNA. We examined a high-throughput alternative to qPCR using the
MassARRAY platform (Agena Biosystems) as part of workflow/instrumentation
consolidation efforts for the clinical laboratory. Methods: DNA extracted from
formalin-fixed paraffin embedded tumor tissue was treated with sodium bisulfite,
which converts unmethylated cytosines to thymines, and purified using a
commercially available kit from Zymo. Bisulfite converted DNA was then PCR
amplified with a forward primer specific to the MLH1 promoter region and a T7promoter tagged reverse primer. These primers amplify both methylated and nonmethylated samples. The amplicon was treated with shrimp alkaline phosphatase
followed by in vitro transcription and enzymatic cleavage at uracil residues, which
results in fragments that differ in size and molecular mass based on differences in
the methylation status of 5 CpG islands in the MLH1 promoter. The fragmented
amplicon was analyzed by MALDI-TOF mass spectrometry on the MassARRAY
instrument. Subsequent analysis of mass-to-charge ratio and percent methylation
was performed using Agena EpiTYPER software. Results: A titration series was
prepared using human methylated and non-methylated DNA standards. Methylation
percentages of 100%, 50%, 10%, 5%, 1%, 0.5%, and 0% were tested, and a limit of
detection of 5% methylation was determined. Additionally, 41 clinical samples that
had been evaluated by a validated methylation-specific qPCR MLH1 promoter
methylation assay were run on the MassARRAY platform. All 29/29 methylated
samples and 12/12 non-methylated samples were concordant with qPCR results
using a cutoff of 10% methylation averaged over 5 CpGs tested by the
assay. Conclusions: The MassARRAY platform appears to be a viable, highthroughput alternative to methylation-specific qPCR assay for detection
of MLH1 promoter methylation. Additional clinical samples will be processed to
assess reproducibility and confirm percent methylation cutoff.
ST07. A Verification Study of the GeneReader NGS System in a Routine
Laboratory Setting
A. Boesl, M. Pfisterer, A. Neugebauer, C. Nemes, J. Schneider, C. Steger, P.
Cerkl, A. de Vries, F. Offner
Academic Teaching Hospital, Feldkirch, Austria.
Introduction: The goal of this verification study was to evaluate the performance of
QIAGEN’s GeneReader NGS System together with the Actionable Insight Tumor
Panel (AIT) for the detection of targetable mutations in various tumor mass and liquid
biopsy samples. Methods: Fifty FFPE tumor samples (colorectal cancer; n=29, lung
cancer; n= 10; melanoma; n=11) and 20 liquid biopsies (CRC; n=15, lung cancer;
n=5) were characterized for mutations in the BRAF, EGFR, KRAS and NRAS genes
using Pyrosequencing or Real-Time PCR (RT-PCR) in 2016. DNA extraction from
FFPE material was performed using the QIACube and circulating tumour cell DNA
using the QIASymphony SP, respectively. All samples were then re-analyzed in
2017 with the GeneReader NGS system and the AIT panel covering the following
genes: KRAS, NRAS, KIT, BRAF, PDGFα, ALK, EGFR, ERBB2, PIK3CA, ERBB3,
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
ST09. Molecular and Clinicopathologic Features Associated with PD-L1
Expression in Lung Adenocarcinoma
S. Yang1 , D. Steiner2 , G.J. Berry1 , J.W. Neal1 , J.L. Zehnder1 , C.A. Kunder1
1Stanford University, Stanford, CA; 2Stanford University, Palo Alto, CA.
Introduction: Studies have shown that PD-L1 expression on tumor cells and tumor
mutational burden may predict response to immune checkpoint therapy. However,
the interaction between these immune biomarkers and their association with
conventional pathologic and genotypic variables are not well characterized. Herein,
we explored potential clinical and molecular correlates of PD-L1 expression and
tumor mutational burden in patients with lung adenocarcinoma. Methods: Targeted
next-generation sequencing using a clinically validated 130-gene panel and PD-L1
immunohistochemistry with the 22C3 pharmDx antibody were concurrently
performed on 112 lung adenocarcinoma samples. Predicted total mutation load
(PTML) for each sample was calculated using a previously described algorithm that
allows estimation of total exonic mutational burden from targeted clinical sequencing
panels. Results: Fifty-three samples (47%) showed positive PD-L1 expression as
defined by ≥1% Tumor Proportion Score (TPS). The frequency of PD-L1 positivity
was higher in tumors from metastatic locations versus intrapulmonary parenchyma
(60% versus 39%, P = 0.033, Fisher’s exact test) and those with TP53 mutations
1009
AMP Abstracts
versus wild-type TP53 genotypes (60% versus 38%, P = 0.022). Conversely, the rate
of PD-L1 expression was lower in tumors with mutations in STK11 (19% versus
52%, P = 0.015) and KEAP1 (8% versus 52%, P = 0.0047). There was no
association with age, gender, smoking history, clinical stage, EGFR status
or KRAS status. In addition, 56 samples (50%) showed high PTML as defined by
predicted mutational burden of >100 mutations. PTML was not associated with any
clinical or genotypic features except for smoking history (60% versus 39%, P =
0.037). Likewise, the median PTML of smokers was significantly greater than that of
never-smokers (169 versus 63.5, P = 0.0015, Mann-Whitney U test). There was no
correlation between PD-L1 TPS and PTML (ρ = -0.035, P = 0.71, Spearman’s rankorder correlation). Conclusions: Alterations in STK11 and KEAP1 were associated
with lack of PD-L1 expression whereas metastatic progression and TP53 inactivation
were associated with increased PD-L1 expression. In addition, smoking history was
correlated with higher mutational burden as estimated from our targeted sequencing
analysis. No concordance between tumor mutation load and PD-L1 expression was
identified in this study. Overall, our findings suggest that recurrent mutations in tumor
suppressors may be involved in modulating the tumor immune microenvironment.
Using the mutational status of these genes as predictive biomarkers for immune
checkpoint therapy is an intriguing possibility that warrants further study.
ST10. Evaluation of NGS Based Methods to Detect the Recurrent Gene
Arrangements in Lung Cancer
A. Tilak1, M. Schomaker1, J. Daniel2, B. Billstein2, R. Kincaid1, L. Nguyen1, B.
Thyagarajan1, A. Nelson1
1University of Minnesota Medical Center, Minneapolis, MN; 2University of Minnesota,
Minneapolis, MN.
Introduction: Lung cancer is the leading cause of cancer-related death globally and
non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung
cancer cases. Genomic studies have identified a number of mutations and gene
rearrangements with potential therapeutic significance in NSCLC patients, which
demands multiplexed testing strategies on routine clinical samples. Currently, most
clinical laboratories use FISH as the primary method of rearrangement detection, but
analysis of multiple individual probes is inefficient and is often limited by sparse
tissue biopsies. Therefore, NGS-based solutions for fusion detection performed on a
single nucleic acid extraction are ideal for translation into clinical practice for lung
cancer testing. Method: Two multiplexed NGS based panels (FusionPlex CTL from
Archer and QuantideX RNA Lung cancer from Asuragen) for the detection of ALK,
RET or ROS1 gene arrangements were evaluated in this study. Total nucleic acid
(TNA) was extracted using the Qiagen FFPE extraction kit from 5 unstained sections
from 03 known ALK/ROS1 positive and 02 negative patient samples (orthogonally
confirmed by FISH) along with commercial positive and negative fusion controls
(Horizon Discovery). NGS libraries were prepared per manufacturer’s protocols and
sequenced on the MiSeq platform (Illumina). Data analysis was performed with
manufacturers’ specific software package. Results: All patient samples and
commercial controls positive for gene fusions passed the bioinformatics QC filter.
Both panels successfully identified EML4-ALK rearrangement in all known positive
patient samples. The Archer panel specifically detected an NPM1-ALK fusion in a
known positive lymphoma patient sample; although the Asuragen panel does not
specifically target this fusion an ALK rearrangement was detected by 3’/5’ imbalance.
A Horizon positive control (carrying fusions of EML4-ALK, CCDC6-RET, and
SLC34A2-ROS1) was used to perform serial dilution studies with negative control
RNA at 1:5, 1:10, 1:20, 1:25 and 1:125 to determine the detection limits of both
panels. Results showed that all 03 rearrangements in the commercial control were
detected by the Asuragen panel at 1:20 dilution while none of these rearrangements
were detected at 1:20 dilution using the Archer panel. However, the Archer panel
was able to detect all 3 fusions only at 1:5 dilution. No false positive gene fusions
were identified in the known negative patient samples or commercial negative
controls by either pipeline. Conclusions: Both panels showed equivalent
performance in a small number of patient samples. Mixing studies showed that
Asuragen panel is more sensitive for the detection of the ALK, or ROS1 or RET gene
arrangements.
ST11. Assessment of UltraSEEK Colon Cancer Panel for Detection of Low
Frequency Somatic Mutations in Blood
R. Avula1, W. Highsmith1, M.C. Liu1, J.S. Voss1, D. Irwin2, R.T. Birse3, B.R. Kipp1
1Mayo Clinic, Rochester, MN; 2Agena Bioscience, Inc., Queensland,
Australia; 3Agena Bioscience, Inc., San Diego, CA.
Introduction: Current guidelines suggest that all patients with metastatic colorectal
cancer (mCRC) have tumor tested for KRAS, NRAS and BRAF mutations to predict
for response to EGFR inhibitors. Testing for these alterations has traditionally been
performed on tissue biopsies which are invasive and sometimes not feasible. Bloodbased analyses (“liquid biopsy”) may overcome these limitations by evaluating
circulating cell-free DNA (cfDNA) from peripheral blood. This study evaluates the
performance characteristics of the Agena Bioscience UltraSEEK chemistry on the
MassARRAY System to detect low frequency mutations in cfDNA. Method: The
UltraSEEK Colon Panel assesses 107 mutations in 5 genes (KRAS, NRAS, BRAF,
PIK3CA and EGFR). The method uses as little as 10ng of total DNA and involves a
multiplex PCR followed by a mutation-specific single base extension reaction. The
extension reaction utilizes multiple, mutation-specific chain terminators labeled with a
moiety for solid phase capture. After capture, the extension products (analyte) are
desalted, transferred to an Agena Bioscience SpectroCHIP Array and loaded into the
1010
MassARRAY System. Data acquired by the MassARRAY System is processed by
the MassARRAY Typer software and reports are generated. Results: We analyzed
55 samples including DNA controls (Horizon Discovery), FFPE tumor DNA, cfDNA
samples, and cell line DNA. Twenty Horizon Discovery DNA samples evaluated gave
accurate mutation results. Mutations detected in 10 FFPE tumor DNA samples
matched results from previously run NGS data. A total of 18 blood samples were
assessed; cfDNA from 10 mCRC patients gave concordant results with matched
tumor tissue DNA analyzed by Digital Droplet PCR or NGS, and 8 samples from
normal controls showed no mutations. A subset of cell line, FFPE and cfDNA from
mCRC patients were also run in triplicates with concordant results. DNA samples
representative of KRAS G12R, G13D and BRAF V600E spiked into wild type DNA
were detectable down to 0.5% mutant frequency. Horizon Multiplex cfDNA reference
standard DNA containing KRAS G12D, NRAS Q61K, NRAS A59T, PIK3CA E545K
and BRAF V600E mutations were detected at 1% allelic frequency. In addition, in
normal plasma spiked with mix of mutant cell line DNA mutations were detected at 510 copies. Conclusion: The UltraSEEK Colon Cancer Panel on the MassARRAY
System has excellent accuracy and is able to detect mutations in KRAS, BRAF,
NRAS, PIK3CA and EGFR at low mutation frequencies with a short turnaround time
of 2 days. Validation with additional blood samples collected from patients with
mCRC and known tissue mutation status is underway.
ST12. Validation of a Neuro-Oncology Next Generation Sequencing 50-Gene
Panel
E. Barr Fritcher, J. Voss, Z. Tu, J. Balcom, J. Winters, Y. Sakai, X. Wu, T.
Kollmeyer, E. Klee, R. Jenkins, C. Giannini, C. Ida, B. Kipp
Mayo Clinic, Rochester, MN.
Introduction: A recent update of the WHO classification of central nervous system
(CNS) tumors incorporates molecular parameters. Therefore, in addition to
identifying prognostically and potential therapeutically relevant alterations, genetic
testing is critical for an integrated histological-molecular classification of multiple
CNS tumor entities including diffuse gliomas. In this validation study, we evaluated
the performance characteristics of a next generation sequencing (NGS) assay
targeting 50 genes (and the promoter region of TERT) associated with CNS tumors
in formalin-fixed paraffin embedded (FFPE) specimens. This test was designed to
aid in the diagnosis/classification and prognostication of CNS tumors. Methods:
DNA was extracted from 50 FFPE samples including 7 non-neoplastic epilepsy
surgery and 43 CNS tumors (≥40% tumor content). CNS tumor types included adult
diffuse gliomas and pediatric tumors. PCR-based target enrichment consisting of
amplification of 2392 target regions was performed using a Qiagen GeneRead
DNAseq Custom Panels V2 kit (Qiagen Inc., Germantown, MD). Using an orthogonal
sequencing approach, half of the final PCR product was utilized for Illumina library
preparation and sequenced on the HiSeq 2500 (Illumina, Inc., San Diego, CA) as the
primary method; the remaining half of final PCR product was utilized for Ion Torrent
library preparation and sequenced on the Ion Proton System (Thermo Fisher
Scientific, Waltham, MA) as the confirmatory method. Sequencing data were
processed through a custom bioinformatics pipeline and detected variants were
reviewed. Sequencing accuracy, variant detection accuracy, intra/inter-assay
precision, and limit of detection were assessed. Results: Sequencing accuracy was
99.76% compared to HapMap NA12878. Concordance was 100% for 45 variants
previously detected by Sanger sequencing in IDH1, IDH2, and TERT promoters.
Illumina sequencing (primary method) identified an additional 139 variants in gene
regions not previously interrogated. All but one variant detected by Illumina
sequencing were confirmed by Ion Proton (138/139, 99%). Concordance between
replicates was 100% for intra and inter-assay reproducibility. Allele frequency dilution
series (20%/10%/5%) and in silico read elimination experiments estimated analytical
sensitivity as 10% variant frequency with at least 100X depth of coverage.
Conclusions: We demonstrated analytical validity and robustness of a targeted 50gene Neuro-oncology NGS panel test suitable for clinical testing of FFPE specimens
that can be used to aid in the diagnosis/classification and prognostication of CNS
tumors. This test is accurate, reliable, and can simultaneously detect variants in 50
genes associated with CNS tumors at relatively low frequency (10%).
ST13. Performance Comparison of Two AR-V7 Detection Methods Confirms
That Unexpected Responses to Abiraterone/Enzalutamide in AR-V7 Positive
Patients are Not Due to Assay Differences
C. Bernemann1, J. Steinestel1, V. Humberg1, M. Boegemann1, A.J. Schrader1, J.K.
Lennerz2, J.K. Lennerz2
1University Hospital Muenster, Muenster, Germany; 2Massachusetts General
Hospital and Harvard Medical School, Boston, MA.
Introduction: The validity of the androgen receptor splice-variant 7 (AR-V7) in
circulating tumor cells (CTC's) of prostate cancer patients as a predictive biomarker
for non-response to next generation anti-androgen therapy (abiraterone or
enzalutamide), has recently been questioned. Specifically, a subset of AR-V7
positive patients showed responses to abiraterone or enzalutamide and assay
differences have been blamed as the underlying reason for these unexpected
findings. Methods: Here, we report a head-to-head performance comparison of 2
established mRNA-based AR-V7 detection technologies using either SYBR Green or
TaqMan chemistries. We assessed in vitro performance using 8 genitourinary
cancer cell lines and series dilutions in AR-V7 positive prostate cancer cell lines. In
addition we performed performance comparison in 32 clinical CTC samples from
castration resistant prostate cancer patients. Results: Both assays performed
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
identical in all cell lines and series dilution showed identical diagnostic
thresholds. Performance comparison in 32 clinical patient samples showed perfect
concordance of both assays. In particular, both assays determined AR-V7 positive
CTCs in 3 patients with unexpected responses to next-generation anti-androgen
therapy. Conclusions: To our knowledge this is the first direct performance
comparison of the 2 RNA-based AR-V7 detection assays. Given the identical
performance metrics, technical differences between the assays can safely be
excluded as the underlying reason for the unexpected responses to next-generation
anti-androgen therapy in a subset of AR-V7 patients. Our findings underscore that –
irrespective of the method used – AR-V7 positive patients should not systematically
be precluded from an otherwise safe treatment option.
ST14. Spectrum of Mutations in Metastatic Chondrosarcomas Identified by
Clinical Targeted Next-Generation Sequencing
P.J. Lee1, A.J. Mattis2, E.J. Duncavage1, I.S. Hagemann1
1Washington University, St. Louis, MO; 2Washington University School of Medicine,
St. Louis, MO.
Introduction: The landscape of recurrent somatic mutations in conventional and
variant chondrosarcomas has not been well established. The survival of patients
with chondrosarcoma with high grade and stage is poor, mainly due to the tumor’s
resistance to chemotherapy and radiotherapy. Elucidating the genomic landscape in
clinically aggressive chondrosarcomas may give insight into therapeutic options for
this rare tumor. Methods: Cases of metastatic chondrosarcoma were identified from
unselected clinical samples submitted for targeted clinical gene sequencing between
September 2012 and October 2013. Tumor DNA was enriched by hybrid-capture
covering all exons of 25 or 40 cancer-related genes and sequenced on an Illumina
HiSeq 2500. Single-nucleotide variants (SNV) and small insertions/deletions (indels)
were called using a clinically validated NGS pipeline based on the GATK pipeline (v
1.6). Results: Seven unique patients were identified, including 5 conventional, one
mesenchymal, and one extraskeletal myxoid chondrosarcoma. Five cases (5/7) were
sequenced from tumor metastases and all 7 cases (7/7) demonstrated documented
metastatic disease. Sequencing showed 10 non-synonymous, coding region
somatic variants within captured gene targets across 5 cases, with a mean of 1.4
variants per case (range 1–3 mutations). Two cases did not demonstrate somatic
variants. Mutations were non-synonymous missense variants in IDH1 (3/7 cases),
PDGFRA (2/7), MET (2/7), ATM, APC, and JAK2 (1/7 each). These genes are
commonly mutated in other cancer types. The mutational profile was not correlated
with histologic variant or grade. Conclusion: Our data show that alterations in
canonical cancer-related genes are common in chondrosarcomas. The observed
mutations fall into oncogenic pathways that may affect methylation,
apoptosis/survival, and PIK3A-AKT and RAS-ERK signaling. Some of these
mutations may have therapeutic implications. On the other hand, some tumors were
wild-type by our analysis. An expanded gene set, or testing for other mutation types,
may be necessary to identify actionable alterations in some high-risk
chondrosarcomas.
ST15. Intratumoral Heterogeneity is the Single Source of Assay Variability
During Laboratory Verification of the Prosigna Assay
A. Nelson1, J. Daniel1, M. Waknitz2, B. Billstein1, A. Luketich3, E. Racila1, M.E.
Klein1, D.J. Dabbs3
1University of Minnesota, Minneapolis, MN; 2M Health, Minneapolis, MN; 3University
of Pittsburgh Medical Center, Pittsburgh, PA.
Introduction: The Prosigna Assay is a gene expression classifier (GEC) clinically
validated to predict the 10 year risk of recurrence for breast cancer patients with low
stage, estrogen receptor positive tumors. It is the only FDA-cleared prognostic breast
cancer GEC intended for decentralized test implementation in independent
laboratories. Previous reports indicate intratumoral heterogeneity may cause
variability in breast cancer GEC results (Rouzier et al., Cancer Res 2017;
Gyanchandani et al., Clin Cancer Res 2016). A verification study of the Prosigna
assay in our laboratory demonstrated a high level of technical precision; only one
case with mixed histology resulted in significant discordance.Methods: FFPE breast
cancer samples were reviewed, macro-enriched, extracted, and processed on the
nCounter DxAnalysis system per manufacturer’s FDA-cleared protocols (Nanostring
Technologies, Seattle, USA). For intra-laboratory verification, 10 unique samples
were extracted in duplicate and run in a total of 38 replicates to assess variability.
Three samples were acquired from another academic center to assess interlaboratory concordance. Results: The 10 internal cases produced Risk of
Recurrence (ROR) scores ranging from 1-65, covering low, intermediate, and high
risk categories. The average standard deviation (SD) across replicate sets was 1.71
(range 0.5-2.49), within the manufacturer’s specified maximum variation of +/- 2.9
SD. Three external cases were extracted and tested, producing ROR scores
between 7-67 with SD < 1 across replicates. Two specimens were concordant
across labs (difference < 2 ROR units), while the third specimen showed a 10 point
ROR difference (58 internal versus 68 external). This case demonstrated both solid
micropapillary and mucinous invasive components. Re-extraction of only the
mucinous component was performed and an RNA aliquot of the original extraction
was obtained from the external institution. The internal re-extraction ROR = 64 while
the external RNA sample ROR = 68, identical to the original external result. Standard
deviations for all replicates were < 1. Conclusions: The Prosigna assay
demonstrated high analytic precision during verification of local instrumentation and
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
procedures. Ultimately, inter-laboratory analysis of exchanged samples
demonstrated high precision, including the analysis of an archived RNA sample from
the external laboratory. One inter-lab sample initially showed considerable variability,
but detailed investigation indicated this was due to intratumoral heterogeneity in a
mixed-histology specimen. These results stress the importance of local pathologist
participation in the annotation of tissue and assessment of results in GEC clinical
assays.
ST16. Relationship Between Forkhead Box M1 Gene Expression, KRAS
Mutation Status and Standard Uptake Value (SUV) of Positron Emission
Tomography (PET) in Non-Small Cell Lung Cancer (NSCLC)
W. Mahmud, S. Prasad, L. Buckingham
Rush University Medical Center, Chicago, IL.
Introduction: Forkhead Box M1 (FOXM1) is a transcription factor closely related to
the cell cycle and to expression of genes upregulated in
glycolysis. FOXM1 expression in non-small cell lung cancer (NSCLC) patients has
been shown to be a measure of poor prognosis. Activating mutations in
the KRAS oncogene promote cell proliferation. Standard uptake value (SUV) in
positron emission tomography is an indicator of glucose metabolism and high values
may predict low survival rates. In this study, we aim to find a relationship
between FOXM1 expression and maximum and minimum standard uptake value in
non-small cell lung cancers (NSCLC). Methods: Formalin fixed paraffin embedded
tissue from lobectomies/wedge resection and FDG-PET scan data from 55 patients
with primary NSCLC were included in this study. RNA was isolated and the gene
expression of FOXM1was measured by RT- qPCR. KRAS mutations were detected
by sequencing. Association between KRAS, FOXM1 and standard uptake values
(SUV) were assessed using the Mann-Whitney Test. Results: FOXM1 expression
was detected in 43 (78%) of patient samples. Those with FOXM1 expression had
higher SUVmax (mean 6.54+3.65, p=0.085) and SUVmin (mean 3.04+1.40,
p=0.020) in comparison to those who did not express FOXM1 (SUVmax 4.63+2.98,
SUVmin 1.99+1.22). KRAS data was available for 47/55 patients and was found to
be mutated in 11 (20%). No significant difference in SUVmax, SUVmin
nor FOXM1 expression was found between mutated versus non-mutated KRAS
(p=0.106, p=0.489). Conclusion: FOXM1 expression was found to have a
statistically significant correlation with SUVmin but not SUVmax in lung tumors.
ST17. Comparing Pyrosequencing and MALDI-TOF Mass Spectrometry to
Methylation-Specific qPCR for Quantifying MGMT Promoter Methylation
R.L. Margraf1, J. Raney1, A. Hall1, J. Nelson1, C. Vaughn1, K. Gligorich2, A. Matynia2
1ARUP Institute for Clinical and Experimental Pathology, Salt Lake City, UT; 2ARUP
Institute for Clinical and Experimental Pathology and University of Utah, Salt Lake
City, UT.
Introduction: Glioblastoma multiforme (GBM) is the most aggressive glial cancer
with low survival rates. Traditional treatment includes surgical resection followed by
radiotherapy plus concomitant and adjuvant chemotherapy with the alkylating agent
Temozolomide (TMZ). Methylation of the O6-methylguanine DNA methyltransferase
(MGMT) promoter region has been shown to be associated with a more favorable
prognosis and a strong predictor of responsiveness to alkylating chemotherapy, such
as TMZ. While methylation-specific qPCR is commonly used to
measure MGMT promoter methylation, the assay is semi-quantitative and
determination of a methylation cutoff is challenging due to variability of individual
CpG methylation levels. This study compares pyrosequencing and MALDI-TOF
mass spectrometry to methylation-specific qPCR to quantify MGMT promoter
methylation. Methods: DNA was isolated from tumor tissue that was formalin-fixed,
paraffin-embedded (FFPE) and underwent bisulfite conversion using the
ZymoResearch EZ-96 DNA Methylation-Lightning Kit, per manufacturer’s protocol.
Bisulfite-treated DNA was PCR amplified with methylation-non-specific primers
designed to the bisulfite converted MGMT promoter region. Samples that underwent
pyrosequencing were processed on the Qiagen PyroMark Q24 instrument for
analysis of 16 total CpGs, and percent methylation was measured using the Qiagen
PyroMark CpG software. Samples that underwent MALDI-TOF mass spectrometry
were treated with shrimp alkaline phosphatase post-PCR amplification, followed
by in vitro transcription and enzymatic digestion. The fragments were processed on
the Agena MassARRAY instrument, and the percent methylation measured over a
total of 12 CpGs was calculated using the Agena EpiTYPER software. Results:
Using a methylation cutoff of 10% average across CpGs assessed, 17/18 (94.4%)
clinical samples were concordant when comparing pyrosequencing to a validated
qPCR MGMT promoter methylation assay. In addition, 13/14 (92.8%) clinical
samples were concordant when comparing mass spectrometry to qPCR. The
discordant results were most likely due to the reduced sensitivity of methylationspecific qPCR and variation of individual CpG methylation frequencies. A titration
series was run on pyrosequencing and mass spectrometry resulting in a limit of
detection > 5% methylation for both methods. Conclusions: Pyrosequencing and
mass spectrometry appear to be advantageous options compared to methylationspecific qPCR for measuring MGMT promoter methylation. While both technologies
provide accurate quantification without the need to generate a standard curve, mass
spectrometry has increased throughput compared to a 24-well pyro-sequencer.
1011
AMP Abstracts
ST18. Implementation of Rapid Blood-Based Mutation Testing for Patients with
Lung Cancer
T.A. Boyle1,2, E. Toloza1,2, J. Gray1, E. Haura1, A. Magliocco1
1Moffitt Cancer Center, Tampa, FL; 2University of South Florida, Tampa, FL.
Introduction: The feasibility of testing tumor cell free nucleic acid (cfNA) in blood for
somatic mutations in advanced lung cancer is established. Yet, many questions
remain regarding implementation and interpretation of cfNA testing for best care of
patients. Methods: At the Moffitt Cancer Center, we implemented targeted cfNA
testing of blood specimens for EGFR/KRAS/BRAF/ALK mutations/translocations with
a digital droplet PCR based platform (Biodesix GeneStrat, ROS1/RET added later).
Implementation began with a quality assurance and feasibility project from July
through September in 2016 with bill-free testing available to thoracic surgeons and
medical oncologists caring for any patient suspected of having or previously
diagnosed with lung cancer. Genetic testing of tissue by pyrosequencing and/or next
generation sequencing (NGS) or blood by an NGS-based cfDNA assay
(Guardant360) was also performed on many of these cases as a part of routine care
during this time. Turn-around time was tracked from targeted cfNA order to report.
Results from cfNA testing were compared with results of other genetic testing and
any changes to care based on the cfNA results were noted.
Results: Of the first 46 cases with cfNA results, 25 had comparable tissue genetic
results with 78% concordance, 67% positive and 87% negative result agreement. Of
9 comparable blood NGS-based cfDNA results, there was 89% concordance with
100% positive and 75% negative result agreement. Twelve of the 46 (26.1%) cases
had at least one positive cfNA result. One discordant case was positive for
an EGFR exon 19 deletion and p.T790M resistance mutation by cfNA. EGFR exon
19 deletion was also positive by tissue testing and NGS-based cfDNA analysis but
not p.T790M. The only other discordant case with a positive result by cfNA and a
negative result by the most recent tissue had been previously positive for the same
EGFR mutation 4 years prior. Average turn-around time was 3.9 days, including nonworking days, much more rapid than the turn-around time for other genetic results.
Three patients had therapy changed based on cfNA detection of EGFR mutations,
all in the setting of progressive disease. Conclusions: Digital droplet cfNA testing is
an effective way to rapidly screen for somatic mutations and translocations,
particularly in the absence of tissue availability. The rapid turn-around time allows
more efficient enrollment into clinical trials which increasingly have
inclusion/exclusion criteria based on the genetic status of tumors. Careful
interpretation with action based on positive cfNA results and resumption of tissuebased genetic testing with no mutations identified by targeted testing may result in
the best care of patients with advanced lung cancer.
ST19. A Comparison of Mutation Frequencies Observed in Non-Small Cell
Lung Cancer (NSCLC) Patients by Two Different Methods: SNaPshot and
Polymerase Chain Reaction (PCR) Versus Next-Generation Sequencing (NGS)
M. Goudie, D. Gaston, K. Govier, W. Greer
Nova Scotia Health Authority, Halifax Nova Scotia, Canada.
Introduction: Since September 2012, the Molecular Diagnostics laboratory in
Halifax, Nova Scotia, Canada has performed mutation analysis on formalin-fixed,
paraffin-embedded (FFPE) lung tissue obtained from patients with NSCLC in Atlantic
Canada. From September 2012 until December 2016, targeted point mutations were
detected using SNaPshot and small insertions/deletions by fragment analysis. In
December 2016, these methods were replaced by NGS. This study compares the
frequencies of KRAS and EGFR mutations observed by each testing method to the
frequencies reported by My Cancer Genome and the Catalogue of Somatic
Mutations in Cancer (COSMIC). Methods: From 2012 to 2016, the lab used a
SNaPshot assay and PCR to test NSCLC tumors for certain mutations
in KRAS and EGFR, among other genes. In December of 2016, the SNaPshot and
PCR assays were replaced by NGS. The TruSight Tumor 15 Panel from Illumina
was used to prepare libraries which were then sequenced on the MiSeq
System. The SNaPshot and PCR methods were used to analyze 2664 samples, and
NGS was used to analyze 328 samples. A N-1 Chi-Square test was then used to
calculate the p-value to compare the results successfully obtained by each method
to the mutation frequencies reported by My Cancer Genome and COSMIC.
Results: KRAS: According to My Cancer Genome, KRAS mutations are reported in
15% to 25% of NSCLC tumors. This lab reported KRAS mutations at a frequency of
32% using the SNaPshot method and 42% using the NGS method. EGFR:
According the My Cancer Genome, EGFRmutations are reported in approximately
10% of NSCLC tumors in the US. This lab reported EGFR mutations at a frequency
of 8% using the SNaPshot and PCR methods and 12% using the NGS method.
Conclusions: KRAS: KRAS mutations are observed in this lab at a higher rate than
that reported by My Cancer Genome regardless of which testing method is
used. The fact that the NGS method detects a higher rate of KRAS mutations than
the SNaPshot method is partially due to the fact that the NGS method detects
mutations at more loci than does the SNaPshot method. However, when the
methods are compared for the same loci, the NGS method still detected mutations at
a higher rate (ie. KRAS c.34), reflecting a greater sensitivity of the latter
method. EGFR: Overall, EGFRmutations are reported by this lab at a frequency
similar to that reported in My Cancer Genome. The NGS method detects
more EGFR c.2573 mutations than the SNaPshot method and at a greater frequency
than that reported in My Cancer Genome. Since an EGFR c.2573 mutation was
detected by NGS in this lab in only 18 samples, additional samples may need to be
1012
analyzed in order to determine the accuracy of this value.
ST20. Comparison of the Clinical Utility of Microsatellite Instability Detection
Approach between a Novel NGS Based Algorithm and Traditional PCR Method
C. Wang1, X. Li1, Y. Cai1, Y. Yang1, H. Liu2, M. Tan2, M. Ye2, M. Mao3
1Beijing Genomics Institute (BGI)-Tianjin, Tianjin, China; 2Beijing Genomics InstituteGuangzhou, China; 3Beijing Genomics Institute (BGI) Shenzhen, Shenzhen, China.
Introduction: Microsatellite instability (MSI) as a useful marker for risk assessment,
prediction of chemotherapy responsiveness and prognosis in patients with colorectal
cancer. Current clinical practice using MSI-PCR and immunohistochemistry.
However, a comprehensive methodology to detect MSI status using next generation
DNA sequencing (NGS) data remains to be developed. Methods: We investigated a
novel Euclid-distance based algorithm to identify mononucleotide and dinucleotide
MSI loci by focusing the capture design region on 13 specific MSI loci of interest
using a total of 85 patients undergoing resection of colon tumor. Comparing with
traditional approaches, the MSI algorithm considered both BW-alignment and peak
adjusted to call base-pair deviations, which using an alignment-free approach to
measure single-base repetitive sequence length and its corresponding support reads
in each of the MSI risk regions. MSI status was determined by the fraction of
unstable microsatellite loci. Results: We used a total of 85 samples to examine the
accuracy of MSI, which composed of 42 retrospective samples and 43 prospective
samples. The novel approach illustrated to be both diagnostically sensitive (range of
89.26% to 90.24% across 2 cohorts) and specific (range of 50% to 100%) compared
to MSI-PCR as a gold standard. In the retrospective series, the approach was able to
determine 25 MSI-High, 7 MSI-Low with overall accuracy of 76.19%. In the second
series, there were 37 MSI-High and 2 MSS tumors with overall accuracy of 97.19%.
The fraction of unstable microsatellite markers calculated from sequencing data were
found correlated with the number of unstable loci detected by conventional MSI-PCR
testing. Comparing with traditional computational method, the new algorithm shows a
significant improvement with over-all accuracy of 11% to 26%. Phenotype analysis
demonstrated that Right-Sided colon cancer patients prefer to have MSI-High status
(Pearson correlation, cor=0.33, p=0.0025).
Conclusions: In sum, our distance-based approach demonstrated high accuracy
across all testing samples, and in many cases with restricted sets of well-performing
MSI loci. The study highlights the potential utility of our method in classifying MSI
status by InDel-Peak-method, allowing its incorporation into existing NGS pipelines.
ST21. Development and Evaluation of a Pan-Sarcoma Fusion Gene Detection
Assay Using the NanoString nCounter Platform
K.T. Chang1 ,4, A. Goytain2 ,4, T. Tucker3, C. Lee2, T.O. Nielsen2, T.L. Ng2
1KK Women's and Children's Hospital, Singapore; 2University of British Columbia,
Vancouver, Canada; 3British Columbia Cancer Agency, Vancouver, Canada; 4 DukeNUS Medical School, Singapore.
Introduction: Molecular charaterization of sarcomas is often critical for pathological
diagnosis and to direct patient management. The NanoString nCounter assay is a
high-throughput hybridization technique using a CodeSet comprising target-specific
probes that can be customized to test for numerous fusion transcripts in a single
assay utilizing RNA from formalin-fixed paraffin-embedded (FFPE) material. In this
study, we assess the value of the NanoString assay in fusion gene detection
in patient samples. Methods: We designed a NanoString assay targeting 174 unique
fusion junctions in 25 sarcoma types representing the large majority of sarcoma
fusion types and variants. The study cohort comprised resections and core biopsies
from 212 soft tissue tumours. RNA extracted from FFPE tissue was hybridized with
the sarcoma fusion CodeSet according to the nCounter protocol and analyzed with
nSolver software. Results: Ninety-six cases showed fusion gene expression by the
NanoString assay, including all 20 Ewing sarcomas, 11 synovial sarcomas and 5
myxoid liposarcomas tested. Amongst these 96 cases, 14 showed fusion expression
not identified by standard clinical assay, including EWSR1-FLI, EWSR1ERG, BCOR-CCNB3, ZCHB4-BCOR, HEY1-NCOA2, CIC-DUX4, COL1A1PDGFB, MYH9-USP6 and IRF2BP2-CDX1 fusions, leading to a change in diagnosis
in these cases. There were no false positive results, while 4 cases were false
negative when compared to clinically-available fluorescence in situ hybridization or
reverse transcriptase-polymerase chain reaction testing. When batched as 6 cases,
the per sample reagent cost was less than conventional techniques such as
fluorescence in situ hybridization, with technologist hands-on time of 1.2 hours per
case and assay time of 36 hours. Conclusions: The NanoString nCounter Sarcoma
Fusion CodeSet reliably and cost-effectively identifies fusion genes in sarcomas
utilizing formalin-fixed paraffin-embedded material including many fusions missed by
standard clinical assays, and can serve as a first-line clinical diagnostic test for
sarcoma fusion gene identification, replacing multiple individual clinical assays.
ST22. Genome-Wide Copy Number Variation and Targeted Next-Generation
Sequencing Studies of Merkel Cell Carcinoma
M. Carter, D. Gaston, W. Huang, W. Greer, S. Pasternak, T. Ly, N.M. Walsh
Nova Scotia Health Authority and Dalhousie University, Halfiax, Nova Scotia,
Canada.
Introduction: Merkel cell carcinoma (MCC), an aggressive primary cutaneous
neuroendocrine tumor, arises usually due to clonal integration of the Merkel cell
polyomavirus (MCPyV) in neoplastic cells and less commonly due to DNA damage
by ultraviolet (UV) light. Genetic studies on these MCPyV-positive (MCPyV+) and
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
negative (MCPyV-) subsets of the tumor have shown inter-group differences in the
overall mutational burdens and in the mutational profiles of certain tumor suppressor
genes (eg RB1 and TP53). Our goal was to further investigate genetic patterns in a
series of MCCs stratified by viral status. Methods: Formalin fixed paraffinembedded (FFPE) tissue from a cohort of MCCs was used for the study. Twentyeight cases (9 MCPyV+ and 19 MCPyV-) were analyzed by Oncoscan FFPE kit
(Affymetrix, Inc.) to study genome-wide copy number variations (CNV). Targeted
next-generation sequencing (NGS) studies, using a panel of 9 genes commonly
affected in MCC, was performed on 46 MCCs (21 MCPyV+ and 25 MCPyV-) to
identify specific mutations. Results: Significantly more CNVs and a greater fraction
of the genome was changed in MCPyV- tumors relative to MCPyV+ cases (p<0.01
for both comparisons). Extra copies of chromosomes 1p and 3q were commonly
found in MCPyV- tumors, but never in MCPyV+ cases, while gains in 1q, 5p, 6, and
19 were common in both. Loss of chromosomes 3p, 4, 5q, 8p, and 13p was
common in MCPyV-, but not MCPyV+, cases. Copy number loss of RB1 or an
inactivating RB1 mutation (either or both) was common in MCPyV- tumors (15/19,
79%) but not MCPyV+ cases (2/9, 22%). A similar trend was seen for TP53, with
several of the MCPyV- tumors (7/19, 37%) showing gene copy number loss or
inactivating mutations compared to none in the MCPyV+ group (0/9, 0%). Three
focal CNV findings were observed with similar frequency in virus-positive and negative groups. The first was biallelic loss at 8p11.22 (9/28, 32%), a region
encoding ADAM metalloproteinase domains 3A, 5, and 18. The second was high
level gain at 1p13.3 (6/28, 21%), corresponding to genes for glutathione-Stransferase mu. The third, and possibly related, finding was focal loss of genes
encoding glutathione-S-transferase theta at 22q11.23 (8/28, 29%).
Conclusions: Our results support the concept that divergent pathways dependent
on either MCPyV or UV light are involved in the tumorigenesis of MCC. The former is
driven by few viral oncogenic DNA alterations and the latter by abundant UV lightmediated DNA changes. Loss of genes involved in the ADAM metalloproteinase
pathway, already known to occur in high-grade gliomas and neuroblastomas, and
dysregulation of the glutathione-S-transferase pathway reflect pathogenetic
mechanisms common to both subsets of the tumor and are novel findings.
ST23. Study of TMPRSS2-ERG Molecular Translocation in Prostate Cancer and
its Correlation with Clinical and Histopathological Parameters
S. Desai, S. Balagangadharan, O. Shetty, P. Bapat, A. Singh
Tata Memorial Centre, Mumbai, Maharashtra, India.
Introduction: Recurrent gene fusions between TMPRSS2 and ETS (E26
transformation specific transcription factors) family have been identified in prostate
cancer. The most common ETS related gene involved in this type of fusion
is ERG.TMPRSS2–ERG gene fusion has been implicated in androgen receptor
stimulated overexpression of ETS transcription factors, which may modulate prostate
cancer development. We undertook this retrospective study to evaluate the
occurrence of TMPRSS2 and ERG fusion in prostate cancer patients at our institute
with an aim to correlate these translocations with histological, biochemical and
clinical parameters. Methods: Thirty cases of prostatic adenocarcinoma diagnosed
on biopsy (n=23), Transurethral Resection of prostate (TURP, n=1) &
prostatetctomy (n=6) formed the study group. Fluorescent in situ hybridization (FISH)
(ZytoLight SPEC ERG/TMPRSS2 TriCheckProbe, ZytoVision, Germany) and Real
Time (RT) PCR (TaqMan Gene Expression Assay (Catalogue No. 4331182, Assay
ID- Hs03063375_ft, Applied Biosystems, USA) were performed to detect TMPRSS2ERG fusion in all the cases and their results were correlated with the clinical and
histopathological parameters. Results: The median age of the study group was 68
years and median PSA levels were 63.12ng/ml (5.84-3614.85ng/ml). Clinical stage
was known in 29 cases, out of which 4 cases (13.3%) had organ confined prostate
cancer, 8 (26.7%) cases were locally advanced (T3,T4) at presentation, one (3.3%)
case showed lymph node metastasis and 16 (53.3%) presented with distant
metastasis. The most frequently encountered Gleason score was 4 + 5 = 9 and was
seen in 30% cases. Thirteen (43.3%) cases were positive for TMPRSS2–ERG fusion
by both techniques i.e. FISH and RT- PCR. In addition, discordance between the
results of 2 techniques was seen in 2 cases. Concordance between FISH and RT
PCR was 86.6% with a p value <0.0001 (Cohen’s Kappa). Additionally, FISH was
able to detect an increase in ERG and TMPRSS2 gene copy number and presence
of TMPRSS2 abnormality without associated abnormalities of ERG gene. There was
no statistical correlation between the presence ofTMPRSS2–ERG fusion positivity
with Gleason score, PSA levels, stage of disease, biochemical recurrence and
disease free survival. Conclusions: The high prevalence of TMPRSS2–
ERG rearrangement in prostatic adenocarcinoma is suggestive of early occurrence
of these molecular events in the development of prostate carcinogenesis. FISH and
RT-PCR have a high concordance for detection of gene fusion. FISH is useful in
detecting additional molecular abnormalities other than the common TMPRSS2–
ERG fusion. No statistical correlation could be demonstrated between the presence
of fusion positivity and any clinicopathological parameter.
ST24. A Rare Case of HER2 Amplified Invasive Ductal Breast Carcinoma with
Pericentric Deletion of Chromosome 17
B.S. Karir, S.P. Naber, J.M. Cowan, M. Roychowdhury
Tufts Medical Center, Boston, MA.
Introduction: Human epidermal growth factor receptor 2 (HER2) amplification status
in breast cancer is assessed by dual-probe fluorescence in situ hybridization (FISH)
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
or chromogenic in situ hybridization (CISH) assay. The assay is interpreted
as HER2:CEP17 ratio based on the number of HER2 and CEP17 (centromere
enumeration probe for chromosome 17) signals/cell. Multiple studies have shown
gain of chromosome 17 (ranging from 10-50% of cases) in cases with Her2
amplification. CEP17 appears to be related to pathologic complete response to
primary anthracycline based-therapy when this marker is combined with Her2 status.
This is the first report of a case of deletion of CEP17 in a case with Her 2
amplification. Methods: A 55-year-old female patient underwent biopsy of an axillary
mass results of which showed a high grade invasive ductal carcinoma.
Immunohistochemistry was performed for Estrogen Receptor (ER), Progesterone
Receptor (PR) and Her2. Following this, CISH testing for Her2 was performed and
25 cells were counted. To confirm the deletion of only centromeric area of
chromosome 17, FISH for p53 gene (located on short arm of chromosome 17) was
performed which was positive. Results: Immunohistochemistry was negative for
estrogen receptor (ER) and positive for HER2 (3+). CEP17 signals were not detected
in tumor cells, however, mean HER2 signals/cell was 15.2 resulting in an amplified
HER2 status. All adjacent non-neoplastic epithelial and stromal cells exhibited
normal chromosome 17 and HER2 signals. The on-slide external control showed
appropriate hybridization. To exclude the possibility of missing short arm of
chromosome 17, FISH for p53 gene was performed and was positive, confirming the
presence of the short arm of chromosome 17. The result was interpreted as HER2
amplified breast cancer with loss of pericentromeric chromosome 17 appearing as a
tumor specific event, possibly due to a clonal deletion event. Conclusions: This
case brings awareness to the unique finding of CEP17 deletion in a high grade
invasive ductal breast carcinoma. This finding is likely due to a clonal deletion event
in the tumor cells. At this time, clinical and therapeutic implications of this finding are
not known. However, reporting such findings and collecting their database is an
important step for future research of novel mutations.
ST25. Biallelic TP53 Gain of Function Mutations in Rapid Progressing Solid
Tumors and Correlating Immunohistochemistry
C.M. Sande, B. Chang, V. Monga, D. Ma
University of Iowa Hospitals and Clinics, Iowa City, IA.
Introduction: TP53 is the most frequently mutated gene in cancer resulting in loss
of tumor suppressor activity in concert with loss of heterozygosity or from exerting a
dominant-negative effect over the remaining wild-type allele. Recent studies are
discovering TP53 mutations with gain of function (GOF) properties that promote
tumorigenesis via a variety of pathways and mechanisms. To our knowledge, all
compound mutations in oncogenes are reported to be allelic. We identified 2 patients
with biallelic GOF TP53 mutations in their tumors and a third with allelic compound
variants. The correlation with p53 expression was also examined. Methods: We
identified 3 patients with TP53 double mutations. Patient (Pt) 1 was a 20 year-old
male with glioblastoma (GBM); Pt 2 was a 56 year-old male with T3 penile
squamous cell carcinoma; and Pt 3 was a 63 year-old male with stage IV lung
adenocarcinoma. Formalin-fixed, paraffin-embedded tissue blocks were selected
and areas of interest (>50% tumor) were microdissected, followed by DNA extraction
(Qiagen kit). Library construction utilized the Ion AmpliSeq Cancer HotSpot Panel V2
(ThermoFisher) and sequencing was performed on the Ion PGM or S5
(ThermoFisher). Data was analyzed with a laboratory developed pipeline
incorporating the Torrent Variant Caller. Immunohistochemistry studies (IHC) for p53
(Dako M7001; 1:100) were also performed. Results: All mutations were located in
exon 8, which encodes part of the p53 DNA-binding domain. Pt 1 and 2 had
compound GOF mutations (p.R273H and p.R273C) that were present on different
alleles at frequencies of 94% and 48% in Pt 1, and 19% and 27% in Pt 2. Pt 3 had
one GOF variant and one not previously classified as a GOF variant (p.P278S and
p.R283L) present on the same allele both at frequencies of 61%. No other variants
were found in these cases. Germline testing for Pt 1 confirmed wildtype TP53,
but was not done in the others. IHC for p53 showed strong nuclear staining in the
tumors from Pt 1 and 2 and null staining in that of Pt 3. All 3 patients experienced
rapid disease progression, succumbing within 2 years of diagnosis, despite
aggressive treatment. Conclusions: The presence of biallelic TP53 GOF mutations
may account for the unusual early presentation of GBM in patient 1 and rapid clinical
deterioration in all 3 patients. Furthermore, strong nuclear localization detected by
IHC may be associated with biallelic GOF mutations whereas the compound allelic
mutation containing a GOF and non-GOF variant may not demonstrate this
pattern. Targeted therapy against mutant p53 is already of clinical
interest. Implementation of TP53 mutation analysis in clinical practice may predict
patient outcome and inhibition of GOF could represent an attractive target for
therapy.
ST26. 1p Deletion, The Most Common Subtype of Leiomyomas Encountered in
NIPT?
M. Van Ness, T. Boomer, E. Almasri, J. Saldivar
Sequenom, San Diego, CA.
Introduction: Our group has shown that a 7 cm uterine leiomyoma with cytogenetic
abnormalities resulted in a non-reportable NIPT result due to multiple genomic
abnormalities. Leiomyomas (fibroids) can be chromosomal unstable demonstrating
gains and losses of chromosomal material. Although cytogenetic abnormalities
appear in only 50% of leiomyomas, the following recurrent cytogenetic deletions
haven been identified 1p, 7q22, 10q22, 13q, 22q. These deletions can appear with
1013
AMP Abstracts
other alterations, likely representing secondary driver events, often present only in a
subpopulation of tumor cells occurring more often in larger leiomyomas.
Methods: We review 17 individual patient cases with non-reportable MPSS NIPT
that had shown reproducible findings by 2nd aliquoted testing and after genetic
counselor investigation was found to have a clinical history of leiomyomas. We
reviewed these cases for known recurrent chromosome abnormalities seen in
leiomyomas. Results: Among these 17 cases 52 chromosome specific events were
identified. Events per case ranged from 1 to 6 with an average of 3 events per case.
Deletions were more common than duplications. In evaluating our cases for 5
recurrent chromosome abnormalities known to occur leiomyomas (1p, 7q22, 10q22,
13q, 22q), 76% (13/17) with a non-reportable result and a clinical history of
leiomyomas demonstrated one or more of these deletions. The most common
chromosome to have an event was chromosome 1 seen in 53% of cases (9/17). In
our series 1p deletions were always seen with another chromosomal change,
consistent with the literature. All 1p deletions appear to involve the leiomyoma
specific 1p minimally deleted region. The second most commonly involved
chromosome was chromosome 13 seen in 29% of our cases (5/17). Similar to 1p
deletions, 13q deletions were always seen with another chromosomal change.
Conclusions: Commonly used NIPT testing can vary from whole genome evaluation
or just core chromosome (21, 18, and 13) evaluation. Since partial 13q deletions are
the second most commonly encountered abnormality in our series, it is likely that this
known recurrent leiomyoma related abnormality is of clinical significance in the
setting of a non-reportable NIPT. In the setting of a non-reportable NIPT result and a
clinical history of fibroids, 1p deletions are the most common findings, consistent with
a known recurrent cytogenetic subgroup of leiomyomas. As NIPT expands to involve
sub- chromosomal CNVs expertise and clinical correlation will be necessary to avoid
undue procedures (CVS or amniocenteses) for false positive NIPT cases that merely
represent a recurrent cytogenetic finding in the most common pelvic tumor in women
of reproductive age.
a molecular barcode based NGS approach for detection mutations with low mutation
allelic fraction (MAF) in cfDNA samples from patients with colorectal carcinoma.
Methods: cfDNA was extracted from plasma using the QIAamp circulating nucleic
acid kit in 37 patients with metastatic colorectal cancer. Mutations in tumor tissue
and in 18 cfDNA samples (n=18) were detected by using the Ion Torrent platform
and the Ampliseq Cancer Hotspot Panel v2. 20ng cfDNA sample was used for library
preparation using the Oncomine cfDNA colon panel according to the manufacturer’s
instructions (Thermofisher Sc, CA). Six quantified libraries were pooled per run and
sequenced on the Ion S5XL using S530 Chip. Data were analyzed for variant calling
and annotation using the Torrent Suite v5.2 and Ion Reporter, respectively with
250,00x average coverage, 2,500x molecular depth and a minimum 3 molecular
counts as a cutoff. Limit of detection studies were performed using Horizon cfDNA.
Results: cfDNA yield ranged from 3.7 to 284.5 ng/ml among the tumors tested. Tag
sequencing detected 34 of 54 (63%) expected tissue mutations and 43 additional
clones in plasma cfDNA. MAF of new clones detected in cfDNA ranged from 0.1% to
23%. An average sequencing depth of 49,732x and 2,700x molecular depth was
achieved per sample. Lower variant calls missed in tissue might be attributable to the
lower sensitivity of the Ampliseq assay (5%), whereas the variant calls missed in
cfDNA may be due to the time interval between tissue and plasma collection (range,
58 to 1,637 days). 31 variant calls in a subset of plasma cfDNA samples were
compared between Oncomine colorectal cfDNA and the Ampliseq panel. 28 of 31
(90%) expected variant calls were detected by both panels with 15 additional calls
detected by Tag sequencing only (MAF 0.1% to 23%). Sensitivity studies showed a
lower limit of detection of 0.1%. Tag sequencing required less hands-on time (~2 hrs)
compared with Ampliseq sequencing (~6 hrs). Conclusions: Tag sequencing
provides higher sensitivity (0.1%) for detecting low level mutations in cfDNA.
Detection of these low level mutations in patients with colorectal cancer will enable
real-time disease monitoring and may facilitate novel molecular discoveries that may
help in the application of targeted personalized therapies.
ST27. A Novel Non-Invasive Bladder Cancer Recurrence Surveillance Test
Using Urine Sample
P. Piatti1, M. Suwoto2, X. Yang1, W. Guo1, X. Jia1, Y. Chew1, G. Liang3
1Zymo Research Corporation, Irvine, CA; 2Pangea Laboratory, Costa Mesa,
CA; 3University of Southern California, Los Angeles, CA.
Introduction: Bladder cancer (BC) ranks as the 4th most common cancer in males,
with approximately 77,000 new cases and 16,000 deaths estimated in the United
States in 2016. More than 50% of BC patients show tumor recurrence within oneyear after transurethral resection of the bladder tumor (TURBT), and as a result, BC
patients commonly require life-long monitoring using cystoscopy, an invasive and
costly practice. This life-long monitoring procedure makes BC one of the most
expensive cancers to manage and thus, a simple, non-invasive, and cost effective
test to monitor bladder cancer patients is desired. In this study, we introduced a
novel and non-invasive assay, the BladderCARE test, for the bladder cancer
recurrence surveillance by analyzing the methylation status of BC-specific
biomarkers in urine DNA. Method: A) Urine DNA Extraction: Upon urine sample
collection, a uniquely formulated stabilization reagent was added to preserve urine
nucleic acids at ambient temperature. Next, the Quick-DNA Urine Kit was utilized to
obtain high quality urine DNA. B) Marker Discovery: BC-specific methylation markers
were identified by analyzing bladder tumor (t) and normal (n) tissues (t= 228, n= 71)
and urine (t= 36, n= 30) samples from patients and healthy donors using the HM450
human methylation array and RRBS next-generation sequencing. A linear
regression model was utilized to select 12 unique BC biomarkers. C)
BladderCARE Test: We developed a novel non-bisulfite based assay that allows the
simultaneous selection and amplification of methylated DNA. The test allows the
precise quantification of the methylation level of the 12 bladder cancer-specific
biomarkers in less than 4 hours. Results: BladderCARE test was used in a preclinical study on a cohort of 53 individuals (33 BC patients, 20 healthy individuals).
The test was able to discriminate BC samples from control samples with 91.7%
sensitivity and 100% specificity. In addition, we are currently in the process of
validating the performances of BladderCARE test in a larger cohort (n=
130). Conclusions: Here, we presented a successful application of our technology,
the BladderCARE test, in the BC recurrence surveillance. The test is non-invasive,
sensitive, and cost-effective, thereby allowing physicians to simplify and streamline
the BC patients’ monitoring process, from the sample collection to the laboratory
results.
ST29. Effect of Blood Collection Tubes on Circulating Tumor DNA (ctDNA)
Yield and Specificity
D. Murray, N. Boulter, S. Pedersen, L. LaPointe
Clinical Genomics Technologies Pty Ltd, North Ryde, North South Wales, United
Kingdom.
Introduction: Solid tumors, such as colorectal cancer (CRC), may leak DNA
(circulating tumor-DNA, ctDNA) into blood. The fraction of ctDNA within the pool of
circulating cell-free DNA (ccfDNA) extracted from plasma specimens can be
detected by assaying ccfDNA for tumor-specific features such as
methylated BCAT1 and IKZF1. Accurate identification of ctDNA relies on minimum
release of genomic DNA such as from lysed white blood cells during blood collection
and transport. The aim of this study was to evaluate a new blood tube in a cohort of
patients with and without CRC. Methods: Blood was collected in K3EDTA and
PAXgene Blood ccfDNA tubes (Qiagen) from colonoscopy confirmed subjects.
Plasma was isolated by centrifugation within 8 hours of K3EDTA blood collection,
whereas the PAXgene tubes were stored at room temperature for 7 days prior to
plasma isolation. ccfDNA was extracted using either QIAsymphony (QS) DSP or QS
PAXgene Blood ccfDNA extraction kits (Qiagen) respectively. Successful recovery of
ccfDNA (ACTB) was measured simultaneously with detection of ctDNA by assaying
for the presence of methylated BCAT1 and IKZF1 DNA. Results: The 2 systems
were evaluated in 40 cases with no evidence of disease (NED) and 39 CRC cases
(including 9 stage I, 7 stage II, 15 stage III and 8 with no stage information). Higher
ccfDNA yield was measured in the PAXgene system (median: 8ng/mL, IQR: 5–20)
compared to the K3EDTA system (4ng/mL, 2–9). There was no significant difference
in the ctDNA positivity rates (NED: PAXgene, 2.5%; K3EDTA, 12.5%; p = 0.1025.
CRC: PAXgene, 64.1%; K3EDTA, 59.0%; p = 0.4795). There was a high
concordance in blood ctDNA test results for both collection systems with 85% in
NED subjects and 80% in CRC subjects. Conclusions: Storing whole blood for 7
days in the new PAXgene Blood ccfDNA collection tube has no downstream effect
on accurate detection of methylated BCAT1/IKZF1 ctDNA in CRC patients. The
PAXgene Blood ccfDNA system may provide a solution for maintaining preanalytical
sample integrity prior to ctDNA testing.
ST28. Ultra-Sensitive Tag Sequencing for Detection of Low Level Somatic
Alterations in Plasma Cell Free DNA of Metastatic Colorectal Tumors on Ion
S5XL Platform
M. Mehrotra, D.Y. Duose, R.R. Singh, C. Lan, B.A. Barkoh, M.J. Routbort, K.P.
Patel, L.J. Medeiros, S. Kopetz, I.I. Wistuba, R. Luthra
University of Texas M.D. Anderson Cancer Center, Houston, TX.
Introduction: Colorectal tumors are characterized by a high frequency of genetic
alterations. Recently, liquid biopsy has emerged as an alternative to tissue biopsy. A
tumor-derived fraction of circulating cell-free DNA (cfDNA), isolated from blood
samples, allows real-time monitoring of disease and detection of mutations during
treatment. Detection of low level mutations in cfDNA requires highly sensitive
platforms with 0.1 to 0.01% detection sensitivity. Here we describe Tag sequencing,
1014
ST30. Successful Lung Cancer EGFR Sequencing from DNA Extracted from
TTF-1 Immunohistochemistry Slides: A New Means to Extend Insufficient
Tissue
G. Deftereos1,2, V. Carter2, A. Sandoval2, L.V. Furtado1, A. Matynia1, M. Bronner1
1University of Utah, Salt Lake City, UT; 2ARUP Laboratories, Salt Lake City, UT.
Introduction: Lung cancer biopsy material is characteristically limited but required
for morphologic diagnosis, immunohistochemical and molecular testing. The breadth
of testing is commonly, therefore, complicated by insufficient tissue. This may lead to
repeat biopsies that can delay diagnosis and testing, and seriously increase
morbidity risk. Moreover, for some patients, overall lung function status and
comorbidities may render repeat biopsy impossible. As such, there is an increased
need to optimize pre-analytical specimen use for molecular testing. We hypothesize
that both hematoxylin/eosin (H&E) and immunohistochemistry (IHC) slides can be
repurposed for DNA extraction and molecular testing. While H&E slides have been
validated as a source of DNA for many clinical molecular tests, little is known on the
potential of IHC slides for this purpose. IHC exposes tissues to harsh conditions,
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
including high temperature and pressure that could lead to DNA degradation. The
aim of this study is to evaluate whether DNA extracted from TTF-1 lung cancer IHCstained slides (the most commonly used stain for lung adenocarcinomas) can be
effectively tested for EGFR mutation by pyrosequencing, comparing original (nonIHC) FFPE extraction results to post-IHC extractions. Methods: A total of 22
samples of lung adenocarcinoma (11 wild-type and 11 mutated by EGFR testing)
were selected and additional standard aniline blue-stained slides were destained and
submitted for TTF-1 IHC. Following successful TTF-1 IHC staining, slides were
decoverslipped and macrodissected for EGFR testing. Pyrosequencing testing for
mutations in exons 18, 19, 20 and 21 of EGFR was performed comparing results
both with and without IHC. Results: All 22 TTF-1 samples produced successful
sequencing results. Of the 11 cases with EGFR mutations on the original FFPE
EGFR testing, 10 matching TTF-1 IHC samples showed identical mutations,
including point mutations and exon 19 deletions/insertions. One case with an L858R
mutation on the original FFPE EGFR testing with a 21% variant allele frequency was
negative on repeat sequencing of the TTF-1 IHC sample, possibly due to lower
tumor burden on the TTF-1 stained tissue. All 11 cases with wild-type EGFR
originally, showed wild-type EGFR results on the matched TTF-1
samples. Conclusions: The results of this study demonstrate that TTF-1 IHC slides
are a viable source of DNA for molecular testing, especially in small lung biopsies
with insufficient material following morphological evaluation and IHC testing.
ST31. Testing for Segmental Chromosomal Aberrations of Multiple Genes
Using Multiplex Ligation-Dependent Probe Amplification (MLPA) Technique in
Children with Neuroblastoma.
M. Ramadwar, O. Shetty, G. Chinnaswammy, S. Qureshi, T. Vora, S. Medhi, B.
Rekhi, N. Khanna
Tata Memorial Hospital, Mumbai Maharashtra, India.
Introduction: Risk stratification of patients with neuroblastoma (NB) is driven by
tumour biology. It enables selection of specific treatment options. Along with MYCN
gene amplification (MNA), several segmental chromosomal aberrations (SCA) and
ploidy are now known to influence the biologic tumour behaviour of NB. Hence It has
becomes imperative for the treating team to test for these molecular parameters
before initiation of treatment. MLPA is a PCR-based technique which enables testing
for gains, losses and amplifications of multiple genes in one single assay. We
studied the spectrum of genetic abnormalities observed in 40 patients of NB
by MLPA. Methods: Study was performed on formalin fixed paraffin embedded
tissues of NB between 2014 and 2017. Three panels of probe mixes P251, P252
and P253 (MRC Holland, Amsterdam) were used in order to detect segmental
chromosomal aberrations (SCA) at chromosomes 1p, 3p, 14q and 11q, unbalanced
gain of 1q, 11p and 17q and genomic aberrations in ALK, DDX1. Veriti Thermal
cycler was used to carry out the MLPA reaction and amplification followed by
fragment analysis on ABI 3500 Genetic Analyzer.Coffalyser software was used for
data analysis. SIOPEN guidelines were used for nomenclature. Results: The age
group ranged from 1 to 24 months with a median age of 1 year. 18 of 40 patients
were below 18 months of age. Interpretable MLPA results could be obtained in 36
/40 samples. SCA were most commonly found as 17q gains and amplification
(25/36), 11q loss (12/36), 3p loss (11/36), 2p gains and amplifications (19/36
inclusive of MNA) and 17p gains (11/36),1p aberration (15/ 36) and 14q loss
(9/36)cases. MNA was found in 6 /36 patients including one infantile NB. Remaining
17 infantile NB showed SCA in 1p, 11q and 17q. Alk amplification was seen in one
case. A combination of gains, losses and amplifications were seen in all
chromosomal arms. None of the tumours showed 12p
aberrations. Conclusions: SCA were commonly found in 17q, 11q, and 1p, which
are already proved to impart adverse prognostic impact. All other chromosomal arms
except 12p also showed SCA; 3p, 2p, 17p being the commonest. Thus MLPA is an
effective tool which enables accurate testing for SCA as well as MNA in one single
diagnostic system. It has become easier and cost effective to conform to INRG and
SIOP-European Neuroblastoma Biology committee recommendations. However, we
would like to highlight a difficulty in interpretation when the values did not fall into
defined categories. Moreover, quality of DNA is one of the most important
determinant of accurate results while working with the FFPE tissues. This technique
is also enabling us to gather prospective data in order to determine clinical
significance of each of the SCAs detected.
ST32. Multiple Mutations in TP53: Tumor-Specific Patterns and Their
Implications for Breast Cancer Pathogenesis and Variant Annotation
J. Coleman1, W. Zhao2, P. Ru2, S. Caruthers2, J.B. Evans2, C. Miller1, H. Tu2, D.
Jones3
1GenomOncology, Cleveland, OH; 2The Ohio State University, Columbus, OH; 3The
Ohio State University and The James Cancer Hospital & Solove Research Institute,
Columbus, OH.
Introduction: TP53 dysregulation by mutation or deletion is a critical step in the
development of many tumors, including secondary acute myeloid leukemia (AML),
non-small cell lung carcinoma (NSCLC), colorectal cancer (CRC) and breast cancer
(BC). Variant-specific TP53 mutation associations are known but the number of
distinct mutations and their variable effects on loss-of-function (LOF) make
interpretation complex. We studied tumors with multiple co-occurring TP53 mutations
to derive some associations to assist in annotation. Methods: Tumor TP53
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
sequence was obtained from an Ion Torrent amplicon assay (PGM platform) and a
TruSeq DNA repair gene panel (NextSeq). Equivalent sensitivity of the platforms was
established and all TP53 exonic calls down to 4% variant allele fraction (VAF) were
included. Known SNPs and low-confidence calls were excluded. Predicted mutation
effect was derived from the IARC and UMD databases. Effects of TP53 VUS were
imputed using a classifier [random forest (RF) model comprising 100 trees
integrating in silico multiple prediction tools from dbNSFP] trained with TP53 variants
of known effect. Results: At least one missense (ms), nonsense (ns) or frameshift
(fs) TP53 mutation/VUS was seen in 89.9% of metastatic CRC (213/237 cases),
86.2% of BC (56/65, enriched for triple negative (TN) cases), 48.7% of NSCLC
(734/1508) and 12.9% (42/326) of AML. Multiple mutations in TP53 were highest in
TNBC (27/45; 60%, with 3-5 calls seen in 26.7%) and estrogen receptor (ER)+BC
(7/13; 53.8%), seen in 19% of TP53-mutated AML, and uncommon in CRC (3.8%)
and NSCLC (2.4%). Frequency of cis-mutations was at least 7% but not fully
evaluable. Dual-mutated cases with 2 well-characterized LOF mutations (e.g.
R282W and R248W) typically showed concordant VAFs. Tumors with at least one
uncommon ms mutation/VUS showed the widest differences in VAF between cooccurring variants. This unequal VAF group, which was much more common in
TNBC or ER+BC, included a high proportion of mutations/VUS (26 in TNBC and 7 in
ER+BC versus 6 in all others) in the transactivation, proline-rich and oligomerization
domains or uncommon VUS within the DNA-binding domain; nearly all scored as
likely functional with the RF classifier. Conclusions: Complex, subclonal and
uncommon TP53 variant patterns are noted in a subset of breast cancer but are not
restricted to TNBC. The high frequency of non-LOF variants suggests that weak
selective effects on TP53 function (possibly related to complementing mutations in
other DNA repair genes) are operating as one mechanism producing tumor-specific
mutation biases apparent in the public databases. A more sophisticated scoring of
somatic TP53 variants in BC distinguishing atypical versus typical/LOF mutation
patterns is suggested.
ST33. Low Cost Liquid Biopsy Combining Hotspot Mutant DNA Enrichment
with Cost Effective Duplex Sequencing
D. Broemeling1, J. Pel1, M. Wiggin1, W. Choi1, P. Davies1, M. Despotovic1, L.
Gelinas1, A. Leung1, L. Mai1, G. Shibahara1, L. Ung1, S. Walsh1, A. Marziali2
1Boreal Genomics, Vancouver, British Columbia, Canada; 2University of British
Columbia, Vancouver, British Columbia, Canada.
Introduction: Liquid biopsy assays require high sensitivity, specificity, multiplexing
and low cost. Achieving these simultaneously is challenging, and most approaches
are limited by either low multiplexing (droplet digital PCR, ddPCR), reduced
specificity, or high cost (barcoded sequencing). Methods: Since sequencing reads
are often wasted re-sequencing wild-type cell free DNA (cfDNA) fragments, a
powerful solution to improve cost, sensitivity, and specificity is to enrich cfDNA
fragments from wild-type cfDNA so only tumor strands (ctDNA) are sequenced.
Given that in many clinical cases the ctDNA represents < 0.1% of the cfDNA, this
approach can reduce required sequencing up to 10,000 fold, dramatically reducing
cost and false positive errors. OnTarget is able to perform such enrichment on
panels of over 100 mutations while retaining quantitation of mutant alleles through
use of internal controls. To complement OnTarget over regions without hotspot
mutations, we apply our Proximity Sequencing (Pro-Seq), a duplex library
construction method for NGS that achieves comparable performance to duplex
barcoding methods, but at much lower sequencing cost. Results: We report data
from analytical validation studies using OnTarget enrichment from cell line and
plasma samples, demonstrating low limits of detection and showing mutation
detection is improved over ddPCR. We present data from clinical samples including
cancer patients and healthy individuals, as a precursor to clinical studies for
population-wide early cancer detection. Using cfDNA and cell-line DNA, we report
average per-base error rate and per-base analytical specificity of Pro-Seq to be
0.0002% and >99.9997% respectively, demonstrating that Pro-Seq is among the
highest performing liquid biopsy assays in terms of sensitivity and specificity, but with
greatly reduced sequencing costs compared to existing methods. Conclusions: In
summary, OnTarget mutation enrichment and Pro-Seq duplex NGS enable high
sensitivity, and high specificity liquid biopsy tests to be offered at low cost, enabling
commercialization in applications with limited reimbursement.
ST34. Anaplastic Lymphoma Kinase (ALK) Mutation Testing for Pediatric
Neuroblastic Tumors: A Single Institution Experience
T. Qdaisat1,2, S.F. Sarabia1, C.R. Webb2, K. Alvarez2, L. Suarez Ferguson2, C.R.
Perez2, J.H. Foster1, D.H. Lopez-Terrada1, A. Roy1, K.E. Fisher1,2
1Baylor College of Medicine, Houston, TX; 2Texas Children's Hospital, Houston, TX.
Introduction: In neuroblastoma, oncogenic activation of the anaplastic lymphoma
kinase (ALK) receptor tyrosine kinase occurs through somatic ALK (NM_004304.4)
gene amplification or secondary to gain of function mutations in the tyrosine kinase
domain (TKD, exons 21-28, amino acids 1122-1376). Patients with
acquired ALK alterations have a less favorable prognosis and are candidates for
targeted therapy, while deleterious germline variants in the ALK TKD are the
predominant cause of familial neuroblastoma. We test all newly diagnosed and
relapsed neuroblastic tumors for ALK amplification and TKD mutations, and
investigated the efficacy of our algorithm to detect clinically
actionable ALK alterations. Methods: We retrospectively searched laboratory
records for neuroblastic tumors tested for TKD mutations using Sanger sequencing
(n=113) and ALK amplification by either fluorescent in situ hybridization (FISH, n=63)
1015
AMP Abstracts
or microarray (OncoScan, n=5) from January 2009 to April 2017. Clinicopathologic
data including MYCN amplification status (n=84) were collected from corresponding
pathology reports. Results: Data from 113 neuroblastic tumors (99 neuroblastomas,
10 ganglioneuroblastomas, 4 ganglioneuromas) collected from 91 patients (52M,
39F; ages 2d to 18.3y, median 2.5y) from 9 institutions were included in the study.
All 91 patients were tested for ALK TKD mutations, 22 patients underwent
repeat ALK TKD testing on a second sample, and 1 patient had 3 samples tested.
Non-synonymous ALK TKD mutations [p.R1275Q (n=3), p.F1174L, p.P1245I, and
p.I1268M] were detected in 6 initial samples (6.6%; 5 biopsies, 1 excision), and 2
patients underwent repeat testing on subsequent samples. A known p.F1174L
mutation was detected again in a follow-up resection, but a previously detected
variant of uncertain significance (p.I1268M) was absent in a subsequent specimen
with <20% viable tumor. One ALK p.P1245I mutation was confirmed as a germline
variant and no patients received targeted therapy. ALK amplification data were
available for 68 initial (74.7%) and 17 repeat (77.2%) samples. MYCN amplification
was detected in 15.5% of samples (13/84) but no samples
demonstrated ALK amplification. Conclusions: Our data suggest
that ALK alterations are infrequent events, and conclude that reflex somatic testing
at diagnosis may not be necessary for all patients. ALK TKD testing can be reserved
for high-risk or relapsed neuroblastoma patients that are candidates for targeted ALK
inhibitor therapy or clinical trials, and germline ALK testing should be at the
discretion of the clinical provider. OncoScan microarray simultaneously assesses
for ALK and MYCN amplification, tumor ploidy, and chromosomal gains and losses,
and is an effective replacement for standalone ALK FISH.
ST35. EGFR Amplification as a Biomarker of Shorter Overall Survival in Grade
III Gliomas
T. Bale1, J.C. DeWitt2, D.N. Louis2, A.J. Iafrate2, J.K. Lennerz2
1Harvard Medical School, Boston, MA; 2Massachusetts General Hospital and
Harvard Medical School, Boston, MA.
Introduction: The 2016 WHO classification of CNS Tumors redefines specific
glioma entities based on molecular characteristics to optimize prognostication and
treatment. While amplification of EGFR has long been recognized as a frequent
aberration in Glioblastoma, WHO Grade IV, it is also reported in histologic WHO
grade II-III gliomas. Currently conflicting evidence exists regarding its prognostic
significance, particularly in the context of additional relevant molecular genetic
markers (IDH mutations, MGMT promoter methylation, 1p19q co-deletion). It has
been suggested that IDH-wildtype gliomas can be reclassified as other entities, yet it
remains unclear if EGFR amplification is a sufficient, independent surrogate for
progression to GBM. Here, we assessed the role of EGFR amplification in a
molecularly annotated cohort of Grade II-III gliomas with diagnostic imaging and
clinical follow-up. Methods: Seventy-one Grade II-III gliomas where EGFR copy
number status was assessed by FISH and/or next-generation sequencing
(SnapShot) were retrospectively identified Survival by EGFR status was compared
with established clinico-pathological features (age, histology, grade, contrast
enhancement on MRI at diagnosis IDH status, MGMT methylation, 1p19q codeletion) as noted in the clinical record. Results: Of the 71 tumors, 20 were EGFRamplified (28%). All 20 of the EGFR-amplified cases met histologic criteria for Grade
III, as compared with38/51 in the non-amplified subgroup (75%, P=0.014). EGFRamplified tumors occurred on average 5 years later (46 versus 41 years; P<0.001).
EGFR-amplification occurred independently of MGMT promoter methylation, and
was mutually exclusive with co-deletion of chromosomes 1p and 19q. EGFR
amplification was frequently associated with wild-type IDH mutation status (19/20
cases). The majority (70%) of the EGFR amplified tumors demonstrated contrast
enhancement on MRI at time of diagnosis, as did a large percentage of nonamplified tumors (45%). Outcome analysis showed significantly shorter overall
survival (P<0.0001) although multivariate analysis did not establish EGFRamplification as an independent prognostic indicator. Conclusions: EGFR
amplification is associated with significantly shorter overall survival in grade III
gliomas and is associated with established poor prognostic molecular and clinical
indicators. Our findings emphasize that IDH wild-type gliomas represent an
aggressive tumor subgroup, encompassing the majority of EGFR-amplified cases.
Our findings demonstrate that the majority of EGFR amplified grade III gliomas may
be considered undersampled GBMs, and lend further support to suggestions to
reclassify IDH-wild type high grade astrocytomas as GBM, based on molecular
features.
ST36. Papillary Renal Cell Carcinoma Associated with Bi-Allelic SDHA
Mutations
C.R. McEvoy1, L. Koe2, D.Y. Choong1, H.S. Leong1, H. Xu1, O.W. Prall1, D.
Karikios3, A.P. Fellowes1, S.B. Fox1
1Peter MacCallum Cancer Cantre, Melbourne, Victoria, Australia; 2New South Wales
Health Pathology, Sydney, New South Wales, Australia; 3Nepean Hospital, Sydney,
New South Wales, Australia.
Introduction: A 45-year-old male with Stage III papillary renal cell carcinoma (RCC)
progressed to Stage IV disease with peritoneal, hepatic, and pulmonary metastases,
over 7 months. The patient was referred for comprehensive genetic profiling of both
tumor and blood DNA using next-generation sequencing to look for a druggable
target. Methods: Analysis comprised sequencing of 2.34 Mb of genomic regions
implicated in cancer, including the entire coding region of 391 genes and 18 genes
1016
specifically implicated in renal neoplasia. Kappa Hyper libraries were prepared and
target enriched using SureSelectXT. Pooled library pairs were sequenced at
500x/100x mean coverage (tumor/blood) on an Illumina NextSeq sequencer using
paired 75bp reads. Variant calling, copy number estimation (based on off-target and
on-target coverage), gene fusion analysis, and mutational signatures were
determined. Results: Two variants in succinate dehydrogenase (SDH) A were
detected: a germline truncating variant, c.91C>T (p.Arg31*), and a somatic missense
variant, c.1765C>T (p.Arg589Trp). Copy number gains of 2p, 7p (including EGFR),
12p and 17, and losses of 9, 13q and 15q were detected. No gene fusions or
mutational signatures of clinical significance were observed. SDHA loss in the tumor
was confirmed by immunohistochemistry. SDH converts succinate to fumarate in the
Krebs cycle. It comprises 4 subunits, SDHA–D, each transcribed by a separate
gene. SDH-deficient RCC is a rare RCC subtype (estimated frequency 0.05%–0.2%)
and bi-allelic SDHA mutations are exceedingly rare even within this subtype. To our
knowledge, this is the first report of an SDH-deficient RCC caused by bi-allelic
germline and somatic SDHA mutations. The patient had previously failed to
respond to anti-PD-1 immunotherapy but as the genetic results predicted a likely
response to tyrosine kinase inhibitors (TKI) these were administered. Although
positive responses were observed for TKIs adverse reactions necessitated treatment
cessation. Currently, the patient has self-initiated a low carbohydrate diet designed
to abrogate increased hypoxia factor induction caused by SDH accumulation.
Interestingly, the patient has reported significant subjective improvement in clinical
fatigue. Conclusions: This is the first report of SDH-deficient RCC caused by biallelic germline and somatic SDHA mutations. The genetic results have been
perceived as beneficial by the patient as they have increased knowledge and
understanding of his condition, confirmed the appropriateness of the treatment
choices made by his oncologist, and enabled active participation in clinical
management through diet & lifestyle modification.
ST37. Molecular Profiling with ALK, ROS1 and MET Genes FISH Panel in NonSmall Cell Lung Cancers: Indian Tertiary Cancer Institutional Experience
O.A. Shetty, T.D. Pai, R. Kumar, A. Singh, S. Dhanavade, S. Kane, K. Prabhash, S.
Desai
Tata Memorial Hospital, Mumbai Maharashtra, India.
Introduction: In this era of precision medicine, identification of genetic alterations,
like activating mutations in EGFR, fusion genes involving ALK, rearrangements
in ROS1, and amplification of MET gene has revolutionized the management of nonsmall cell lung cancer (NSCLC) patients with the tyrosine kinase inhibitors. Here in
this study, the authors present their experience of these biomarkers in NSCLCs over
a period of 2013-2015. Methods: All EGFR mutation negative NSCLCs were
evaluated for ALK and ROS1 gene rearrangements and MET gene amplification
(n=182). Immunohistochemistry (IHC) for ALK using Anti ALK antibody (D5F3 clone,
Ventana, USA) was used to screen the equivocal or positive cases for reflex testing
by fluorescent in-situ hybridization (FISH). FISH was performed on sections from
formalin fixed paraffin embedded (FFPE) tissue using ALK (VYSIS dual colour Break
apart probe, Abbott, USA), ROS1 (Zytolight SPEC Dual Colour break apart probe,
Zytovision, Germany) and MET gene amplification (Zytolight MET/CEN7 dual colour
probe, Zytovision Germany) probes. Interpretation of the FISH results was done in
accordance to the CAP guidelines. Results: The patient’s age ranged from 27-82
years with median age of 56 years, male to female ratio 1.2:1. Site of biopsies
included lung (57.6%), nodal metastasis in supraclavicular lymph node (22%),
pleural fluid (8.5%), axillary nodes (4.2%) and others (7.6%). ALK gene
rearrangement was detected in 24/182 cases (16.2%).The pattern of the ALK FISH
signals were as follows: Combined split and loss of 5’ centromeric end of ALK gene
signals (28.2%), split signals (17.3%), only 5’ centromeric end of ALK gene loss
(9.0%) and polysomy was detected in 5.1% cases. ROS1 gene rearrangement was
detected in 8/182 cases (5.6%). FISH signal patterns were as follows: Combined
split and loss of 3’ centromeric end of ROS1 (16.2%), only split (14.9%) only 3’
centromeric end of ROS1 loss (9.0%) and polysomy in 1.4% of the cases. MET gene
amplification was detected in 1 case (0.5%) with an average 4.27 MET signals /
average 2.12 CEN7 signals [Ratio 2.02].Majority of the cases were in advanced
stage (III/IV) at presentation. Of the total gene rearranged cases
(n=24 ALK; n=8 ROS1) Crizotinib was administered to 21 patients, Gefitinib and
Erlotinib to each one patient. With a median follow up of 6 months of the patients
taking TKI, 2 patients expired due to progressive disease; one had partial response
and 18 patients had stable disease. Conclusions: The prevalence of ALK,
ROS1 gene rearrangement, MET gene amplification was 16.2%, 5.6% and 0.5%
respectively. All 3 genetic alterations were mutually exclusive. FISH Lung Cancer
Diagnostic panel is an invaluable tool in routine clinical management of NSCLC
patients.
ST38. Real Time PCR Assessment of Actionable Mutations in Non-Small Cell
Lung Cancer
A. Atkinson1, R. Zhang2, J.D. Paterson3, F.B. de Abreu3, Y. Yuan4, M. Lei2, G.J.
Tsongalis3
1Dartmouth Hitchcock Medical Center, Lebanon, NH; 2QuanDx Inc., San Jose,
CA; 3Dartmouth Hitchcock Medical Center and Geisel School of Medicine at
Dartmouth, Lebanon, NH.
Introduction: Of cancer-related mortality, lung cancer accounts for most with over
220,000 diagnosed annually in the United States alone. Of these patients, there is
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
an average 5 year survival rate of only 17.7%, with more than 80% of these
diagnoses being non-small-cell lung cancer (NSCLC). Molecular diagnostics have
become a critical prerequisite and monitoring tool in attempt to improve both early
detection and the poor prognosis of NSCLC. Predictive biomarkers, such as ALK
fusion genes and stereotypical EGFR, KRAS, and BRAF mutations are fundamental
to patient outcomes in driving therapeutic actions, as many treatments are biomarker
specific. Molecular detection of these biomarkers is complicated by factors such as
the pressure to rapidly determine actionable mutations for specific first-line therapies,
biopsies of limited quantity, and fixation artifacts. To avoid patient risk exposure and
payer expense while addressing rising and evolving clinical needs, we evaluated the
Lung Cancer EGFR/KRAS/BRAF Mutation Detection Panel from
QuanDx. Methods: Samples were formalin fixed paraffin embedded (FFPE) tissues
from 16 patient tumors previously tested with our 50-gene Ion Ampliseq Cancer
Hotspot Panel v2 and 13 commercial cell lines. Total nucleic acids were extracted
using Qiagen AllPrep DNA/RNA FFPE extractions and quantified by QuBit HS RNA
and DNA Assays. The Lung Cancer Panel consists of 5 separate reactions
assessing an internal control and 57 separate mutations in BRAF (V600E/K), KRAS
(G12A/V/S/D/R/C, G13D/C), or EGFR [exon 19 LREA deletions, exon 20 in-frame
insertions and point mutations in exons 21 (L858R, L858Q), 20 (T790M, S768I), and
18 (G719A/S/C)]. DNA inputs of 10 ng or 6.6 ng per tube were assessed. Reactions
were run in duplicate per the manufacturer’s protocol on an Applied
Biosystems 7500 instrument. For analysis, baseline values were set manually and
cycle threshold (Ct) values compared to values set by QuanDx. Results: Diagnostic
Ct values for the NSCLC EGFR/KRAS/BRAF Detection Panel gave genotype
specific results matching the known results of commercial cell lines and our Cancer
Hotspot Panel. Complete concordance across 13 cell lines and 16 patient samples at
both 10 and 6.6 ng DNA concentrations was observed. Conclusions: The low DNA
input, accuracy, and rapid 2.5-hour result time makes this assay a valuable tool in
the assessment of NSCLC with less labor as compared to a lab developed
SNaPshot Lung Panel for a few common EGFR/KRAS/BRAF mutations with a
workflow requiring 1 day with 4 sequential reactions prior to fragment analysis. In
comparison, the NSCLC Panel from QuanDx could reduce our workflow without
sacrificing sensitivity or DNA availability, especially for STAT biomarker
determination.
ST39. Integrated Genomic Profiling in Pediatric Solid Tumors: An Institutional
Experience
L.F. Surrey1, S.P. MacFarland1, G.T. Akgumus1, D.J. Gallo1, F. Lin1, F. Chang2, M.M.
Li1
1The Children's Hospital of Philadelphia, Philadelphia, PA; 2WuXiNextCODE,
Cambridge, MA.
Introduction: Pediatric solid tumors are a rare and diverse set of diseases that
contain a variety of genomic alterations including single nucleotide variants (SNVs),
insertions/deletions (indels), copy number variations (CNVs), and gene fusions.
Accurate identification of recurrent and novel alterations is critical for diagnosis,
prognosis, and therapeutic decision making in these patients.
Methods: Our laboratory developed a comprehensive next-generation sequencing
(NGS) panel that interrogates 237 solid tumor related cancer genes and 106 major
fusion partner genes associated with 586 known cancer fusions and many novel
fusions. The DNA-based sequencing portion uses Agilent SureSelectQTX target
enrichment capture-based libraries while the RNAseq-based fusion portion uses
Archer technology with custom designed primers. Sequencing is performed on
Illumina MiSeq platform. Copy number analysis is performed using SoftGenetics
NextGENe software. Data was collected on non-central nervous system (CNS) solid
tumors clinically tested in our laboratory between 1/2016 and 3/2017.
Results: During a 15-month period, 147 solid tumor samples were tested from 141
different patients with an average age of 9 years (range 0-49, 134 cases <19 years).
There were 42 sarcomas (29%), 37 neuroblastic tumors (25%), 26 carcinomas
(18%), 16 embryonal tumors (11%), 8 germ cell tumors (5%), and 18 other tumors
(12%). Clinically significant tier 1 or 2 alterations of any type (SNV, indel, CNV,
fusion) were identified in 130 (88%) cases. A total of 63 SNVs and 25 indels were
reported in 47 genes, with TP53 being the most frequent (9%), followed
by ALK, KRAS, and BRCA2 (3% each). Over 934 CNVs were reported in 100 cases
(68%). The most frequent CNVs in >5% of cases were chromosome 11 and 17
abnormalities (loss or copy neutral loss of heterozygosity), MYCN amplification, and
near triploid genome. Fusion testing was performed on 127 cases with 40 fusions
identified (32%). Half (50%) of sarcomas and 65% of carcinomas were fusionpositive. Nine fusions were novel or had only rare reports in the literature. This
genomic information has significantly impacted patient care by leading to potential
alterations in therapy, based on risk and/or targeted therapy, in 61 cases (41%) and
changing histologic diagnosis for 4 cases (3%). Conclusions: These results
demonstrate that integrated genomic profiling in non-CNS pediatric solid tumors has
a high rate of detecting clinically significant variants and can result in changes to
therapy and diagnosis in a subset of patients. The comprehensive NGS solid tumor
panel is fully implemented as an essential component of patient care for pediatric
solid tumors in our institution.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
ST40. Factors that Predict the Success of RNA Seq Analysis on Solid and
Hematologic Tumor Specimens
R.N. Wehrs, J.L. Winters, J.I. Davila, N. Fadra, A.A. Nair, A.M. McDonald, X.
Wu, J.S. Voss, L. Jin, R.A. Jackson, K. Rumilla, B.R. Kipp, R.B. Jenkins, J. Jen, K.C.
Halling
Mayo Clinic, Rochester, MN.
Introduction: RNA Seq may be helpful in establishing diagnosis, determining
prognosis, and guiding targeted therapy in patients. In this study, we evaluated the
effect of various factors (e.g. RNA input, RIN value, library yield) for the success of
RNA Seq analysis. Methods: Total RNA was extracted from solid tumor tissue
(fresh, frozen, and cell lines) or blood using a Qiagen miRNeasy Micro and Mini kit,
respectively. RNA quality was assessed via RNA Integrity Number (RIN) using an
Agilent 2100 Bioanalyzer and quantified using a Qubit 2.0 Fluorometer. mRNA
isolation by poly-A tail selection and cDNA synthesis were completed using the
TruSeq RNASample Preparation v2 Kit (Illumina) on a Janus Automated Workstation
(Perkin Elmer). Libraries were prepared with a Biomek FXp Liquid Handler (Beckman
Coulter) using custom-built protocols. Post-analytic library yield was assessed prior
to sequencing via the Agilent 2100 Bioanalyzer and Qubit 3.0 Fluorometer. Pairedend, 101 bp-read sequencing was completed on a HiSeq 2500 (Illumina) in Rapid
Run mode. Data was analyzed using MapRSeq, a Mayo Clinic-developed suite of
alignment, fusion detection and filtering programs to detect fusions on a pre-selected
set of 573 genes and associated transcripts. An assessment of the depth of
coverage at different distances from the 3’ was also used. A select set of samples
were used for minimum total RNA input testing. Results: A total of 124/135 (91.9%)
samples met the quality threshold (≥2nM) for library prep and were sent onto
sequencing, of which 119 (96.0%) samples met the sequencing quality threshold
(≥50 million) for mapped reads. RIN value weakly correlated to library prep yield
(R²=0.2) and did not correlate to mapped reads (R²=0.03). Library yields did not
predict the number of mapped reads (R²=0.04). Total RNA input above 62.5 ng
showed a strong correlation to passing library yield (R²=0.9) and moderate
correlation to mapped reads (R²=0.5) but variable correlation with fusion-supporting
read (R²=0.07 to 0.49). Tumor percent did not correlate with fusion-supporting read
count (R²=3E-05). Conclusions: While sample RIN value provides an estimate of
the level of degradation of RNA, the RIN value imprecisely predicts the degree of
sample RNA degradation for our RNA Seq assay. A RIN value threshold of ≥5 and a
library yield threshold of ≥2nM can be used as a first measure to ensure acceptable
levels of degradation, while a mapped reads threshold of ≥50 million can be a useful
metric after sequencing. The degree of RNA degradation can be more precisely
assessed by using the estimated sensitivity at different distances from the 3’end. The
depth of coverage at the exon boundaries can be also used to determine the degree
of degradation across particular genes.
ST41. HPV Genotyping of Solid Tumors Using Real-Time PCR and Multi-Color
Melt Curve AnalysisHPV Genotyping of Solid Tumors Using Real-Time PCR
and Multi-Color Melt Curve Analysis
A. Atkinson1, C.M. Studwell2, L.J. Tafe3, G.J. Tsongalis3
1Dartmouth Hitchcock Medical Center, Lebanon, NH; 2University of Connecticut,
Storrs, CT; 3Dartmouth Hitchcock Medical Center and Geisel School of Medicine at
Dartmouth, Lebanon, NH.
Introduction: Oncogenic Human Papillomavirus (HPV) genotypes are linked to over
600,000 cancer cases worldwide. While the role of HPV in nearly all cervical cancers
is established and screening methods are in place, the status of HPV in other
cancers has only recently become apparent. Oral HPV infection confers ~50% fold
increased risk for oropharyngeal squamous cell carcinomas (OPSCCs). HPVpositive OPSCCs and other anogenital cancers have a more favorable prognosis
and response to therapy than similar HPV-negative cancers. As such, HPV-status
has significant prognostic and treatment implications. Being able to accurately
identify HPV genotypes in patients is of critical importance. To this end, we validated
the use the MeltPro High Risk and Low Risk HPV Genotyping Kits from QuanDx to
assess HPV status on OPSCCs and other solid tumors relative to the Roche Linear
Array. Methods: DNA from 29 formalin fixed paraffin embedded (FFPE) tissues was
isolated using the Qiagen Gentra Tissue Kit. Samples previously tested on the
Roche Linear Array were diluted in water to 25 �L for total DNA ranges from 5 to
1000 ng. Samples were then added to lyophilized High-Risk or Low-Risk MeltPro
Kits and run on a SLAN-96 real-time PCR instrument (QuanDx/Zeesan Biotech, San
Jose, CA) per kit instructions over 2.5 hours. Instrument software provided sample
genotypes via multi-color melting curve analysis (MMCA); these results were then
compared to those obtained from the Roche Linear Array. Results: For all 44
reactions from 11 samples run on the Low Risk HPV MeltPro Kit there was complete
concordance between assays across all DNA dilutions. The MMCA was well within
temperature ranges set by the manufacturer and HPV6 and HPV11 were
distinguished from all negative and high risk samples. For 45 reactions from 18
samples, the High Risk HPV MeltPro Kit was 97.8% concordant; however, upon
repeat of this sample, the correct results were obtained for all samples. The QuanDX
kits correctly called all negative samples, low risk HPV6 and HPV11 together with
high risk HPV genotypes: 16, 18, 33, 35, 45, 51, 52, and 59. Conclusions: The
QuanDx High-Risk and Low-Risk MeltPro Kits offer an accurate genotyping solution
relative to the Roche Linear Array. In comparison of the workflows between the 2
assays, the MeltPro Kits require half the hands on time of 30 minutes with shorter
turnaround times of 2.5 versus 6 hours for the Linear Array.
1017
AMP Abstracts
ST42. Validation of a Low DNA Input Hotspot Solid Tumor Assay on the Agena
Bioscience MassARRAY System Utilizing Reference Standards and FFPEDerived Clinical Samples
T. Neuwerth, M. Buckway, P. Bradley, D. Loughmiller, C. Johnson
Intermountain Precision Genomics, St. George, UT.
Introduction: Frequently, FFPE tissues from solid tumor malignancies have low
DNA extraction yields leading to cancelation of testing or prioritization of other testing
methods in place of NGS (especially in NSCLC). We validated a 5-gene hotspot
panel from Agena Bioscience as an LDT for use on solid tumor FFPE samples.
Previously tested, known positive and negative clinical samples and synthetic
reference standards were utilized for validation, in a CLIA certified and CAP
accredited laboratory. Methods: The iPLEX Pro chemistry based assay comprises
multiplex PCR and single nucleotide extension with genotype detection by MALDITOF Mass Spectrometry on the MassARRAY System. The assay requires 10-20 ng
of input DNA and evaluates mutational hotspots in 5 genes commonly mutated in
solid tumor malignancies: BRAF (codons 469, 594, 500, 600,
601), EGFR (substitutions and indels in exons 18, 19, 20, 21), KRAS (codons 12, 13,
61, 146), NRAS (codons 12, 13, 61), and KIT (substitutions in exons 11, 13, 17). We
used SeraCare Solid Tumor Mutation Mix-II as reference standards; each contained
5 substitutions and 2 indels detectable by the assay in a background of GM24385
cell line DNA at 10%, 7% or 4% Allelic Frequency(AF). Results: We assessed 30
replicates of AF10% to determine performance characteristics. Sensitivity was
92.55%, specificity was 99.94%, positive predictive value (PPV) was 98.29%, and
negative predictive accuracy (NPA) was 99.67%. We assessed 10 replicates each of
AF10%, 7%, and 4% to evaluate the limit of detection. The assay reliably detects
variants at 10% AF (49/50 variants detected), and often detect variants at lower AF
with less precision 7% AF (40/50 variants detected), and 4% AF (10/50 variants
detected). We established the limit of detection for the assay at 10% AF for
substitutions and indels. We assessed 21 known positive or known negative clinical
samples (FFPE tissues) initially characterized by a hybrid capture NGS assay
(targeted exome), sequenced on an Illumina instrument. Our overall concordance
was 80.85% (38/47 concordant and 9/47 discordant variants). Conclusions: The
Agena Bioscience MassARRAY system is a robust method for detection of somatic
variants in low yield samples. The assay can reduce the DNA input requirement for
multi-gene/targeted sequencing especially when compared to NGS and provide
additional options to assess somatic variation when FFPE tumor tissue is scant or
other testing is prioritized.
ST43. Validation of an Anchored Multiplex PCR-Based Next Generation
Sequencing Assay for the Detection of MET Exon 14 Skipping
K.D. Davies, J. Haney, M. Seager, D.L. Aisner
University of Colorado Anschutz Medical Campus, Aurora, CO.
Introduction: Mutations and deletions in MET that result in aberrant splicing that
excludes exon 14 from the transcript have recently been characterized in several
cancer types, including, most prominently, lung cancer. Several recent studies have
reported dramatic responses to HGFR-directed therapy in lung cancer patients
positive for these genetic abnormalities, and clinical trials are now underway that are
further testing this paradigm. Therefore, the ability to detect MET exon 14 skipping
in clinical samples is critical. However, many of the reported genomic deletions that
result in exon 14 skipping involve large regions, and large indels can be difficult to
detect via DNA-based sequencing techniques. An alternative approach is to instead
test RNA, whereby, regardless of the underlying genomic aberration, exon 14
skipping is simply observed as ‘fusion’ of exon 13 to exon 15. In this study, we
validated the Archer FusionPlex Solid Tumor assay, an RNA-based anchored
multiplex PCR approach, for detection of MET exon 14 skipping. Methods: Total
nucleic acid (TNA) was extracted from FFPE processed tumor samples. Libraries
were created from the TNA via the Archer FusionPlex assay per the manufacturer’s
instructions with minor modifications. Libraries were sequenced on the Illumina
MiSeq using v3 chemistry. Confirmation of MET exon 14 skipping mutations was
achieved using DNA-based next-generation sequencing assays. Results: Our
validation cohort consisted of 5 samples known to be positive for mutations expected
to result in exon 14 skipping. Four of these samples passed assay quality control
(QC) metrics, and all 4 clearly demonstrated sequencing reads
supporting MET transcripts that lacked exon 14. The fifth sample did not
demonstrate exon 14 skipping, but this sample failed QC and was deemed
uninformative due to suboptimal RNA quality. However, a re-extraction of a different
sample from the same patient was positive for exon 14 skipping via the assay. No
unexpected MET exon 14 skipping events were observed. Thus, in the validation
cohort, the assay demonstrated 100% sensitivity and specificity for samples with
acceptable QC metrics. Following validation, routine clinical testing has revealed 5
additional positive cases to date. Importantly, for 2 of these cases, an orthogonal
DNA-based assay did not reveal exon 14 skipping related mutations. Conclusions:
The Archer FusionPlex assay exhibits good performance in terms of the ability to
detect MET exon 14 skipping. Use of an RNA-based assay may be preferable to
DNA-based assays due to the prevalence of difficult to detect genomic deletions that
result in exon 14 skipping. RNA-based assays require stringent QC metrics to
identify cases in which negative findings are uninformative.
1018
ST44. TERT Promoter Mutation Status in Morphological Variants of Urothelial
Carcinoma
D. Pradhan, W.L. Ernst, S. Mercurio, M. Barbi De Moura, M. Nikiforova, R. Dhir, S.
Roy
University of Pittsburgh Medical Center, Pittsburgh, PA.
Introduction: Telomerase is deactivated in most of the somatic cells due to
suppression of the Telomerase reverse transcriptase (TERT) gene after
differentiation. However, point mutations in the promoter region of the TERT gene
that creates de novo binding sites for E-twenty-six/ternary complex (ETS/TCF)
transcription factors causing cancer-specific telomerase reactivation are the most
common genetic alterations in urothelial carcinoma (UC) of the bladder. The aim of
this study was to evaluate the prevalence of TERT promoter mutations (TPM) in
different variants of UC. Methods: The TERT promoter was amplified by PCR on
extracted DNA and interrogated for the hotspot mutation loci for g.1295228C>T (c.124C>T; C228T) and g.1295250C>T (c.-146C>T; C250T) using bi-directional
Sanger sequencing. Results: The following variants constituted our cohort of 74
UCs; squamous differentiation (n=20), high grade papillary UC (n=10), micropapillary
(n=9), small cell (n=7), plasmacytoid (n=4), nested (n=1), primary adenocarcinoma
(n=1), and high grade UC, not otherwise specified (HGUC NOS, n=22). The median
age of the patients was 71 years and male to female ratio of 2.9:1. Overall 56 of 74
(75.7%) urothelial carcinomas harbored TPM. All cases of micropapillary, small cell
and nested variants harbored TPM and were present at lesser frequency in papillary,
plasmacytoid and squamous differentiation (Table 1). HGUC NOS revealed TPM in
almost half of the cases and a lone primary adenocarcinoma was negative for TPM.
TPM C228T was most common mutation followed by C250T mutation. C250T
mutation was seen in UC with squamous differentiation (n=3) and one in each case
of HGUC NOS, papillary, micropapillary and small cell variant. TPM C228T was
observed in all the variants in varying proportion, except adenocarcinoma. There
was no significant difference in overall survival and disease specific survival based
on TPM status. Conclusions: TPM was highly prevalent in micropapillary, small cell
and nested variant and was less prevalent in UC with papillary, plasmacytoid and
squamous differentiation. Almost half of HGUC NOS harbor TPM. TPM status can
be potentially used in a clinical diagnostic setting (e.g., urine cytology, cystoscopic
biopsy).
Table 1:
UC Variants
# Cases
Micropapillary
9
Small Cell
7
Nested
1
Papillary HGUC
10
Plasmacytoid
4
Squamous Differentiation 20
HGUC-NOS
22
Adenocarcinoma
1
Total
74
TPM
9 (100%)
7 (100%)
1 (100%)
9 (90%)
3 (75%)
15 (75%)
12 (54.5%)
0 (0%)
56 (75.7%)
C250T
C228T
1
8
1
6
1
1
8
3
3
12
1
11
7/56 (12.5%) 49/56 (87.5%)
ST45. Development of a Breast and Lung Cancer Research Panel To Target
Therapeutically Relevant Copy Number and Gene Fusion Variants from Blood
J. Schageman, V. Bagai, K. Lea, P. Kshatriya, J. Gu, K. Bramlett
Thermo Fisher Scientific, Austin, TX.
Introduction: With recent advances in next-generation sequencing (NGS)
technologies, it is now possible to detect somatic mutations with allele frequencies in
blood samples as low as 0.1% from circulating tumor DNA. A natural extension to
this achievement is adding the ability to simultaneously detect copy number variants
and gene fusions. A panel such as this addresses a full repertoire of variant classes
found to be linked with certain tumors and would enable researchers additional to
profile cancer samples more dynamically thus enriching current diagnostic tool sets.
Here, we present progress on such an approach and apply current NGS technology
to achieve our goals. Results: Using control samples, we can reproducibly
demonstrate detection of ERBB2 (HER2/neu) and FGFR1 gene amplifications with
high statistical significance and as low as a 1.4 fold difference versus non-amplified
loci in titration experiments. In addition, the FGFR1 gene amplification was detected
in the context of a validated breast cancer somatic mutation panel in which no
negative impact was exhibited and mutation detection specificity and sensitivity were
both greater 90%. Lastly, we developed an additional panel to detect gene fusions
relevant to lung cancer. Using the titration approach above, the EML4-ALK fusion
variant was shown to have a limit of detection near 1% with no negative impact on
detection sensitivity and sensitivity when combined with the validated lung cfDNA
somatic mutation panel with a 0.1% limit of detection. Conclusions: From the
outcomes of these experiments, we have shown the ability to reproducibly and
simultaneously detect copy number and gene fusion variants as well as somatic
mutations at very low limits of detection in a cell free DNA background derived from
blood samples.
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
ST46. Targeted Mutational Analysis of Predictive and Prognostic Biomarkers in
Colorectal Carcinoma
A.M. Olofson, X. Liu, S.J. Deharvengt, F.B. de Abreu, J.D. Peterson, A.A.
Suriawinata, M. Lisovsky, G.J. Tsongalis
Dartmouth Hitchcock Medical Center and Geisel School of Medicine at Dartmouth,
Lebanon, NH.
Introduction: Recent guidelines outlining the application of molecular testing to the
evaluation of colorectal cancer have recommended mutational analysis of the
genes KRAS, NRAS, and BRAF p.V600. Monoclonal antibody therapies targeting
the epidermal growth factor receptor have become a mainstay in the treatment of
colorectal cancer. Mutations in RAS codons 12 and 13 of exon 2, codons 59 and 61
of exon 3, and codons 117 and 146 of exon 4 can serve as predictive biomarkers for
anti-EGFR antibody therapies. BRAF p.V600 mutational analysis can provide
information relevant to prognostic stratification. Our institution currently employs a 50
gene hotspot panel for the mutational analysis of all colorectal carcinomas. The goal
of this retrospective study was to evaluate the occurrence of the aforementioned
mutations in comparison to other mutations included in our multi-gene
panel. Methods: Tumor specimens from patients diagnosed with colorectal
carcinoma between the years of 2012 and 2017 were selected for analysis. Genomic
DNA was extracted from formalin-fixed, paraffin-embedded (FFPE) tissues
processed at the Dartmouth-Hitchcock Medical Center (DHMC). DNA specimens
were multiplexed and sequenced on the Ion Torrent 318v2 chips using the Ion
PGM System. Next generation sequencing was performed using a 50 gene hotspot
panel (AmpliSeq Cancer Hotspot Panel v2, Life Technologies) which includes the
KRAS, NRAS, and BRAF oncogenes. Variants were identified using the Variant
Caller Plugin (v.4.0), and annotation and functional predictions were performed using
Golden Helix SVS (v.8.3.4). Results: A total of 619 colorectal tumor specimens were
evaluated. Of these, 103/619 (16.6%) harbored BRAF p.V600 mutations and
236/619 (38.1%) contained at least one of the RAS mutations recommended for
analysis. Remaining mutations were most commonly identified in the following
genes: TP53 (55%), APC (36.3%), PIK3CA (16.2%), FBXW7 (8.1%), and SMAD4
(6.5%). In total, 276/619 (44.6%) of the tumors tested harbored actionable mutations.
Conclusions: Our study indicates that recent guidelines outlining the mutational
evaluation of colorectal cancers can provide treatment recommendations for almost
half of the colorectal cancers encountered at our institution. With guidelines now
available, further consideration can be made to limiting mutational analysis to
guideline-recommended genes with prognostic and predictive value as opposed to
large panels that include predominantly non-actionable genes.
ST47. Evaluation of Targeted Next Generation Sequencing of Circulating CellFree Tumor DNA for Clinical Diagnosis Using Archer Reveal ctDNA Assay
A.A. Stence, N.V. Guseva, R.Y. Walder, K. Sompallae, A.N. Fillman, B. Chang, M.
Deqin, A.D. Bossler
University of Iowa, Iowa City, IA.
Introduction: The ability to detect fragments of tumor DNA, or cfDNA, in the
bloodstream can potentially be used as a non-invasive means of early detection and
treatment response of cancer. Here we assess the performance of a next generation
sequencing (NGS)-based assay utilizing Anchored Multiplex PCR technology
(Archer Reveal ctDNA) for the detection of circulating ctDNA from 28 genes
commonly found mutated in solid tumor type cancers. Methods: Whole blood
specimens were collected in EDTA and Streck Cell-Free DNA BCTs (Streck, La
Vista, NE), and cfDNA was extracted using the Circulating DNA kit (Qiagen, Hilden,
Germany). This DNA, and a commercially available multiplex cfDNA reference
standard set (Horizon Discovery, Cambridge, United Kingdom) was interrogated by
sequencing analysis of 28 gene using the Archer Reveal ctDNA Panel (ArcherDX,
Boulder, CO) and the Illumina Miseq (Illumina, San Diego, CA). Three negative
samples (2 healthy donors and 1 negative from reference standard set), and 3
positive samples (5, 1, and 0.1% multiplex reference standards) were included. Data
was analyzed with Archer Analysis Pipeline 5.0. Multiplex cfDNA reference standard
set and negative control was run with in-house EGFR T790M ddPCR assay on the
Bio-Rad QX200 platform (Bio-Rad, Hercules, California). Results: The Archer
Reveal ctDNA assesses hotspots in 28 genes, with full exonic coverage
for TP53. The Horizon Discovery cfDNA multiplex reference standards have 8
established mutations across 3 allelic frequencies. Expected gene variants were
detected in 5, 1, and 0.1% levels of the multiplex reference standards. Digital droplet
PCR was performed on reference standards and negative control for EGFR T790M.
The laboratory developed ddPCR assay confirmed results found in the NGS-based
assay for the T790M variant. Conclusions: After addressing read depth issues,
sample analysis, and analysis settings on the pipeline, we detected the expected
mutations in the 5, 1 and 0.1% % cfDNA reference standards. Negative specimens
were absent of deleterious mutations as expected. The Archer Reveal ctDNA NGSbased assay is a reliable platform for simultaneous detection of multiple gene targets
commonly mutated in solid tumor type cancers on cfDNA.
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
ST48. Cell Free DNA in Patients with Pancreatic Adenocarcinoma: Evaluation
of a Commercial Assay and Clinicopathologic Correlations
T. Theparee1, M. Akroush1, L. Sabatini1, K. Mangold1, S.J. Stocker1, M.
Talamonti1, K. Kaul2
1NorthShore, Evanston, IL; 2NorthShore University Health System, Evanston, IL.
Introduction: Pancreatic adenocarcinoma has a high mortality rate, is often
detected in advanced stages, and is often inoperable. Detection of circulating tumor
DNA (ctDNA) from plasma cell free DNA (cfDNA) has shown promise for diagnosis,
therapeutic targeting, and monitoring. Multiple extraction methods and platforms
including digital PCR and next generation sequencing have been designed for
detection of ctDNA with differing detection rates, input DNA, and sensitivities. This
study explores the use of commercial ctDNA detection methods and
clinicopathologic factors associated with ctDNA detection in patients with pancreatic
lesions. Methods: Patients undergoing surgical resection of pancreatic lesions were
enrolled with informed consent and plasma samples were collected prior to surgery
and stored at -80°C. Cell free DNA was recovered from 4-6 ml of plasma according
to the manufacturer’s instructions (ZymoResearch Quick-cfDNA Serum & Plasma
Kit). DNA recovery was assessed using the 4200 TapeStation (Agilent
Technologies). 18.9 to 50 ng of DNA was sequenced using the Oncomine Lung
cfDNA Assay (ThermoFisher) on the Ion Torrent S5 sequencing platform. 0.1%
multiplex I cfDNA HD779 (Horizon Discovery) cfDNA reference standards were used
as a comparison. Statistical analyses using Chi-squared test, Mann-Whitney U test,
and linear regression were performed. Results: cfDNA was successfully recovered
from 37 of 38 plasma samples ranging from 22.5-836.8 ng cfDNA (3.8 to 139.5 ng/ml
of plasma) at the 190 and 360 bp bands. Depth of sequence coverage
for KRAS ranged from 26,000x to 98,000x while the mean depth of coverage for all
targets on the panel ranged from 29,000x to 113,000x.
Eight patients (21.1%) showed a total of 6 KRAS mutations in codon 12 and 3 TP53
mutations. Allele frequencies detected ranged from 0.07-2.03%. Detectable ctDNA
mutations were more frequent in patients with poorly differentiated tumors. Patients
without detectable ctDNA mutations showed longer survival (medians of 12.5 months
versus 3.5 months, p=0.006). Conclusions: The use of commercial cfDNA
purification and next generation sequencing kits successfully detects low frequency
circulating tumor DNA in pancreatic adenocarcinomas, particularly in poorly
differentiated adenocarcinomas. Further clinical and pathologic correlations are
underway.
ST49. Improved Detection of Low Abundance Somatic Mutations of KRAS,
BRAF, NRAS and PIK3CA in Melanoma Using iPLEX HS, a New Highly
Sensitive Assay for MassARRAY
B.C. Sutton1, R. Birse2, R.T. Birse2, K. Maggert1, T. Ray1, J. Hobbs1, A.
Ezenekwe1, S. Hummel2, J. Kish1, A. Bullock3, Z. Shi3, S. Stack3, D. Irwin2
1Foundation Medicine, South Bend, IN; 2Agena Bioscience, San Diego,
CA; 3University of Notre Dame, South Bend, IN.
Introduction: Malignant melanoma incidence is increasing more rapidly than any
other malignancy, except lung cancer. According to the American Cancer Society, in
2017 an estimated 87,110 new cases will be identified and 9,730 patients will die of
melanoma in the United States. Identification of genomic mutations associated with
melanoma has led to the development of actionable biomarkers that currently
provide prognostic information, and predict effectiveness of specific targeted
therapies. However, analytical challenges remain to be resolved such as the need
for reliable detection of low abundance somatic mutations, particularly in small
specimens with a low percentage of tumor cells. In this study, we assessed 59
cases of melanoma previously tested for BRAF, KRAS, NRAS, and
PIK3CA mutations using a novel analytic approach that reduces wild type signal,
allowing for detection of low mutation load approaching 1%, the iPLEX HS chemistry
for MassARRAY System (Agena Bioscience, San Diego, CA). Methods: Archived,
frozen deoxyribonucleic acid (DNA) samples were searched for melanoma cases
previously tested for BRAF, NRAS, KRAS and PIK3CA mutations using the
OncoFOCUS Panel on the MassARRAY system. Specimens were deidentified prior
to entry into the study. DNA originated from formalin fixed, paraffin embedded
(FFPE) human tumor tissue samples and all histologic diagnoses were confirmed by
a pathologist. Eight cases were primary tumor; 2 were recurrent lesions; 4 lesions
were of uncertain stage; and the rest were metastatic deposits. DNA was extracted
using the QIAamp DNA FFPE Tissue Kit (Qiagen, Boston, MA). Prior to repeat
testing, specimens were assessed for DNA integrity using the iPLEX Pro Sample ID
Panel, and all specimens with adequate amplifiable DNA were then interrogated with
a new, highly sensitive iPLEX HS panel that includes more than 86 common
mutations in BRAF, NRAS, KRAS and PIK3CA, both using the MassARRAY System.
Results: The OncoFOCUS assay has a mutation detection limit of 5-10% mutant
allele burden, while the iPLEX HS is more sensitive at about 1%. In this pilot study of
59 patient samples, the iPLEX HS panel confirmed all previously
identified BRAF mutations (n=26; 26/59=44%), and NRAS mutations (n=9;
9/59=15%); one additional BRAF V600E mutation and one KRAS G12C mutation
were detected. Two PIK3CA mutations were also identified in this group.
Conclusions: These results show that this approach is a powerful method for the
detection of rare somatic mutations in key Melanoma mutations. This study is an
example that the iPLEX HS is readily amenable to scaling and high through-put; with
a rapid and simple workflow and minimal sample input requirements to detect lowlevel mutations in tumor samples.
1019
AMP Abstracts
ST50. Low Level METex14 Skipping Is Observed at Low Frequencies in
Patients with Solid Tumors from the NCI-MATCH Clinical Trial
V. Datta1, R.D. Harrington1, H. Robinson2, D.J. Sims1, S. McDermott1, C.
Karlovich1, E.P. Mitchell3, J. Wright4, A. Chen4, K. Flaherty5, D.S. Hong6, E.L.
Kwak7, D. Catenacci8, M. Routbort6, B. Conley4, J. Sklar9, S. Hamilton6, J. Iafrate2
1Frederick National Laboratory of Cancer Research, Frederick, MD; 2Massachusetts
General Hospital, Boston, MA; 3 Sidney Kimmel Cancer Center at Jefferson,
Philadelphia, PA; 4National Cancer Institute, Bethesda, MD; 5Massachusetts General
Hospital-Cancer Center, Boston, MA; 6University of Texas M.D. Anderson Cancer
Center, Houston TX; 7Massachusetts General Hospital, Boston, MA; 8Univeristy of
Chicago, Chicago, IL; 9Yale School of Medicine, New Haven, CT.
Introduction: MET exon 14 skipping (METex14), which results in decreased
ubiquitination and delayed downregulation of the MET receptor after ligand
stimulation (Chi et al. 2011), has recently been described as a potential driver
alteration in lung cancer. The NCI-MATCH (EAY131) clinical trial is a signal finding
study, assigning treatment to patients with relapsed/refractory solid tumors,
lymphomas, myelomas, or rare tumors. The subprotocol C2 of EAY131 assigns
eligible patients whose tumors had METex14 for treatment with crizotinib, an oral
small-molecule inhibitor of MET, ALK, and ROS1 receptor tyrosine kinases. In this
abstract, we describe 20 cases of lung adenocarcinoma and other tumor histologies
with METex14. All patients were considered eligible for treatment with crizotinib per
EAY131-C2 clinical protocol. Methods: METex14 was detected in RNA extracted
from FFPE tumor biopsy material using the NCI-MATCH AmpliSeq assay. Validation
of the assay was performed using RNA extracted from lung adenocarcinoma
samples deemed METex14-positive by an orthogonal DNA assay. A conservative
threshold was set at 1000 read counts after discussion with industry experts. Binary
Alignment Map (BAM) files from low-level METex14 cases were scrutinized using
Integrative Genomic Viewer (IGV). Results: Twenty patient cases with various tumor
histologies had METex14 above the assay threshold of 1000 read counts,
representing an overall prevalence in the EAY131 screened population of 0.40% (n =
4956). None of the 20 patients had concomitant MET amplification. Read counts
were distributed in a bimodal fashion. Four of the 20 patients had a diagnosis of lung
adenocarcinoma and had a read count of more than 100,000. However, in 16
patients METex14 was less than 5,500 read counts, including one lung
adenocarcinoma. We visualized the BAM files for low-level cases in IGV and
identified 3 instances of METex14 with additional base substitutions and/or indels
near splice junction sites. Several low-level METex14 cases also had concomitant
KRAS/BRAF variants. Twelve of the 16 low-level METex14 patients were enrolled in
EAY131-C2 and received treatment with crizotinib. Conclusions: METex14 is an
alternative RNA splice variant of MET transcription, and clinical activity has been
observed in METex14 lung tumors treated with crizotinib (Paik et al, 2015). In
EAY131-C2, METex14 was identified in histologies other than lung adenocarcinoma,
but thus far only low-level events were observed. We hypothesize that low-level RNA
read counts for intragenic events like METex14 will not predict response and should
likely be considered separately from high read count cases. Analysis of low level
METex14 cases at the DNA level is ongoing and will be presented.
ST51. Assessing Sensitivity of NGS RNA Fusion Assays Using a Multiplexed
and Well Characterized Linearity Panel
C. Huang, P. Kamineni, D. Philkana, M. Ryder, B. Anekella
SeraCare Life Sciences, Gaithersburg, MD.
Introduction: Many targeted NGS-based panels have recently been introduced to
detect RNA fusions useful for prognosis and therapy selection in
cancer. Understanding the lower limits of detection (LOD) and the reportable range
of these assays is important for clinical adoption and regulatory
compliance. However, there are few tools available to assess LOD. We generated
an RNA panel that represents decreasing levels of 16 different RNA fusions and
exon skipping events compared to the total cellular RNA. Fusion RNAs were
sequence verified and their concentrations measured by dPCR. The panel is being
used to assess LOD and reportable range of an NGS panel. Methods: GM24385
cell line was engineered to contain 15 different RNA Fusions and an exon skipping
event. These engineered cells went through formalin fixation and RNA was extracted
using the Maxwell RSC FFPE RNA kit. Similarly, RNA was extracted from fixed, nonengineered, GM24385 cells. RNA from engineered “fusion” cells and from normal
cells was quantitated by Qubit RNA HS Assay and normalized to 25 ng/µL. The
fusion RNA was diluted into normal RNA in serial 1:5 dilutions. The panel was
characterized by fusion-specific, TaqMan digital PCR assays: Panel member #1
(highest member) was characterized for all 16 fusions, while all panel members were
tested for 3 fusion RNAs including CD74-ROS1, EML4-ALK and SLC34A2ROS1. NGS testing used the ArcherDx FusionPlex Solid Tumor panel with 250 ng of
input RNA and an Illumina MiSeq instrument. Results: Panel member #1 had fusion
RNAs present at extremely high levels, ranging from 18,430 copies/µL for FGFR3TACC3 up to 95,200 copies/ µL for EGFR variant III. The mean among the 16
fusion targets was 43,779 cp/µL and the median was 33,960 cp/ µL. Digital PCR
analysis of all panel members for select fusion targets showed consistent 5-fold
decreases in concentration. For example, EML4-ALK concentrations were 18700,
4220, 772, 184.4, 39.4 and 6.5 copies/µL across the panel members, which has
R2 value of 0.99. NGS testing of panel member 2 on the ArcherDx FusionPlex panel
gave high positive results. The unique start sites (a semi-quantitative indicator of the
1020
number of unique molecules) of between 104 for CD74-ROS1 to 706 for TPM3NTRK1. Additional testing is ongoing to determine the concentration at which failures
of detection are first observed. Conclusions: Determining limits of detection and
reportable range is important when validating an NGS assay. The panel described
here is a proof of concept of how reference materials can help establish these limits.
Additionally, such panels may be useful when evaluating competing NGS
technologies or when evaluating changes to sequencing workflows or bioinformatics
analysis pipelines.
ST52. Clinical Cancer Whole Exome and Transcriptome Sequencing of
Pediatric Tumors at Columbia University Medical Center: Laboratory
Perspective at Three Years
S.J. Hsiao, J.A. Oberg, D. Pendrick, A.N. Sireci, A.T. Turk, M. Sulis, J.H. Garvin, D.J.
Yamashiro, J.L. Glade Bender, M.M. Mansukhani
Columbia University Medical Center, New York, NY.
Introduction: Molecular profiling of tumors offers much potential in identifying
therapeutic targets, in detecting prognostic markers, and in establishing or
confirming clinicopathological diagnoses. These data are being increasingly
incorporated into clinical practice to guide patient management. Methods: Cancer
whole exome sequencing of paired tumor-normal samples and transcriptome
sequencing of tumor samples was performed on specimens from pediatric oncology
patients. Exome sequences were captured using Agilent SureSelect v5 + UTR
reagents and sequenced on the Illumina HiSeq 2500. Variant calling and annotation
were performed using laboratory developed pipelines. RNA libraries were prepared
using the TruSeq Stranded Total RNA LT sample prep kit. Fusions were identified
using FusionMap software and gene expression outliers were identified using a
custom pipeline. All cases were reviewed at a multi-disciplinary molecular tumor
board prior to case sign out and transmission of the clinical report to the patient’s
electronic medical record. Results: Two hundred fourteen clinical samples were
sequenced; from 2014 to 2016, this represented an approximate 70% in growth in
case volume per year. These 214 samples (3.7% blood, 17.8% bone marrow,
38.8% FFPE, and 39.7% frozen tumor tissue) encompassed a broad range of
pediatric malignancies (31.3% CNS, 25.2% hematological, 1.4% renal, 4.7% liver,
5.6% neuroblastoma, 4.7% rhabdomyosarcoma, 10.3% other sarcomas, 4.2% Wilms
tumors, 7.4% mesenchymal, and 5.1% other tumors). A high success rate (>98%)
for sequencing was seen; in FFPE tissue, the DNA sequencing success rate were
similar to those seen in other sample types, and the RNA sequencing success rate
was 80.5%. Potentially actionable findings were seen in 34.6% of cases. Of the
actionable findings, 40.5% were detected by transcriptome analysis. Germline
findings were seen in 8.9% of cases (inclusive of cancer syndromes as well as
ACMG secondary findings). Conclusions: Cancer whole exome and transcriptome
sequencing is increasingly being utilized at our institution and yields actionable
findings in a substantial percentage of cases. Transcriptome analysis and
identification of germline variants notably enhances the utility of genomic analysis of
pediatric tumors.
ST53. Pre-Designed Gene Content Enables Rapid Deployment of HighQuality Customized Enrichment Panels
A.J. Barry1, C. Hendrickson2, K. Patel2, A. Emerman2, S. Bowman2, A. Scott2, C.
Elfe2, F. Stewart1, S. Russello1, T. Davis1
1New England Biolabs, Ipswich, MA; 2Directed Genomics, Ipswich, MA.
Introduction: Efficient utilization of targeted gene panels for translational research
applications is challenged by the wide variation in gene constituents specific to a
given study. While focused gene panels provide the necessary depth of coverage for
variant detection and discovery, the high costs and design challenges associated
with de novo panel design presents challenges. Methods: The NEBNext Direct
technology utilizes a novel approach to selectively enrich nucleic acid targets ranging
from a single gene to several hundred genes, without sacrificing specificity or
coverage uniformity. The approach rapidly hybridizes both strands of genomic DNA
with biotinylated oligonucleotide baits prior to a straptavidin bead capture, enzymatic
removes of off-target sequence, and conversion of captured molecules into
sequencer-ready libraries containing an 8 basepair sample index and a 12 basepair
unique molecule index, or UMI. The UMI individually barcodes each molecule prior to
the final PCR amplification of the library, enabling identification of PCR duplicates.
The result is a 1-day protocol that enables the preparation of sequence-ready
libraries from purified genomic DNA specific to the gene content included in the
panel. Results: We have designed and developed baits specific to the full exonic
content of >900 genes across a variety of disease areas. These are balanced and
pooled on a per gene basis, and can be combined into customized panels, allowing
rapid turnaround of specific custom gene subsets. Conclusions: Here, we present
the ability to rapidly deploy custom gene panels across a variety of panel sizes and
content, while maintaining high specificity, uniformity of coverage across target
conetn, and sensitivity to detect nucleic acid variants to drive translational research
applications.
ST54. EGFR Gene Mutations Analysis in Non-Small Cell Lung Cancer Using
Cobas Assay in FFPE and Plasma Specimen Types
L. Cai, C. Wilson, S. Tuck, M. Stanley, S. Hood, M. Eisenberg, L. Kam-Morgan
Laboratory Corporation of America Holdings, Research Triangle Park, NC.
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
Introduction: The FDA approved cobas EGFR Mutation Test v2 is a real-time PCR
test for the qualitative detection of defined mutations of the EGFR gene in DNA
derived from formalin-fixed paraffin-embedded (FFPE) or plasma human non-small
cell lung cancer (NSCLC) patients. The test is intended to aid in identifying patients
with NSCLC whose tumors have defined EGFR mutations and for whom safety and
efficacy of a drug have been established. In this study, we have evaluated the
clinical and analytical performance features of the assay. Methods: Genomic DNA
was isolated from the tumor specimens using the cobas DNA Sample Preparation
Kit. Mutation detection is achieved through real-time PCR analysis on the
cobas z480 analyzer. DNA from NSCLC specimens were used to evaluate accuracy,
repeatability, reproducibility and analytical sensitivity of the assay. Results: Of the
specimens tested during validation, 20 FFPE specimens with known mutations in
EGFR had cobas results that were 90% concordant and 20 plasma specimens were
100% concordant. The lower concordance in FFPE specimen type was due to
borderline positive specimens, specimen degradation and detection sensitivity
difference between the 2 platforms. Repeatability was 90% concordant for FFPE
specimens and 100% for plasma specimens; reproducibility was100% concordant for
both FFPE and plasma specimens. The lower concordance for FFPE repeatability
was due to borderline positive specimens. This assay can detect 5% of mutant DNA
in a background of wild type genomic DNA when the input DNA is 50ng for FFPE
specimens and 100 copies for plasma specimens.
The cobas EGFR Mutation Test v2 has been offered as a clinical test in LabCorp. Of
the 389 FFPE specimens tested, 53.47% were negative, 42.67% were positive with
1-3 EGFR mutations. 27.50% specimens had one variant, 11.31% contained exon
19 deletion/T790M variants, 2.83% had T790M/L858R variants, and 1.03% had 3
variants or 2 other variants. Results could not be obtained in 3.85% specimens due
to specimen degradation or low DNA yield. Of the 513 plasma specimens tested,
59.84% were negative, 39.96% were positive with 1-3 EGFR mutations. 26.12%
specimens had one variant, 7.4% contained exon 19 deletion/T790M variants, 5.46%
had T790M/L858R variants, and 0.98% had 3 variants or 2 other variants. Results
could not be obtained in 1 (0.19%) specimen. There was no difference in the
mutation rate and distribution between males and females. The higher detection
sensitivity of the cobas platform has resulted in the higher mutation detection rate
than other platforms. Conclusions: The cobas EGFR Mutation Test v2 is a robust,
reproducible, sensitive, and fast assay for molecular diagnostic utilization in NSCLC
using FFPE or plasma specimen types.
ST55. Early Evaluation Site Experience with a Liquid Biopsy Kit Designed for
Next Generation Sequencing of Circulating Tumor DNA
S. Gunn, S. Verma, C. Sims, M. Moore, P. Cotter
ResearchDx/PacificDx, Irvine, CA.
Introduction: Liquid biopsies provide a minimally invasive source of circulating
tumor (ctDNA) for next generation sequencing (NGS) with the potential to
revolutionize treatment planning and longitudinal surveillance of many solid tumor
types. However, for most laboratories, the challenges associated with accurate
detection of mutations at < 1% allele frequency create a barrier to entry in routine
use of liquid biopsy for clinical research applications. Challenges include design of a
panel with clinical relevance across multiple tumor genome profiles, adequate
sequencing coverage over regions of interest, integration of liquid biopsy samples
into the routine laboratory workflow, and bioinformatics support. Methods: In the
current study, our CAP/CLIA accredited laboratory was selected to participate in an
Early Evaluation Program (EEP) of the research use only (RUO) AVENIO ctDNA
Analysis Targeted and Expanded Kits (Roche Sequencing Systems, Pleasanton,
CA). The Targeted and Expanded kits interrogate cancer personalized profiles by
deep sequencing (CAPP-Seq) and integrated digital error suppression (IDES)
technology to overcome key challenges of liquid biopsy and provide clinical
laboratories with an end to end capture-based NGS liquid biopsy solution on the
NextSeq platform (Illumina, San Diego CA.) In the EEP, conducted across 3 sites
including our own, 48 normal plasma and contrived cell line DNA samples with
known mutations were evaluated for intra and inter-laboratory accuracy at allele
frequencies of 1%, 0.5%, and 0.25% using 30 ng of input DNA for both
kits. Results: All 3 sites showed 100% sensitivity and specificity at 1% and 0.5%
allele frequencies and > 90% at 0.25%. Median coverage across regions of interest
was 2,700x to 6,400x for both the Expanded and Targeted kits. The kits integrated
seamlessly into our laboratory’s 5-day NGS workflow from DNA extraction through
reporting. Conclusions: The ctDNA kit solution has enabled our laboratory to
perform accurate NGS analysis of liquid biopsy clinical research samples using
existing equipment and staff with minimal disruption of our routine NGS workflow.
Future directions include an expanded 60 sample intra-laboratory evaluation of the
ctDNA kit at allele frequencies below 0.25% using contrived samples prepared from
normal human plasma and fragmented DNA with known SNV’s created from our
onsite repository of over 100 characterized cell lines.
ST56. Clinical Utility of Large Scale Genomic Sequencing of Solid Tumors at a
Large Academic Medical Center
N.A. Brown, B.L. Betz
University of Michigan, Ann Arbor, MI.
Introduction: FoundationOne is a commonly used molecular panel for screening
solid tumors for both standard of care and investigational biomarkers. However, the
clinical utility of this type of large-scale genomic profiling in place of standard of care
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
molecular testing is uncertain. Methods: FoundationOne reports were reviewed for
all University of Michigan patients for whom FoundationOne testing was performed
over a 3-year period. All available medical oncology notes were reviewed for each
patient to determine what action, if any, was taken on the basis of the results.
Results: Testing was successful in 355 patients. The most common cancer types
included colorectal cancer (CRC; 33.8%), non-small cell lung cancer (NSCLC;
17.5%), breast cancer (6.5%), ovarian cancer (5.9%), pancreatobiliary cancer (3.9%)
and head and neck cancer (3.9%). Potentially actionable variants (based on
FoundationOne reporting) were identified in 87.9% of samples and were classified as
relevant for an FDA-approved therapy in the patient’s tumor type (33.8%; 18.6%
if RAS mutations in CRC are excluded), relevant for an FDA-approved therapy in
another tumor type (60%), and relevant for clinical trials (87.0%). 42.5% of the first
category were RAS mutation in CRC patients. FoundationOne results altered patient
management in 47.3% of cases. However, 72.6% of these clinical actions reflect
standard of care molecular findings (predominantly RASmutation status in CRC
patients). Excluding standard of care molecular findings, FoundationOne results
altered patient management in 13.5% of patients. Of these, 43.8% (7.6% of total)
were treated with an FDA-approved drug for an off-label indication, and 56.3% (5.9%
of total) were enrolled in a clinical trial. Of note, all of the variants that affected
patient management are included in the TruSight Tumor 170 Assay (Illumina) and
93.8% are included in the Oncomine Comprehensive Assay (Thermo Fisher
Scientific). Conclusions: The number of patients whose management is affected by
FoundationOne sequencing results is far lower than the number variants designated
actionable within FoundationOne reports and most variants affecting patient
management reflect standard of care molecular analytes. In addition, most, if not all,
variants that affect patient management are detectable using smaller, commercially
available oncology panels that can be implemented in local clinical laboratories.
ST57. Epi proColon, Septin 9 Gene Methylation Detection Assay as a
Screening Tool for Colorectal Cancer
L. Cai, H. Scott, M. Stanley, D. Rutledge, J. Riojas, C. Best, M. Eisenberg, L. KamMorgan
Laboratory Corporation of America Holdings, Research Triangle Park, NC.
Introduction: The Epi proColon test is a qualitative in vitro diagnostic method for the
detection of methylated Septin 9 DNA in EDTA plasma derived from patient whole
blood specimens. Methylation of the target DNA sequence in the promoter region of
the SEPT9_v2 transcript has been associated with the occurrence of CRC. This test
is indicated to screen adults of either sex, 50 years or older, defined as average risk
for CRC, who have been offered and have history of not completing CRC screening.
Methods: DNA is isolated from plasma and treated with bisulfite. Real-time PCR is
performed on the ABI 7500 Fast Dx to detect the methylated form of Septin 9
DNA. DNA from CRC specimens were used to evaluate accuracy, repeatability and
reproducibility of the assay. Results: Of the specimens tested during validation, 10
specimens with known CRC clinical status and positive Septin 9 status had analysis
results that were 100% concordant when tested at 2 different clinical sites; 10
specimens with no CRC clinical status and negative Septin 9 status had analysis
results that were 75% concordant. The false positive rate from 130 data points was
13.25%, which was lower than the manufacture’s 20% rate. Repeatability (intraassay precision) was 100% concordant for Septin 9 positive specimens and 88.9%
for Septin 9 negative specimens; reproducibility (Inter-assay precision) was100%
concordant for Septin 9 positive specimens and 88.9% for Septin 9 negative
specimens. The Epi proColon test has been offered as a clinical test in LabCorp
based on the successful performance features in the validation data. Of the 3047
specimens tested, 73.94% were negative and 23.57% were positive. Among the
positive specimens, 67.78% specimens had one positive result among the triplicate
runs, 19.11% had 2 positive results and 13.11% had 3 positive results. Results could
not be obtained in 2.49% specimens due to low amounts of DNA recovered from the
plasmas. The positive detection rate was similar to the manufacturer’s rate during
assay clinical validation and there was no gender difference. A subset of 350
specimens was from patients <50 years, the positive rate from this group was
16.29% with females at 11.35% and males at 19.61%. The positive rate for patients
≥50 years but <75 was 22.39% (2238 specimens) with females at 19.86% and males
at 25.74%. For patients ≥75 years, the overall positive rate was 34.86% (459
specimens) with no significant difference between males (35.08%) and females
(34.86%). False positive rate increases with age. Patients with a positive Epi
proColon test result should be referred for diagnostic colonoscopy. Conclusions:
The Epi proColon test is a robust assay for molecular screening in CRC cancer
patients using plasma specimen types.
ST58. Application of the GeneReader NGS System in Testing of Actionable
Mutations in Tumor and Blood Samples
C. Mayo de las Casas, M.D. Garzón Ibañez, N.D. Jordana Ariza, A.D. Balada, J. D.
Bertrán-Alamillo, B.d. Garcia, S. Villatoro, S. Rodriguez, E.d. Aldeguer, Z. Yeste, L.
Alonso, R. Campos, S. Viteri Ramirez, M. Gonzalez-Cao, A. Martinez-Bueno, N.
Karachaliou
Quiron Dexeus University Hospital, Barcelona, Spain.
Introduction: Traditional tests such as PCR, FISH and IHC are usually employed to
detect critical alterations in tumor tissue and cancer blood samples. Multiplex
techniques can significantly reduce the quantity of sample needed, turnaround time
and price in this setting, but require a careful evaluation to establish its performance.
1021
AMP Abstracts
Methods: We used the GeneReader NGS System (QIAGEN) and the Actionable
Insights Tumor Panel (AIT), which covers actionable mutations in 12 genes. A total
of 43 FFPE standards (cell lines) and 104 FFPE tumor samples (57 lung, 19
colorectal, 21 endometrium, 2 melanomas and 5 other tumors) were analyzed.
Extraction was performed manually, using GeneRead DNA FFPE kit (QIAGEN). For
analytical performance of the panel in liquid biopsy, we also tested 86 circulating free
DNA samples isolated from different bodily fluid sources: blood (n=75), cerebrospinal
fluids (n=5), pleural fluids (n=2), urines (n=2) and peritoneal lavages (n=2). cfDNA
was isolated using the QIASymphony automatic extractor (QIAGEN). The FFPE and
liquid biopsy DNA samples were then processed and analyzed using the
GeneReader. The mutational status of FFPE and cfDNA samples were confirmed
with PNA-Taqman or Sanger sequencing. Results: In 0/43 FFPE standards, 29/104
FFPE tumor and 62/86 cfDNA samples, DNA concentration was below specifications
(FFPE<2.5 ng/ul, cfDNA<1 ng/ul) but analysis was continued. Concordance between
the GeneReader and alternative analysis was 100% in reference standards. NGS
results were obtained in 103 FFPE samples, revealing EGFR (n=14), KRAS (n=18),
PIK3CA (n=12), NRAS (n=3) and BRAF (n=2) and KIT (n=1) mutations; concordance
with previous genotyping was 100%. Analysis could be completed in 84 cfDNA
samples, detecting EGFR (n=28), KRAS (n=19), BRAF (n=12), NRAS (n=3) and ALK
(n=2) mutations. Concordance was observed in 75 samples (89.3%). In 5 cases
(5.9%) the results were discordant; in 4 remaining cases (4.8%) the GeneReader
detected the EGFR-sensitizing mutation but not the concomitant p.T790M or
p.C797S. The C797S mutation was not in the original panel design but later included
in the design of a lung cancer-specific panel. Conclusions: The GeneReader NGS
System can be used to investigate actionable mutations in FFPE samples, including
those with limited tumor tissue. Further technical refinement is expected to enable
usage in liquid biopsies.
ST59. Rare BRAF Inactivating Mutation G466E and Literature Review
M. Kruzel1, A. Sharma2, B. Middleman1, J. Waddle1, K. Champion1, M. Weindel1, M.
Araneda3
1Med Fusion, Lewisville, TX; 2Texas Oncology PA, Denton, TX; 3Texas Oncology,
PA, Harlingen, TX.
Introduction: BRAF mutations have been shown to occur in nearly 50% of
melanomas, and 10% of colorectal carcinomas. We report 2 patients with tumors
harboring the rare BRAFG466E mutation. This mutation is extremely rare and has
been shown to be an inactivating mutation leading to decreased kinase
activity. Standard BRAF targeted therapy may be ineffective or less effective in this
setting. The use of EGFR inhibitors and a 5-FU analog have proven effective and
have thus far provided progression free survival for one of our patients. The use of
immune checkpoint inhibitors has also shown to be effective in cases where a
mutation leads to decreased kinase activity. Methods: All genomic interrogation was
completed at med fusion, LLC a CAP and CLIA certified. Briefly, solid tumor biopsies
were obtained from the patients’ tumor site and sent to med fusion for molecular
profiling. All variants for this case report were identified via Life
Technologies ionTorrent Personal Genome Machine, massively parallel
sequencing. Variants were analyzed utilizing Genome Oncology a third-party
analysis provider and then verified by internal genomic analysts. Results: Patient
one has metastatic melanoma with lung and CNS involvement. Following
explanation of therapeutic options and potential outcomes, the patient refused
treatment and elected to enter hospice. Patient 2 presented with colorectal cancer
(sigmoid mass) and was treated with resection followed by adjuvant chemotherapy
(FOLFOX). Following treatment, a PET scan revealed metastases/recurrence within
the perirectal region and left lower quadrant mesentery. The patient was started on
FOLFIRI and bevacizumab but eventually progressed on that regimen. The patient is
currently being treated with panitumumab and capecitabine without noted disease
progression. Conclusions: The advent of massively parallel sequencing has yielded
the discovery of an increased number of variants within cancer. BRAF has
implications across several tumor types and thus, these variants can impact therapy
and prognosis in multiple clinical settings. The NCCN guidelines are clear regarding
the treatment of activating BRAF mutations, but lack therapeutic guidance on
treating inactivating mutations. It has been reported that immune checkpoint
inhibitors are useful in patients with inactivating mutations. Reports have indicated
BRAF G466E can stimulate the MEK-ERK pathway by activation of CRAF in CRC
but no literature exists for this mutation in melanoma. Our report indicates that the
use of EGFR and 5-FU in this setting may demonstrate clinical benefit. The
continued utilization of NGS will further the discovery of rare mutations within various
tumor types which may lead to additional novel therapeutic targets.
ST60. Ion Torrent Next Generation Sequencing – Detect 0.1% Low Frequency
Somatic Variants and Copy Number Variations Simultaneously in Cell-Free
DNA
Y. Li1, J. Gu1, J. Schageman2, B. Dima1, K. Lea1, M. Jasti1, P. Kshatriya1, V.
Bagai1, K.S. Bramlett1
1Thermo Fisher Scientific, Austin, TX; 2Thermo Fisher Scientific, San Francisco, CA.
Introduction: Genetic variation information from cell-free DNA (cfDNA) will provide
valuable opportunities in cancer research and impact future oncology treatment and
diagnosis. We have previously reported the Oncomine Breast cfDNA Assay enables
detection of somatic mutations in plasma down to a level of 0.1% variant allelic
frequency in breast cancer relevant genes. Here we extend this technology to
1022
simultaneously detect single nucleotide variants (SNVs) as well as structural variants
such as copy number variation (CNV) from a single cfDNA sample. Methods: cfDNA
was extracted from purchased normal donor blood plasma using MagMAX Cell-Free
DNA Isolation Kit. To evaluate sensitivity and specificity of SNV detection, we used
reference DNA containing 0.1% or 0.5% variant allelic frequency based on
fragmented AcroMetrix Oncology Hotspot Control material that mimicked the size of
cfDNA. The variant frequencies of the reference DNA were verified with dPCR. For
CNV sensitivity and specificity evaluation, cfDNA from CNV positive cell lines were
isolated and titrated into normal donor plasma cfDNA background and copy number
status was verified with dPCR. Libraries were prepared manually, templated with Ion
Chef automated templating instrument (Thermo Fisher Scientific) and sequenced
with Ion S5 Sequencing instrument (Thermo Fisher Scientific). Data analysis was
performed using the variantCaller plugin in Torrent Suite Software for SNV target.
Additional plugin tool was developed for CNV analysis. Results: The new
developments including CNV analysis extend our current Oncomine Breast cfDNA
Assay from low frequency SNV detection to both SNV and CNV detection. The new
panel covers 160 hotspots in 10 breast cancer related genes (AKT1, EGFR, ERBB2,
ERBB3, ESR1, FBXW7, KRAS, PIK3CA, SF3B1, TP53) plus 3 CNV regions
(ERBB2, FGFR1 and CCND1) commonly related to breast cancers. Additionally it
extended coverage of the tumor suppressor gene TP53 to more than 80% coding
region. This new breast panel is optimized for low allelic frequency detection at 0.1%
in hotspot region and CNV detection as low as 1.4 fold amplification (equivalent to
10% tumor frequency at 5 fold) from cfDNA. Conclusions: The new Ion Torrent
breast assay currently in development provides an easy, quick and reliable solution
for researchers in detecting both low frequencies SNV and CNV simultaneously in
cfDNA.
ST61. Investigation of Mutational Burden in Urothelial Tumors Using a
Targeted NGS Panel
W. Zhang, M. Mansukhani, A. Sireci, T. Gindin, A. Turk, S. Hsiao, H. Fernandes
Columbia University Medical Center, New York, NY.
Introduction: Immune checkpoint inhibitor therapy is likely to be effective in the
treatment of tumors with a higher number of nonsynonymous mutations, resulting in
elevated Tumor Mutational Burden (TMB). Durable clinical responses that correlate
with higher TMB are often seen in non-small cell lung cancers and melanomas. We
assessed the TMB in patients with metastatic urothelial carcinoma (UC) whose
tumors were sequenced using the 467 gene targeted Columbia Comprehensive
Cancer Panel (CCCP). Methodology: DNA from macrodissected lesional tissue of
15 cases with metastatic UC was extracted on the QIAcube (Qiagen, Hilden,
Germany). A minimum of 50ng of sheared DNA was subject to capture using Sure
Select reagents from Agilent (Santa Clara, CA) and sequencing was performed on
the Illumina HiSeq2500 (San Diego, CA) NGS platform. The CCCP gene panel TMB
was calculated as the number of deemed somatic, single nucleotide variants and
indels, per megabase (Mb) of genome interrogated. Based on the variant information
from 250 tumors previously sequenced with the CCCP, the average number of
mutations per megabase was 4.5. For the UC cohort, the 25th, median and 75th
quartile thresholds for TMB were 6.1, 8.6, and 9.6 respectively. Results:
The median age of the patients was 71 years old with a male: female ratio of 2:
1. Four of the fifteen (26%) cases, displayed TMB below the 25% quartile (6.1
mutations per Mb) and 2 of them were classified as low grade tumors. 47% (7/15) of
the tumors with TMB between the 25th and 75th quartile (6.1 - 8.6 mutations per Mb)
were high-grade tumors. 27% of the tumors had TMB above the 75% quartile (9.6
mutations per Mb) and 3 of the 4, were high grade tumors. Interestingly, almost all of
the tumors with TMB near or above the median had frame-shift and/or nonsense
variants. The most frequently mutated genes at 40% each were TP53, PIK3CA,
KDM6A, and MLL2 followed by FGFR3 at 33%. Other significantly mutated genes
included ARID1A, CREBBP, TSC1 and STAG2. At a glance, there was a trend
towards higher TMB in specimens with variants in the TP53 and PIK3CA genes.
Conclusions: The TMB in patients with metastatic UC was diverse. There was a
tendency for increased TMB in high grade tumors with deleterious mutations and
tumors with variants in TP53 and PIK3CA genes. Further investigations with
correlation to clinical outcome in a larger cohort of patients with UC are needed to
determine any association with response to immunotherapy.
ST62. Integrated Molecular Diagnostic Call Criteria for MET Exon 14 Skipping
in Lung Cancer
R.J. Schmidt1, S.R. Digurmarthy2, Z. Zheng3, C. Wang2, A. Macleay2, M. Pacula2, S.
Duraisamy2, E. Dominguez Meneses2, S.H. Al Turki4, M. Rivera2, V. Nardi2, D. DiasSantagata2, L.P. Le2, A.J. Iafrate2, R.S. Heist2, J.K. Lennerz2
1Harvard Medical School, Boston, MA; 2Massachusetts General Hospital, Boston,
MA; 3City University of Hong Kong, Kowloon, Hong Kong; 4Ministry of National Guard
Health Affairs, Riyadh, Saudi Arabia.
Introduction: Preclinical evidence and dramatic responses to MET inhibition in
patients with lung cancer harboring MET exon 14 skipping (METex14) demonstrate
its potential as a novel therapeutic target. METex14 is an oncogenic RNA splicing
isoform whose transcription has been reported in the setting of somatic DNA-level
mutations that affect the splicing of exon 14. Additionally, METex14 has been
observed in the setting of MET gene amplification, indicating that this phenomenon
may also be associated with high-level expression. As a result, multiple molecular
diagnostic modalities are currently needed to assess the biological etiology and
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
clinical significance of METex14. Here we describe our experience in a CLIAcertified lab with combined tissue-based DNA, RNA and copy-number assessment to
derive METex14 call criteria. Methods: We used anchored multiplex PCR in
conjunction with next-generation sequencing (NGS) for targeted DNA and RNA
sequencing of consecutive clinical FFPE tumor specimens over a 2 year period.
DNA sequence, DNA copy number, and RNA transcript variants were detected using
a laboratory-developed bioinformatics pipeline. Copy number alterations were also
assessed using FISH. Expression of METex14 was validated using RTPCR. MET expression was assessed by RT-PCR and IHC. Response criteria
followed RECIST guidelines. Results: Of the 2734 tested samples (1432 non-lung),
we identified N=812/1113 lung cancer patients without detectable METex14 reads
(<2 reads). We used the remaining N=221 lung cancer patients (with more than 2
METex14 reads; range: 2-2466 reads) to derive the following call criteria for
reporting a clinically significant METex14 event: 1) a concurrent DNA mutation
predicted to affect exon 14 splicing; 2) quantitative predominance of the METex14
transcript among all detected alternative MET transcripts; 3) absence of
concurrent MET gene amplification; 4) absence of concurrent driver mutation (i.e.,
KRAS, ALK-, or ROS1-rearrangement etc.). In a separate cohort consisting of N=53
non-neoplastic lung tissue samples, we confirmed that METex14 can be detected at
low levels in MET expressing (RT-PCR/IHC-positive) tissues with up to 48 reads.
METex14 patients fulfilling call criteria treated off-label with a MET inhibitor showed
that 4 of 4 patients had significant tumor shrinkage with 3 patients achieving at least
partial response (2 partial, 1 complete response).
Conclusion: Our data indicate that METex14 transcripts and other alternative
transcripts represent a normal phenomenon of MET expression and low-level
METex14 expression in lung cancer may have different biological and therapeutic
significance. We propose call criteria for an integrated DNA, RNA and copy numberbased diagnosis of METex14 skipping in lung cancer.
ST63. Gene Expression Profiling of Traditional Immunohistochemical Tumor
Biomarkers Using Nuclease Protection Coupled with Targeted Next-Generation
Sequencing
M. Reinholz, Y. Lipovka, J. Cooley, G. Reinholz, Q. Liu, I. Botros, D. Thompson, B.
LaFleur
HTG Molecular Diagnostics, Tucson, AZ
Introduction: The HTG EdgeSeq PATH Assay enables investigators to measure the
expression of 470 human mRNA targets using a proprietary quantitative nuclease
protection chemistry coupled with targeted next-generation sequencing (NGS). This
panel measures the expression of genes that encode proteins commonly detected
by commercially available immunohistochemistry (IHC) antibodies. The HTG
EdgeSeq chemistry allows for extraction-free sample preparation and is compatible
with limited amounts of tissue from formalin-fixed, paraffin-embedded (FFPE)
samples and with cell line controls. Methods: Three studies were conducted to
characterize the performance of the HTG EdgeSeq PATH Assay in FFPE and cell
lines including: 1) sample input; 2) repeatability; and 3) IHC agreement. The sample
input study evaluated a range of sample input amounts for breast FFPE samples and
2 cells lines (MCF-7 and MDA-MB-231; luminal A and triple negative breast cancer,
respectively). Optimal input range was established based on sample quality (using
post-sequencing quality control) and biomarker expression
(sensitivity). Repeatability across technical replicates also was performed for both
sample types. IHC agreement was established for estrogen receptor (ER),
progesterone receptor (PR), Ki-67, and human epidermal growth factor receptor 2
(HER2) expression. Results: The recommended sample input for FFPE samples is
6 mm2 of 5 micron sections and the recommended sample input for cell lines is 2500
cells. Acceptable performance was observed at as low as 1.5 mm2 for FFPE and 313
cells. Pairwise Pearson correlation coefficients for technical replicates were greater
than 0.90 for at least 95% of the well pairs (between processors and days). Current
CAP/CLIA guidelines were used for IHC interpretation. High Pearson correlation
was demonstrated between gene expression and IHC for 14 breast cancer samples.
Gene expression to IHC correlation coefficients for PR, Ki-67, HER2, and ER were
0.77, 0.79, 0.85 and 0.96, respectively. Accuracy, measured by the area under the
Receiver Operating Characteristic (ROC) curve (AUC), was 0.97, 1.0, 0.96, 1.0, for
PR, Ki-67, HER2, and ER, respectively. Conclusions: The HTG EdgeSeq PATH
Assay is a reliable tool for retrospective gene expression profiling to complement
traditional IHC. This assay allows investigators to assess mRNA expression in
translation biomarker development using known cell line control samples as well as
when formalin-fixed, paraffin-embedded sample availability is limited.
ST64. Spectrum of Variants Detected In a Large Cohort of Lung
Adenocarcinomas at New York-Presbyterian Hospital
G. Ramrattan1, J. Baum1, K. Park1, L. Cong1, C. Martin1, Q. Pan2, H. Rennert2
1NewYork-Presbyterian Hospital, New York City, NY; 2Weill Cornell Medicine, New
York City, NY.
Introduction: Adenocarcinoma of the lung is the most prevalent subtype of nonsmall cell lung cancer, accounting for about 40% of cases and with a 5-year survival
rate of ~15%. The overall low survival rate is due in part to late diagnoses, and
inadequate cytotoxic chemotherapies. Recently, targeted therapies aimed at specific
genetic alterations have become a promising alternative; however the number of
targetable alterations within lung adenocarcinoma remains low. In current clinical
practices, Next-Generation Sequencing (NGS) panels are becoming more frequently
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
used to identify diagnostic/prognostic markers. At New York-Presbyterian, we have
used a 50-gene targeted cancer hotspot panel to sequence over 2,500 samples
encompassing a broad range of samples and tumor types; 35% of which are lung
adenocarcinomas. Using this robust cohort, we aim to further enhance the genomic
understanding of this disease as well as share novel findings that may hold
therapeutic utility. Methods: A total of ~940 cases were tested from 2015 to 2017.
DNA was extracted from the tumor-enriched area of FFPE sections and sequenced
on the Ion Torrent Personal Genome Machine (PGM) after library preparation and
enrichment using AmpliSeq Cancer Hotspot Panel v2 (Thermo Fisher Scientific).
Sequencing data was analyzed using the Ion Torrent Suite. Variants present in the
Catalog of Somatic Mutations in Cancer (COSMIC) and an in-house knowledge
database were considered reportable. Results: We were able to successfully
sequence 98% (922/940) of cases, with 82% having reportable alteration(s)
detected: EGFR (30%), KRAS (31%), STK11 (1%), BRAF (4.5%),
PIK3CA (2.7%), ERBB2 (1.2%), and TP53 (30%). In comparison to public data sets
such as TCGA, we observed similar patterns with several exceptions such
as EGFR being mutated twice as frequently in our cohort, possibly due to a higher
incidence of patients from Asian-descent. Interestingly, 23% of cases had genetic
alteration(s) in both an oncogene and a tumor suppressor; specifically,
9.5% EGFR/TP53, 8.5% KRAS/STK11, and 7% had KRAS/TP53 mutations. In
addition, we identified one case having mutations in both EGFR and KRAS.
Conclusions: Actionable genetic alterations were detected in ~30% of lung
adenocarcinoma patients. We identified several rare alterations as well as those
having unusual concomitant variants.
ST65. Development of a Targeted NGS Cancer Gene Panel Using Multiplex
PCR-Based Enrichment in an Integrated Fluidic Circuit
H. Gong, X. Wang, S. Meyers, D. Do, J. Qin, R. Ramakrishnan, P. Chen
Fluidigm, South San Francisco, CA.
Introduction: Next-generation sequencing (NGS) is being rapidly adopted for cancer
gene mutation detection, which is useful in clinical research to align actionable
mutations in tumors to targeted therapies or clinical trials. Compared with whole
genome and whole exome analysis, targeted sequencing can provide an affordable
solution with a much easier data analysis. We have developed a targeted NGS
cancer gene panel based on multiplex PCR enrichment in an integrated fluidic circuit
(IFC). Methods: Multiplex PCR employing target-specific primers is conducted in an
LP—48.48 IFC on the Fluidigm Juno system. Up to 48 DNA samples can be
processed simultaneously. A unique barcode is incorporated in one of the PCR
primers to distinguish individual samples. PCR products harvested from the IFC are
pooled and purified using double-sided solid-phase reversible immobilization (SPRI)
with Agencourt AMPure XP magnetic beads. A second PCR step using a universal
primer pair is performed to add sequencing adapters and generate enough material
for sequencing. After a second purification, the DNA library is sequenced on a
NextSeq 500 system. A pipeline developed by Fluidigm is used for data analysis.
Results: A total of 1,540 primer pairs with an average insert size of 155 bp were
designed to cover single nucleotide variants (SNVs), small insertions/deletions
(indels), and copy number variations (CNVs) in 53 oncogenes and tumor
suppressors. The assays were separated into 38 pools to minimize interaction
between assays and improve performance. NA12878 was used as DNA input for
technical characterization of the panel performance. With a NextSeq 500/550 Mid
Output Kit (300 cycles) an average base coverage depth of 3,500x can be obtained
by paired-end sequencing. After 3 rounds of iterative primer design and reaction
condition optimization, 1,474 out of 1,540 assays (95.7%) have reads higher than
0.2x mean reads. The percentage of reads mapped to genome was about 98.9%,
and the percentage of reads mapped to amplicons was about 96.8%. Horizon Tru-Q
HDx Reference Standards were used to investigate the analytical performance.
Preliminary results revealed a sensitivity of 98.9 % and positive predictive value of
98.7% at a variant allele frequency greater than or equal to 5%. The variant call
between replicates had concordance rates between 0.943 and 1.0. Conclusions: A
targeted NGS cancer panel employing PCR-based enrichment on an IFC has been
developed. This panel covers actionable mutations in 53 cancer genes and utilizes a
microfluidic device to provide a simple streamlined workflow for library preparation of
up to 48 DNA samples per IFC. Such library alone or in combination with other
libraries could be analyzed on multiple Illumina platforms.
ST66. Assessment of Tumor Mutational Burden and Microsatellite Instability
with Illumina’s TruSight Tumor 170 Panel
S. Zhang, A.S. So, S. Kaplan, J. Yao, K.M. Kruglyak
Illumina, San Diego, CA.
Introduction: In recent years, immune checkpoint inhibitors have shown great
promise in treating various cancer types; however, only a fraction of patients respond
to this type of immunotherapy. PD-L1 protein expression measured by quantitative
immunohistochemistry (IHC) is an FDA approved companion diagnostic or
complementary assay for some immune checkpoint inhibitors. Recently, studies
have shown that emerging genomics markers, such as tumor mutational burden
(TMB) and microsatellite instability (MSI), correlate with improved efficacy of immune
checkpoint inhibitors. Methods: Illumina’s TruSight Tumor 170 (TST170, research
use only; RUO) panel is a comprehensive next-generation sequencing (NGS) assay
that covers the coding regions of 170 genes associated with solid tumors. Designed
to capture mutational changes, including single nucleotide variants, indels,
1023
AMP Abstracts
amplifications, splice variants and fusions, TST170 targets both DNA and RNA
variants from the same FFPE tumor sample in a single sequencing run. Here we
evaluate the performance of TMB and MSI assessment with the TST170 panel.
Results: In this study, we compared the estimated TMB of 27 FFPE tumor samples
profiled with both WES and TST170 and saw high correlation (R2=0.998) between
the 2 methods. Using TST170 content to estimate TMB was further validated in
silico in larger cohorts (n=5336) by filtering WES data to TST170 target regions.
Next, we developed a bioinformatics algorithm to determine MSI status using 104
homopolymer repeat loci covered by the TST170 panel. The repeat length of each
homopolymer locus was calculated and compared to a baseline of normal samples,
and the final MSI score was calculated using the number of unstable sites divided by
total tested sites. Each sample was then classified as MSI-high or microsatellite
stable (MSS). We assessed the MSI status of forty colon samples assayed with
TST170 and MSI-PCR. Using TST170 homopolymer loci to determine MSI status
was also further validated in silico in larger cohorts (n=84). We also found that MSI
high samples had significantly higher TMB compared to MSS samples.
Conclusions: In summary, our analysis indicates that the panel content of TST170
can be used to accurately estimate TMB and MSI from FFPE tumor samples.
ST67. Mutational Spectrum in a Multi-Gene Panel of Germline and Somatic
Ovarian Cancer in Singapore
S. Ho1, L. Chiu1, K. Poon1, K.M. Tan1, T.S. Png1, D.S. Tan2, S. Lee2, E.S. Koay3, B.
Yan1
1National University Hospital, Singapore; 2National University Health System,
Singapore; 3National University of Singapore, Singapore.
Introduction: Germline and somatic mutations in ovarian cancer in Singapore are
not well-characterised. In this study, we aim to identify the mutational spectrum of
germline and somatic ovarian cancer in our local population using a 94-multigene
sequencing panel. Methods: Germline samples (n=30) were analysed using the
TruSight Cancer Sequencing Panel that targets 94 genes including BRCA1/2.
Multiplex Ligation-dependent Probe Amplification (MLPA) was performed to
detect BRCA1/2 gain/loss. BRCA1/2 were positive in 8 and negative in 22 samples.
Of the 22 negatives, 15 paired FFPE samples (7 high grade serous, 5 clear cell, 1
seromucinous, 1 endometrioid, 1 mixed mucinous and endometrioid) were retrieved
to analyse for somatic variants. Sequencing data from MiSeq was aligned against
hg19 using MiSeq Reporter software followed by Somatic Variant Caller for regions
specified in the TruSight Cancer manifest file. Filtering was performed according to
previously optimised protocols. Sequencing reads were manually examined using
Integrative Genomics Viewer in ambiguous variant calls while ClinVar was used to
determine pathogenicity. Results: Of the 8 BRCA1/2 positive samples, there were 3
pathogenic variants in BRCA1 and 1 had BRCA1 loss detected by MLPA. Somatic
variants were found in APC, ATM, BRCA1, BRCA2, BRIP1, CDH1, CHEK2, EPCAM,
MSH6, PTEN, STK11, TP53, TSC1. Variants of uncertain significance (VUS) were
found in TP53, STK11, TSC1, and BRIP1 where clinical trials are available.
Conclusions: The identification of pathogenic variants in multigene panels require
an established database in our local population. At present, the mutational spectrum
of germline and somatic ovarian cancer in our local population is not well-established
and eligibility in clinical trials may be missed.
ST68. A Droplet Digital PCR Assay for Detection of Methylated BCAT1 and
IKZF1 in Circulating Tumor DNA
N. Boulter, D. Murray, S. Pedersen
Clinical Genomics Technologies Pty Ltd, North Ryde, New South Wales, Australia.
Introduction: Circulating tumor DNA (ctDNA) can be detected in patients with
colorectal cancer (CRC) using a qualitative real-time PCR test (qPCR) assaying for
methylated BCAT1 and IKZF1 DNA in blood. Quantitative assessment of ctDNA
levels may have clinical utility, including informing effectiveness of tumor debulking
and response to chemo/radio therapy. The aim of this study was to analytically and
clinically evaluate a quantitative test for BCAT1 and IKZF1 using droplet digital PCR
(ddPCR). Methods: Blood samples were collected in K3EDTA tubes, processed to
plasma and stored at -80°C until further testing. DNA extracted from 4mL of plasma
was bisulphite converted and assayed for methylated BCAT1 and IKZF1 DNA using
PCR assays developed for either the QX200 ddPCR (Bio-Rad) or LightCycler 480 II
(Roche) platforms. Analytical samples included pooled plasma, from presumed
normal donors (age < 30yo), spiked with methylated DNA (range: 2.3 to 300 pg/mL).
Clinical samples were obtained from colonoscopy confirmed subjects and each
sample (4mL) was split 50/50 between the ddPCR and qPCR assays. Qualitative
detection of either methylation marker was deemed positive for ctDNA. Results:
Analytical testing determined an equivalent Limit of Detection for both assays
(qPCR, 14.8 pg/mL; ddPCR, 15.9 pg/mL; Probit regression analysis). The ddPCR
assay was quantitative down to 75pg/mL, which equates to 90 copies of methylated
DNA per 4mL of plasma. In contrast, the qPCR assay was not quantitative in the
concentration range tested (i.e. the qPCR limit of quantification was >300pg/mL).
The 2 assays showed 82% concordance in 60 clinical samples, including 20 with no
evidence of disease (NED), 20 adenoma and 20 CRC. There was no significant
difference in the qualitative positivity rates between the 2 assays (CRC, 12/20 in both
assays; Adenoma, ddPCR 2/20, qPCR 1/20; NED 4/20 in both assays). Despite only
assaying half of each clinical sample with either method, 6 of the 12 CRC positive
cases were within the quantitative range of the ddPCR assay. The CRC positivity
rate, and the number of samples that fall within the quantitative range of the assay, is
1024
expected to increase when assaying the entire sample. The ddPCR assay was
further successfully developed as a triplex, using ACTB as a positive control gene,
which maintained the distinction for all 3 markers using only 2 fluorophores.
Conclusions: Detection of methylated BCAT1 and IKZF1 using ddPCR is
comparable to the published qPCR method. 50% of ddPCR-positive CRC cases
were within the quantitative range. Further clinical evaluation of the triplex assay is
required to appraise the utility of the quantitative assay.
ST69. Validation of CD274/PD-L1 FISH as a Predictive Biomarker for the Use of
Immune Check Point Therapies in Undifferentiated Malignancies
K. Devereaux, S. Zhao, G. Charville, D. Bangs, M. van de Rijn, A. Cherry, Y.
Natkunam
Stanford University, Stanford, CA.
Introduction: Undifferentiated malignancies (UMs) pose a therapeutic challenge
with poor clinical response to most chemotherapeutic agents. Given the high degree
of genetic instability in undifferentiated neoplasms, checkpoint inhibitors may provide
an alternative, more effective treatment approach. Recent studies show that tumors
with a genetic basis for PD-L1 expression are highly sensitive to PD-L1 inhibitor
immunotherapy. In our ongoing work, we have observed that the majority of UMs
abberantly express PD-L1; however, genetic alterations at the CD274 (PD-L1) gene
locus have yet to be tested in this class of tumor and may offer clinical utility in
further predicting which patients will respond favorably. Methods: Formalin-fixed,
paraffin-embedded (FFPE) tissue sections from UMs, including undifferentiated
carcinomas (35 cases) and undifferentiated malignant neoplasms (55 cases) from
indiscriminate anatomic locations were selected. Immunohistochemistry (IHC) for
PD-L1 was scored by staining intensity (0 - negative, 1- weak, 2 - moderate, 3strong). FISH was performed on available cases using probes targeting CD274 (PDL1), PDCD1LG2 (PD-L2) and the centrometric region of chromosome 9 and
malignant cells were scored such that a target:control probe ratio of ≥3:1 =
amplified, between 1:1 and 3:1 = relative cogain and 1:1 = polysomy (>2 copies) or
disomy (= 2 copies). Results: Of the cohort of UMs, the majority of undifferentiated
carcinomas (69%; 24/35) and undifferentiated malignant neoplasms (51%; 28/55)
display moderate to strong PD-L1 expression by IHC. Overall, alterations in the PDL1 and PD-L2 loci detected by FISH show concordant genomic scoring. Coamplification or co-gain of PD-L1 and PD-L2 is observed in 15 of 24 (63%) UMs by
FISH, with the mean PD-L1 expression intensity being highest amongst tumors
with PD-L1 amplification. Conclusions: Here we show that the majority of
undifferentiated neoplasms demonstrate chromosomal alterations inCD274/PDL1, providing a genetic basis for aberrant PD-L1 expression in many of these tumors.
Such findings provide molecular data and support for using checkpoint inhibitors as a
novel treatment approach for this tumor class. Validation and further assessment of
the use of PD-L1 FISH in conjunction with PD-L1 IHC in future studies will be
important in helping better predict which patients will benefit most from PD-1
blockade.
ST70. Calculation of Mutational Burden in a Clinical Targeted Next-Generation
Sequencing (NGS) Panel for Solid Tumors
P. Velu, A. Bigdeli, P.R. Hess, J.J. Roth, J.N. Rosenbaum, J.J. Morrissette
University of Pennsylvania Health System, Philadelphia, PA.
Introduction: Developing biomarkers that can predict patient response to PD-1 and
PD-L1 blockade immunotherapy is of high interest. Recent studies using NGS
whole-exome and large targeted gene panels have shown a correlation between
high tumor mutational burden (TMB) and treatment response. Many clinical
laboratories now use targeted gene panels (TGP) with tens to hundreds of genes
that are known to be implicated in cancer and can assess TMB; however, TGPs offer
narrower coverage and are biased toward disease-associated genes. Here we
explore the feasibility of using a TGP for calculation of TMB, and determine whether
this measure provides additional information beyond clinically reported
variants. Methods: Data included MAF files from clinical brain, lung, melanoma, and
GI samples (n=577) that were sequenced using a clinical targeted NGS panel for
solid tumors (153 genes, 0.5 mb coverage) and MAF files of the same tumor types
from the TCGA Data Portal. For comparison, a simulated TGP was created by
limiting TCGA MAF files to only genes present on the clinical panel. TMB was
determined by dividing total mutations by coverage size in mb after removal of noncoding mutations, germline variants, and known somatic alterations in the COSMIC
database. Data processing, analysis, and visualization were performed using R
scripts. Results: In the TCGA data, the amount of correlation between the TMB
calculated from whole exome and TMB calculated from the simulated TGP was
similar across different variant callers (average R2 difference <0.05) but varied widely
depending on level of TMB and tumor type (R2 range 0.18 to 0.99). Correlation
tended to be poor at lower TMB (<10 mutations/mb) and in tumor types with few data
points outside this range. Removing synonymous mutations reduced TMB
significantly, but limiting variants to those with population allele frequencies less than
0.1% rather than less than 1% did not, and neither change affected correlations
significantly. In the institutional data, the total number of mutations clinically reported
after variant review (VR) compared to the total calculated by the TMB algorithm
showed no correlation (R2<0.1). The frequency distribution of total mutations per
patient revealed a distinct population with a higher number of mutations in all 4 tumor
types using TMB, but only in GI and melanoma using VR. Conclusions: Small
TGPs can be used for TMB calculations. Poor correlation between whole exome and
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
simulated TGP TMB in data sets with only low TMB values is likely due to sampling
noise. Calculation of TMB from institutional data reveals a distinct population of
patients with higher TMB in each of the tumor types. Correlation with future matched
outcomes data is needed to determine the clinical utility of this measure.
ST71. Molecular Characterization of a Series of Solitary Fibrous Tumors,
Tested for NATB2-STAT6 Fusion Transcripts Using Reverse
Transcriptase(RT)–Polymerase Chain Reaction(PCR) Technique: an Indian
Experience
B. Rekhi, O. Shetty, P. Tripathi, P. Bapat, M. Ramadwar, J. Bajpai, A. Puri
Tata Memorial Hospital, Mumbai Maharashtra, India.
Introduction: Recent studies, mostly from the west, have unraveled a specific fusion
transcript, namely NAB2-STAT, underlying a solitary fibrous tumor (SFT) and STAT6
as a fairly sensitive and specific immunohistochemical (IHC) marker for diagnosing a
SFT. Methods: Twenty-six cases of SFTs, after review, were tested for 8 variants
of NAB2-STAT6 fusion transcripts, using qualitative endpoint reverse-transcriptase
(RT)-PCR technique. All these tumors were also tested for IHC expression of
STAT6, which was further graded, based on intensity (mild, moderate and strong)
and percentage of tumor cells showing positive immunostaining, ranging from 125%(1+); 26-50%(2+); 51-75%(3+) and in more than 75%(4+)tumor nuclei.
Additionally, 9 other tumors were tested for NAB2-STAT6 fusion transcripts. RNA
extraction was performed using Recover All Total nucleic acid extraction kit (Ambion,
Thermo Scientific, USA). Extracted RNA was quantified using Nanodrop. RNA was
reverse transcribed using Revertaid H-minus First strand cDNA synthesis kit
(Fermentas, Thermoscientific, USA) along with No RT controls to rule out genomic
DNA contamination. The integrity of the cDNA was assessed by performing PCR for
beta actin (ACTNB) gene. The selected cases were screened for all the 8 fusion
transcripts using 8 primer pairs for NAB2 and STAT6 gene spanning ex
6/17,2/19,7/2,5/3,6/18,4/5,5/8,3/3 respectively with product size ranging from 200330bp. Immunohistochemistry for STAT6 antibody was performed by polymer
detection technique. Results: Twenty-six SFTs occurred in 15 men and 11 women
(M: F=1.4:1), with age varying from 13-72 years (average=49.9); across various
body sites. Out of 37 initially reviewed SFTs, 26(70.2%) cases yielded amplifiable
RNA. By RT-PCR technique, 23/26(88.4%) cases of SFT were found to be positive
for NAB2-STAT6 fusions, including NAB2ex7/STAT6ex2 (5 cases),
NAB2ex6/SAT6ex17 (4cases), NAB2ex2/STAT6ex19 (4cases), NAB2ex5/STAT6ex3
(4cases), NAB2ex3/STAT6ex3 (4cases), NAB2ex6/STAT6ex18 (1case), NAB2ex4/S
TAT6ex5 (single case) and NAB2ex5/STAT6ex18 (none). None of the 8 variants
of NAB2-STAT6 were found to be positive in 5 cases of synovial sarcoma, 2 of
Ewing sarcoma and 2 cases of MPNST, tested for the same (100% specificity).
Immunohistochemically, most SFTs displayed moderate (9/26) and strong (14/26)
immunostaining for STAT6, including 3+ (7/26cases) and 4+ (13/26cases)
immunostaining patterns. Conclusions: This study, constituting as the first one from
India, reinforces high sensitivity and specificity of NAB2-STAT6 fusion transcripts in
SFTs. The most common variant observed was NAB2ex7-STAT6ex2. Strong and
diffuse IHC expression of STAT6 correlated with the NAB2-STAT6 fusion, which
appears as the current gold standard for diagnosis of a SFT.
ST72. Prospective Analysis of the Clinical Impact of Expanded Genomic Tumor
Testing on Management and Outcomes of Adult Oncology Patients at a Large
Academic Medical Center
A. Sireci, R. Carvajal, S. Hsiao, K. Komatsubara, E. Lim, L. Roman, H.
Fernandes, A. Turk, H. Remotti, A.R. Sepulveda, T. Gindin, M. Mansukhani
Columbia University Irving Medical Center, New York, NY.
Introduction: The Columbia Combined Cancer Panel (CCCP) is a 467 gene panel
that detects single nucleotide variants (SNV), small indels and copy number
alterations in both solid and hematologic tumors. We have seen an increase in
demand for this testing by oncologists over time. However, reimbursement has been
slow to materialize with many payers considering expanded genomic testing
“experimental” thereby limiting access to this powerful diagnostic modality. We are
embarking on a prospective QI initiative to describe the ordering patterns and clinical
impact of CCCP-style testing on the management and outcomes of adult oncology
patients at CUMC. These are our results to date. Methods: Consecutive cases
submitted for CCCP testing with results finalized after March 1, 2017 were reviewed
and discussed during a weekly adult molecular tumor board. Reason for testing and
category of clinical impact were discussed among expert panel members and
recorded. Additionally, payer information and details regarding clinical outcomes
(e.g., enrollment into clinical trials or results of germline genetic testing) were
documented for each patient. Results through May 9th, 2017 are presented.
Results: To date we have prospectively gathered data on 34 patients with both solid
and hematologic malignancies. The most common reason for ordering testing was to
gain information on a resected tumor with a high risk of recurrence or treatment
failure on conventional therapy. 74% (25/34) of CCCP findings were categorized as
clinically actionable or impactful: 52% (13/25) would inform future management via
experimental protocols; 16% (4/25) led to referral to genetic counseling for potential
germline predisposition testing; 8 % (2/25) had results which could impact future
management with approved targeted agents. 56% (14/25) of patients with impactful
results had molecular results qualifying them for clinical trial enrollment: 4 of these
patients are currently awaiting enrollment; 7 patients are in remission or responding
to standard therapy; 3 patients are ineligible due to performance status or trial
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
closure. Thirty-one percent of all cases were insured by a government payer that
does not pay for this testing. Of commercially insured patients for whom
preauthorization with third party payers was attempted, 2/12 (17%) were denied
preauthorization. Conclusions: The current data support the impact of CCCP
testing in an adult oncology population where at least one third of the testing is not
reimbursed. We expect that additional patients will be recruited to clinical trials based
on CCCP results as they fail standard therapy. This data will inform reimbursement
patterns and indications for expanded genomic tumor testing.
ST73. RNA Sequencing Using Non-Cell Block Cytology Slides and FFPE
Specimens Augments a DNA-Based Next Generation Sequencing Panel for
Non-Small Cell Lung Cancer
K.J. Hampel, J.D. Armstrong, N. Sidiropoulos
University of Vermont Medical Center, Burlington, VT.
Introduction: Gene fusions are a significant category of clinically relevant driver
mutations in non-small cell lung cancer (NSCLC). Assessment with florescence in
situ hybridization (FISH) and immunohistochemistry is common, but direct
sequencing of the genes, though challenging, is adventageous. RNA next generation
sequencing (NGS) is validated to complement a previously validated DNA-based
NGS panel, to interrogate NSCLC samples for ALK, RET and ROS1 fusions.
Presented are informatics and design considerations for a validation study that not
only assesses RNA NGS, but also evaluates it as a complementary process to DNA
NGS for single nucleotide variants and Insertion/deletions. Described is the
experience with both FFPE and non-cell block cytology samples that can be used for
RNA NGS. Methods: Amplicon sequencing interrogates select exons
in ALK, RET and ROS1. Amplicon libraries were prepared using ArcherDx
FusionPlex reagents and pooled libraries were sequenced on an Illumina MiSeq.
Bioinformatics analysis, results review and annotation was performed using an
ArcherDx analysis suite hosted in the Clinical Genomicist Workspace (CGW) at
PierianDx. Results: Quality parameters of RNA extracted from FFPE tissue and
non-cell block cytology samples were evaluated to establish metrics with consistently
sufficient quality to be used in a clinical NGS workflow. RNA extracted from non-cell
block cytology samples is higher quality than that extracted from FFPE tissue as
qualified by reverse transcription qPCR. In addition, detection of gene fusions is
enhanced by the expression of the non-oncogenic fusion partner. This resulted in a
limit of detection (LOD) for RNA NGS that is 2-fold lower than the LOD of the
complementary DNA NGS assay. The clinical validation of previously assayed
clinical samples resulted in RNA NGS identifying one clinically relevant fusion event
that was missed by traditional FISH due to the close chromosomal proximity of the
fusion partners. Data generated during the first 4 months of clinical testing is
additionally reported. Conclusions: RNA NGS for fusion events in NSCLC can be
successfully integrated into a clinical NGS workflow where FFPE and non-cell block
cytology samples are routinely used. Validating RNA NGS for gene fusion detection
streamlines tissue utilization of small biopsies, easily enables integrated genomic
reporting, and adequately interrogates fusion events of ROS1 intron 31, a site of
clinically actionable fusions that is difficult to consistently detect with DNA-based
NGS. Analysis and reporting using the custom built RNA Sequencing solution within
CGW streamlines sign-out and facilitates integration of RNA and DNA genomic
results into unified reports.
ST74. The Importance of Tumor-Normal Sequencing For Accurate Somatic
Variant Determination in Genomic Cancer Testing
T. McDaniel, T. White, A. Kurdoglu, J. Lobello, G.D. Basu, T. Izatt, R. Halperin, T.
Royce, D. Craig, T. McDaniel
Translational Genomics Research Institute, Phoenix, AZ.
Introduction: The characterization of tumors with large next-generation sequencing
gene panels is growing in popularity as a method for identifying therapeutic options
for cancer patients. Several papers in the literature have raised the question as to
whether the laboratory can effectively perform such analysis by sequencing only
tumor DNA, or if it is necessary to make tumor calls only after subtracting the
sequence of a matched normal sample. Methods exist for in silico identification of
somatic variants in tumor-only sequences, for example by filtering away common
germline variants found in public single nucleotide polymorphism (SNP) databases; if
such methods are effective at filtering out benign background variation, then tumoronly sequencing would be a useful option due to simplified logistics and cost savings
in sequencing. If, however, such methods are not effective, for example due to the
preponderance of rare germline SNPs, then applying them could result in benign
germline variation being wrongly reported as actionable somatic mutations.
Methods: DNA was extracted from formalin-fixed paraffin-embedded tumors and
matched normal from blood and sequenced using a 562-gene panel employing a
hybrid capture technology (Agilent, Inc.) using custom-designed oligonucleotide
baits. Libraries were sequenced on Illumina HiSeq 2500 machines then base called
and aligned using standard methods. Somatic variant calling was performed either
using paired tumor and normal sequences and a joint variant caller or from tumor
sequences alone followed by filtering against SNPs in the public databases. Variants
were then assigned therapeutic actionability using a literature-based rules engine
database. Results from the 2 analysis methods were compared and false positive
rates, determined as somatic calls present in the tumor-only pipeline that were
absent in the joint tumor-normal pipeline, were then tallied for each report. Results:
In an analysis of 115 tumors, an average of 3.8 false positives per report were
1025
AMP Abstracts
generated using the tumor-only analysis. Overall, 103 of 115 (98.2%) reports had at
least one false positive when tumor-only analysis was applied. Conclusions:
Consistent with results published by other groups, these results suggest that tumoronly methodology is prone to widespread over-reporting of tumor mutations due to
confusion with likely benign germline variants.
ST75. Added Value of Non-Cell Block Cytology Slides Compared to FormalinFixed Paraffin-Embedded (FFPE) Specimens for Targeted Genomic Profiling of
Solid Tumors
K. Hampel, A. Balla, M.P. Cameron, N. Sidiropoulos
The University of Vermont Medical Center, Burlington, VT.
Introduction: Non-cell block (NCB) cytology samples are frequently collected at the
time of biopsy of advanced malignancies and have been demonstrated to perform
well when subjected to downstream molecular diagnostics. This sample type
however is subjected to a myriad of cytopreparatory methods and remains
underutilized due to the complexities of comprehensive validation. Reported is the
experience of clinical next generation sequencing (NGS) using NCB samples and
the added value of this practice for the laboratory, clinicians and patients.
Methods: DNA extracted from NCB cytology samples representative of 3
cytopreparatory types (Thin Prep, PAP and MGG) was subjected to hybrid-capturebased DNA NGS to interrogate the exon sequences of 29 clinically-actionable
genes. Hybrid capture was performed using Agilent Sure Select and pooled libraries
were sequenced on an Illumina MiSeq. Bioinformatic analysis, variant review and
annotation were performed by the Clinical Genomicist Workspace
(PierianDx). Results: DNA was extracted for 203 clinical tests: NCB cytology slides
(52%), FFPE surgical biopsies (39%), cytology cell blocks (6%), and a combination
of specimen types (2.4%). The number of sequencing reads required to achieve 1fold coverage of the 134 kilobase pair target capture region demonstrates that noncell block cytology samples had a 20% lower read requirement than FFPE tissue
specimens. In addition, analysis of nucleic acid quality shows DNA from NCB
cytology samples is consistently high quality permitting lower NGS input quantity and
protocols for extraction, quantification and qualification of this DNA is less timeconsuming than that for FFPE samples. Utilization of NCB cytology samples
mitigated the number of cases that would have otherwise been unsatisfactory for
NGS and ancillary tests complementary to NGS (florescence in situ hybridization,
PD-L1 immunohistochemistry, and centralized laboratory testing for clinical trials).
Conclusions: Validation and utilization of NCB cytology samples for NGS has
added value for clinical NGS. This sample type consistently exhibits high quality
nucleic acid that results in lower input quantity and supports more effective use of
small biopsy samples for an ever expanding portfolio of tests ancillary to morphologic
evaluation. This sample type permits efficient clinical NGS workflow and mitigates
samples otherwise deemed inadequate for ancillary testing. The latter adds value to
clinical care and the patient experience by decreasing additional biopsies and the
morbidity and inherent delays in care inherent therein.
ST76. The Assessment by Next-Generation Sequencing of FFPE Derived
Tumor DNA Using an Ovarian Cancer and a Custom Solid Tumor
Hybridisation-Based Enrichment Panel Approach
J. Chan1, S. Eckert1, L. Georgieva1, W. Wright2, G. Speight1
1Oxford Gene Technology, Begbroke, United Kingdom; 2Oxford Gene Technology,
Inc., Tarrytown, NY.
Introduction: Solid tissue tumour biopsies are typically archived as formalin-fixed,
paraffin embedded (FFPE) blocks. This approach unfortunately can significantly
compromise the quality of the resulting extracted nucleic acids. Various approaches
are utilised to overcome these issues, one of which is using the SureSeq FFPE DNA
Repair Mix, prior to NGS library preparation. In addition hybridisation-based
enrichment can be more amenable than PCR-based enrichment for poorer quality
(shorter fragments) DNA inputs. Here we individually assess the utility of the repair
mix and compare the uniformity of coverage generated by the 2 main enrichment
approaches. We then assess the hybridisation approach further using a
commercially available Quantitative Multiplex Formalin Compromised Reference
Standard from Horizon with a custom SureSeq myPanel. Methods: We utilised the
SureSeq FFPE DNA Repair Mix, Library Preparation Kit and Ovarian Cancer Panel
(BRCA1, BRCA2, ATM, TP53, ATR, NF1 and PTEN) from Oxford Gene Technology
(OGT) to assess the impact of the repair mix on a set of FFPE derived DNA. The
DNA samples used had a broad range of DNA quality (as defined by DIN number).
We then compared the uniformity and coverage of BRCA1 and BRCA2 with an
alternative PCR-based enrichment method. We also then created a custom solid
tumour panel to analyse the commercially available formalin compromised reference
sample. Results: Using the FFPE repair mix, we consistently achieved greater than
1.5 fold improvement in mean target coverage and on target rates. The lower the
DIN number the greater the fold improvement. Using the Ovarian cancer panel with
16 samples per MiSeq lane we achieved (post repair) a mean target coverage of
800x (de-duplicated) and a uniformity of greater than 98% of bases covered at 20%
of the mean. When we compared this uniformity (BRCA1 and BRCA2 only) to a
PCR-based enrichment, this was only 91% of bases covered at 20% of the mean.
The custom solid tumour panel detected 100% of all known variants in the
compromised reference standard, including the EGFR variant (T790M) at an allele
frequency of 1.00%. Conclusions: We have shown the successful detection of key
variants in formalin compromised samples as well as the superior uniformity of
1026
coverage generated from FFPE derived tissue samples in comparison to PCR-based
enrichment. The utility of this hybridisation-based enrichment approach now permits
the analysis of somatic variation in these (and other) key genes associated with solid
tumours from difficult to use samples.
ST77. A Comparison of EGFR Mutation Detection between Targeted NextGeneration Sequencing and Real-Time PCR Assay in Non-Small Cell Lung
Cancer
Y. Cheng, M.A. Jakubowski
Cleveland Clinic, Cleveland, OH.
Introduction: Despite the poor prognosis of NSCLC, clinically responsive drugs
targeting Epidermal Growth Factor Receptor (EGFR) gene are approved for treating
non-small cell Lung cancer (NSCLC), which accounts for nearly 80% of all lung
cancer. The gain-of-function or driver EGFR mutations have been well documented
to constitutively activate the gene’s kinase function that leads to uncontrolled cell
proliferation in cancer. Several EGFR inhibitors targeting either primary mutations
or a second-site, acquired drug-resistant variant (resistant to erlotinib or gefitinib)
have been developed and conferred with increased responses in treating NSCLC
patients. Allele specific real-time PCR, therascreen (QIAGEN, Manchester UK), has
been commonly used to detect disease-associated EGFR sequence changes in
NSCLC. It has the advantages of quick turnaround time, using suboptimal genomic
DNAs, and high detection sensitivity. However, therascreen is limited in identifying
low level T790M, a common EGFR drug-resistant mutation. Methods: NSCLC
genomic DNA was extracted from formalin fixed paraffin embedded tissues and fine
needle aspirate specimens. Mutation hotspots of a 50-gene panel corresponding
to AmpliSeq cancer hotspot panel v2 was multiplex amplified, adaptor ligated, and
analyzed using the NextGENe (SoftGenetics, State College, PA) bioinformatics
tool. Variants in EGFR mutation hotspots with greater than 2% allele fractions were
called and examined. The therascreen assay was performed according to the
manufacture instruction. The amplification steps were performed in a Rotor-Gene Q
5plex HRM instrument. The proper ΔCt values used for calling each EGFR mutation
were based on those indicated in the manufacture manual under section of
“EGFR mutation data analysis”. Results: In this presentation, we show next
generation sequencing (NGS) consistently determined T790M variants at 2-5% allele
fractions. The ability of NGS to identify variants at low allele fractions is especially
important for detecting mutations that were acquired due to the initial anti-cancer
therapy. Other examples of EGFR variants that may be more accurately determined
using NGS than therascreen are also discussed. Conclusions: We conclude that
deep sequencing of NGS and real-time PCR approaches is complementary to
accurately identify EGFR mutations.
ST78. Analysis of Active Oncogenic Signal Transduction Pathways in Ovarian
Cancer
P.v. Wiel1, J. Helleman2, E.d. Biezen-Timmermans1, J. Wrobel1, H.v. Ooijen1, E.
Berns2, A.v. Stolpe1
1Royal Philips Eindhoven, Noord Brabant, The Netherlands; 2Erasmus MC,
Rotterdam Zuid-Holland, The Netherlands.
Introduction: Patients with ovarian cancer in general have a bad prognosis due to
frequently advanced disease at diagnosis associated with limited therapeutic
options, and recurring disease. Use of targeted drugs in ovarian cancer is not very
advanced, one major reason being insufficient knowledge on tumor driving signaling
pathways. We have developed novel assays to identify activity of oncogenic signal
transduction pathways in cancer tissue, based on Bayesian reasoning models which
infer from mRNA levels of pathway transcription factor-specific bonafide target genes
the probability that the corresponding signaling pathway is in the active mode
(Cancer Res 2014 Jun 1;74(11):2936-45). Methods: Affymetrix U133 Plus 2.0
microarray analysis was performed on fresh frozen samples from a cohort of patients
with high grade serous (HGS) ovarian cancer, consisting of 37 patients, some with
samples from multiple different metastatic location, resulting in a total of 90 samples
from primary untreated ovarian carcinoma tissues. The Affymetrix models were also
ported to a RT-qPCR based platform for which we developed 96 real-time RT PCR
assays. The PCR test was performed on the same samples. Subsequently activity of
the AR, ER, PI3K-FOXO, Hedgehog (HH), TGFbeta, Wnt and NFkB pathway was
determined using the computational pathway models. In addition a number of
publically available Affymetrix datasets were analysed with respect to pathway
activity. Results: Two pathway combinations were identified: NFkB active (n=23)
and PI3K- HH-TGFbeta pathway active (n=13). In a 36M2 serous ovarian cancer cell
line cisplatin treatment induced activity of NFkB suggesting that this pathway
combination presents a less aggressive cancer type. Spatial heterogeneity between
local metastatic locations of the same patient was observed mainly for TGFbeta and
HH pathway active tumors. One patient with 12 samples, all with the
PI3K/HH/TGFbeta signature, was primary resistant to chemotherapy and died within
6 months. In an ovarian cancer cell line (A2780) HH pathway activity was associated
resistance to cisplatin (GSE15709). ER and PI3K pathways are thought to play a role
in ovarian cancer. In one out of 2 patients of which samples were available before
and after chemotherapy, PI3K pathway activity appeared after standard
chemotherapy. Conclusions: While too few patients to reach statistically significant
results, results suggest that the PI3K-HH pathway activity combination may be
associated with a worse prognosis and potential resistance to chemotherapy. In
addition the pathway models were successfully ported to the qPCR platform. In the
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
future this will allow also FFPE material to be analyzed.
ST79. Circulating Cell Free DNA (cfDNA) Isolated and Amplified from the
Plasma of Pancreatic Cancer Patients as Reference Material for ctDNA Assays
Y. Konigshofer1, M. Egyud2, T. Godfrey2, B. Anekella1, H. Safran3, J. Wands3, R.
Garlick4
1SeraCare Life Sciences, Gaithersburg, MD; 2Boston University School of Medicine,
Boston MA; 3The Warren Alpert Medical School of Brown University, Providence,
RI; 4SeraCare Life Sciences, Milford, MA.
Introduction: Circulating tumor DNA (ctDNA) has great potential for early cancer
detection and for therapeutic or resistance monitoring. However the analytical
challenges to consistently detect and quantify oncogenic variants is quite high as
ctDNA often represents just a few copies within 300-3,000 genome equivalents of
cfDNA per ml. Assays whether based on PCR or NGS need to be very efficient and
have very high sensitivity and specificity. Absolute standards, produced in quantity,
are required to optimize ctDNA recovery, variant detection and for analytical
comparison of different methods because amounts of native donor-derived cfDNA
are severely limited. The goals of this study are to evaluate a novel circulating cell
free DNA amplification method and to determine if the method produces reference
material in sufficient quantity and quality for multiple analyses while maintaining
original genotype profile. Methods: Two patients with late stage pancreatic ductal
adenocarcinoma were recruited. Patient derived ctDNA was prepared from 10 ml
blood draws collected into Streck Tubes. The cfDNA in the plasma fraction was
extracted using the QIAGEN QIAamp Circulating Nucleic Acid kit and the cfDNA
therein was used to generate a library that was subsequently amplified and used to
prepare reference material. A contrived biosynthetic mixture of genomic DNA from
GM24385 cells and synthetic spike-in sequences for 9 variants was also prepared at
~0.6% VAF for each variant. Variant allelic frequencies were measured by digital
PCR (dPCR), by the SimSen-Seq NGS assay and by several other NGS-based
assays. SimSen-Seq is able to detects variants < 0.1% allele frequency with input
DNA at <50 ng. Results: The Streck Tube collection, cfDNA isolation and library
synthesis yielded about 2,000 amplifiable genome equivalents DNA for each patient
sample. Subsequent amplification yielded micrograms of DNA. The dPCR and
SimSen-Seq assays were able to detect KRAS p.G12R at ~3.4% VAF for patient 1
and KRAS p.G12D at ~0.6% VAF for patient 2. As a control, both assays were able
to specifically detect synthetic spiked in KRAS p.G12D at ~0.6% VAF in the
multiplexed reference material. Conclusions: Circulating cell free DNA from late
stage patients with pancreatic ductal adenocarcinoma were individually isolated and
amplified to yield sufficient DNA for numerous replicate assays. The method
produced large quantities of cfDNA and retained a cfDNA-like size distribution.
KRAS p.G12R and KRAS p.G12D were detected in the amplified samples
demonstrating that this sample amplification method has the potential to reproduce,
at large scale, rare ctDNA patient samples for use in assay validation, method
comparisons and proficiency testing.
ST80. Performance Comparison of Commercially Available Gene Fusion Next
Generation Sequencing Panels
K.E. Bartow, S. Springborn, A.C. Mackinnon
Medical College of Wisconsin, Milwaukee, WI.
Introduction: Fusion genes, described as genetic aberrations resulting from
chromosomal translocations, inversions, or interstitial deletions, play a vital role as a
diagnostic tool with high prognostic and treatment value. With the recent emergence
of targeted treatment options for oncogenic gene fusions, the importance of gene
fusion identification and characterization in cancer patients has grown. Although
fluorescence in situ hybridization (FISH) is the gold standard, the need for higher
throughput methods with increased sensitivity has driven the development of next
generation sequencing assays (NGS) capable of detecting gene fusion events. To
further investigate the capabilities of gene fusion detection using NGS, we performed
a side by side comparison of 2 commercially available gene fusion
panels. Methods: A cohort of 14 samples was used for the direct comparison of 2
fusion panels: Archer FusionPlex CTL (ArcherDx) and TruSight RNA Fusion
(Illumina). RNA was extracted from formalin-fixed, paraffin embedded (FFPE) cancer
specimens with previously known fusions events. Libraries were prepared in parallel
using the Archer®FusionPlex CTL Kit and the TruSight RNA Fusion kit. Sequencing
was performed on the Illumina MiniSeqs platform using Illumina’s High Output
Sequencing kits. Sequence data was analyzed with the Local Run Manager RNA
Fusion module or the Archer Analysis 5 package for the TruSight RNA Fusion panel
and the Archer FusionPlex CTL panel, respectively. Fusions detected by either NGS
panel were compared to FISH. Results: The FusionPlex CTL panel detected 12 of
the 14 cases containing either ALK or ROS1 gene fusions. Unexpectedly, one ROS1
positive case by FISH had 2 additional ALK fusion events detected by NGS. Eleven
of the 14 samples were sequenced using the TruSight RNA Fusion panel. The
TruSight RNA Fusion panel detected 6 of the 11 fusions events identified by the
FusionPlex CTL panel. Two samples with either an ALK or a ROS1 fusion were
missed by both panels. Conclusions: Preliminary results indicate that both NGS
panels had lower detection rates compared with the original FISH results. They also
varied in concordance for detection of the same gene targets. Possible causes for
the lower detection of fusion events with NGS-based methods include sample quality
due to specimen type (FFPE), the age of the sample, and RNA stability, which is
adversely affected by enzymatic and/or chemical modifications during specimen
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
processing. Additional testing of a larger cohort of samples and sample types will
further establish the suitability of implementing NGS-based detection methods in the
clinical laboratory.
ST81. Tumor in Normal or Normal in Tumor: What to Do When Somatic
Mutations Are Detected in “Normal” Germline Control Used for NGS-Based
Targeted Somatic Mutation Testing
A. Yemelyanova, M.J. Routbort, H. Chen, A. Rashid, S. Roy Chowdhuri, A.J.
Lazar, J. Manekia, M. Mohammad, G. Mantha, R.R. Singh, R. Luthra, J.L.
Medeiros, K.P. Patel
University of Texas, M.D. Anderson Cancer Center, Houston, TX.
Introduction. Analysis of somatic mutations in solid tumors using targeted NGSbased assays has become a part of routine oncology practice. Extended gene
panels require paired tumor-normal analysis to increase confidence of somatic calls
by subtracting germline variants. Peripheral blood (PB) is a preferred source of
germline DNA; formalin-fixed paraffin embedded (FFPE) normal tissue can be used
as well. Detection of somatic variants within “normal” DNA sample is occasionally
encountered and presents a challenge in the interpretation of results. Herein, we
report 14 cases where somatic mutations were detected in germline sample.
Methods. We searched our archives for cases with known somatic mutations in both
tumor and normal DNA samples. Mutational analysis of FFPE tumor samples was
performed using the Oncomine cancer panel (ThermoFisher Scientific), which allows
for the detection of somatic mutations in 128 genes and selected copy number
variations in 49 genes. The assay requires normal DNA for subtraction of germline
variants which was obtained from PB or FFPE. Results. The study group included
14 cases: 1 melanoma and 13 adenocarcinomas involving colorectum (n=5), lung
(n=4), endometrium (n=1), cervix (n=1), pancreatobiliary tract (n=1), and urothelium
(n=1). DNA for germline control was derived from FFPE (n=8) or PB (n=6). Control
FFPE tissue included an alternative paraffin block (n=5) or dissected non-neoplastic
tissue from the same block as the tumor (n=3). Somatic mutations in PIK3R1, PTEN,
TP53, ERBB2, APC, PIK3CA, KRAS, CTNNB1, FBXW7, NRAS, TET2, TSC1,
ATM, and SMAD4 were observed in control DNA extracted from FFPE at allelic
fractions (AFs) less than 5%. The AFs of mutations in these genes in the
corresponding tumor samples ranged from 7-80%. In cases in which PB was used
as a source of germline DNA, mutations in FLT3, DNMT3A, JAK2, EGFR (p.T790M),
and ERBB4 were detected; the AFs ranged from 3-48% and exceeded or equaled
those detected in the tumor in EGFR, JAK2, and DNMT3A. Conclusions. Nonneoplastic FFPE tissue, a commonly available source of germline DNA, may contain
morphologically an apparent tumor that results in low level contamination of control
DNA. These somatic mutations are present at AFs usually substantially lower than in
the corresponding tumor samples. Alternatively, mutations seen in hematologic
conditions (known or incidental) can be present in both PB and leukocytes in FFPE
blocks. Germline mutation EGFR (p.T790M) has been described and may be
detected in cases of lung adenocarcinoma. Detection of somatic mutations in DNA
used as germline control may complicate analysis of NGS data. Awareness of such
occurrences and careful evaluation of normal DNA source is important for accurate
reporting of sequencing results.
ST82. Development and Validation of a Genomic Classifier to Predict
Aggressive Prostate Cancer from Diagnostic Biopsy Tissue
E. Davicioni1, T.J. Triche2, E.A. Gibb1, Q. Wang1, N. Erho1, M. Alshalafa1, V.
Choerung1, K. Youseffi3, A.E. Ross2, T. Bismar3, E. Davicioni2
1GenomeDx Biosciences Inc., La Jolla, CA; 2GenomeDx Biosciences Inc., San
Diego, CA; 3University of Calgary, Calgary, Alberta, Canada.
Introduction: Increased clinical adoption of genome-wide expression analysis in
urology practice provides an opportunity to prospectively collect massive amounts of
genomic data. We have previously shown how these data can be used to further
refine our understanding of the genomic landscape of prostate cancer on a
population level. Here, we show how these data may also be used to develop a
novel genomic classifier that further improves prediction of aggressive prostate
cancer. Methods: Genome-wide expression profiles from 11,170 localized PCa
patients were extracted from the Decipher GRID registry (NCT02609269). The
training cohort of radical prostatectomy (RP) tumor sample profiles included 545
cases from a retrospective cohort with long-term follow up and 1,939 from
prospective clinical utilization of Decipher (GC1, 22 biomarkers). A generalized linear
model was trained to predict a composite genomic-pathologic endpoint (GC1>0.7
and Grade Group>3) in the prospective cohort resulting in a final 60 gene model
(GC2). Model characterization was performed in an additional 6,739 and 1,694
prospective RP and biopsy samples. Model validation was conducted on 253 biopsy
samples from a retrospective cohort of men treated with RP. GC1 and GC2 models
were compared using area under the ROC curves, uni- and multi-variable logistic
regression for the prediction of high-grade (Grade Group ≥3), lymph node
involvement (LNI), biochemical recurrence (BCR) and metastasis. Results: GC2
achieved an AUC of 0.92 on 10-fold cross validation to predict the composite
endpoint. In prospective biopsy the AUC for predicting high-grade disease by
independent pathology review among the GG1 tumors were 0.64 and 0.66 for GC1
and GC2; compared to 0.57 and 0.52 for age and PSA, respectively. In the
retrospective biopsy validation cohort the GC1 and GC2 models had an AUC for
predicting LNI of 0.73 and 0.77, for BCR 0.79 and 0.86 and for metastasis 0.70 and
0.82, respectively. Addition of the models to NCCN clinical risk groups increased the
1027
AMP Abstracts
AUC of NCCN from 0.62 to 0.76 and 0.83 for prediction of the metastasis end-point.
In multivariable analysis adjusting for age and NCCN risk groups, only the genomic
models remained significant risk factors for predicting metastasis with hazard ratios
of 1.33 (p=0.013) and 1.68 (p<0.001) per 10% increase in GC1 and GC2 score,
respectively. Conclusions: This study has demonstrated for the first time how
genomic data collected prospectively through the clinical use of a first-generation
genomic test may be used to train a new and improved second generation risk
model for prediction of aggressive disease from prostate needle biopsy specimens.
ST83. Uncommon and Novel BRAF Fusions Detected by Targeted Next
Generation Sequencing and Their Impact on Clinical Management
V.A. Paulson1 ,3, M. Longoni2 ,3, L. Le2, M.R. Toups2, V. Nardi2, J.K. Lennerz2, A.
Bardia2, J.V. Cohen2, A.J. Iafrate2, C. Azzoli2, D. Dias-Santagata2
1Brigham & Women's Hospital, Boston, MA; 2Massachusetts General Hospital,
Boston MA; 3Boston Children's Hospital, Boston, MA.
Introduction: The role of kinase inhibitors has become increasingly prominent in the
treatment of neoplasia, as kinase signaling pathways drive many of the phenotypic
hallmarks of cancer and frequently confer oncogenic dependency. BRAF gene
rearrangements affect multiple tumor types and involve a wide array of fusion
partners. They are rare compared to the more prevalent BRAF V600E mutation and
are resistant to treatment with approved inhibitors such as vemurafenib and
dabrafenib. Preclinical studies and a limited number of clinical reports suggest that
tumors harboring BRAF rearrangements may instead be sensitive to treatment with
MEK inhibitors. Unbiased clinical approaches, capable of identifying not only wellcharacterized translocations but also novel gene rearrangements, have the ability to
identify rare potentially targetable fusion events across multiple tumor types.
Methods: From September 2013 to May 2017, over 3,500 tumor specimens from
patients with advanced disease were tested for gene rearrangements as part of
routine clinical cancer genotyping at the Massachusetts General Hospital. Total
nucleic acid from formalin-fixed paraffin embedded (FFPE) tumor tissue was tested
using a commercially available next generation sequencing assay based on
anchored multiplex polymerase chain reaction (PCR) (ArcherDx). This assay, which
targets 52 genes recurrently rearranged in cancer, is designed to identify fusion
events where only a single gene partner is known. Results: We detected 12
rearrangements involving BRAF fused to 9 unique partners, 4 of which have not
been previously described, including: KIAA1549 (pilocytic astrocytoma, 2
cases), AGK and SND1 (thyroid carcinoma, one and 2
cases respectively), TRIM24 (lung adenocarcinoma and duodenal
adenocarcinoma), NRF1 (colon adenocarcinoma), SRPR (pancreatic
adenocarcinoma), CCT8 and DIP2B (cutaneous melanoma, one case each),
and SCAPER (invasive ductal breast carcinoma). All fusions preserved the intact
kinase domain of BRAF and removed the auto-inhibitory Ras binding regulatory
domain. BRAF pathway inhibition was considered for 2 patients, and ultimately
pursued in the patient with a TRIM24-BRAF fusion-positive lung adenocarcinoma.
This patient has been treated with trametinib monotherapy (MEK inhibitor) for 3
months and her disease remains stable with a modest reduction in tumor burden.
Conclusion: The BRAF fusions described in this report, in addition to representing
uncommon and novel mechanisms of kinase activation, are also potential
therapeutic targets. Expansion of clinical molecular diagnostic panels to detect fusion
events where only one partner is known will be necessary to identify a significant
number of patients who may benefit from targeted kinase inhibitors.
ST84. Head to Head Comparison of Archer VariantPlex/FusionPlex Solid Tumor
and the Illumina TruSight Tumor 170 Assays
O. Rouhi1, T. Schneider2, H. Jones1, J. Hauenstein1, H. Tillson3, B. Cooper3, B.
Lutz3, K. Tucker4, J. Mellott4, A. Kuraishy4, J. Kissiedu1, V. Avadhani1, K. Hanley1, X.
Li1, K.R. Magliocca1, G. Sica1, C. Hill1, M. Rossi1
1Emory University Hospital, Atlanta, GA; 2EGL Genetics, Tucker, GA; 3Archer,
Boulder, CO 4 Illumina, Atlanta, GA.
Introduction: Use of combined DNA and RNA multiplex gene panels for clinical
diagnostics is gaining favor for laboratories that want to consolidate the detection of
mutations, copy number and structural rearrangements into more comprehensive
workflows. Recently, the Illumina TruSight Tumor 170 (TST170) panel and Archer
VariantPlex/FusionPlex assays have become available for laboratories that use
Illumina NextSeq sequencers. The TST170 kit is a dual DNA and RNA hybridcapture assay which targets 151 genes for SNV and indels, 55 genes for fusions and
splice variants, and 59 gene amplifications. The Archer VariantPlex Solid Tumor
DNA assay queries SNVs, indels and copy number variation in 67 genes and can be
paired with the FusionPlex Solid Tumor RNA kit which detects fusions and splice
variants in 50 genes. We chose to evaluate the Illumina TST170 and the Archer
VariantPlex/FusionPlex assays head-to-head with a diverse group of 20 FFPE tumor
specimens for their ability to detect clinically actionable and diagnostically relevant
abnormalities. Methods: Formalin-fixed paraffin embedded (FFPE) tissue sections
were prepared from 5 lung adenocarcinoma, 1 lung sarcomatoid tumor, 2
glioblastoma, 3 head and neck squamous cell carcinoma, 2 melanoma, 2 colon
adenocarcinoma, 1 ovarian steroid cell tumor, 1 uterine carcinoma, and 1 breast
invasive ductal carcinoma. Genomic DNA and total RNA were prepared from a
minimum of 4 unstained serial sections that met established tumor cellularity
requirements and nucleic acid quantity and quality per Qubit and Agilent
BioAnalyzer. Aliquots were apportioned for respective library preparations. All
1028
libraries were prepared over a 3 week period and sequenced using an Illumina
NextSeq. Data analysis was performed using Illumina BaseSpace tools and Archer
analysis software. Results: We are still evaluating the entire dataset, but have
identified strong concordance in the detection of fusions (ex. ALK/EML4,
FGFR3/TACC3 and ROS1/SDC4), SNVs and amplifications in 80 overlapping genes
in the Archer and Illumina TST170 assays compared with FISH, copy number array
(Affymetrix OncoScan) and Illumina TruSight Tumor 26. Summaries of our findings
and the appropriateness of these assays for the tumor types that were sequenced
will be discussed. Conclusions: Although our results are preliminary, we provide
evidence that dual DNA and RNA assays appropriate for Illumina NextSeq
sequencing instruments can be performed in academic molecular pathology
laboratories in a 7-10 day timeframe. These assays provide a more comprehensive
assessment of driver mutations in a variety of solid tumor types and provide a new
series of data management challenges for laboratories with limited bioinformatics
resources.
ST85. Use of Synthetic Mutation Standards to Bolster Validation of DNA Based
NGS Panels for Detection of Translocation and Large Indels
P.M. Rindler, A. Bolia, R.R. Bastien, R.L. Margraf, J.A. Raney, A.A. Hall, S.
Hellwig, G. Deftereos, A.H. Grossmann, A.P. Matynia, P. Bernard, M. Bronner, B.
Kennedy, L.V. Furtado, E.P. Gee, K.M. Gligorich
ARUP Laboratories, Salt Lake City, UT.
Introduction: We have developed a 48-gene hybrid-capture solid tumor NGS panel
to detect clinically actionable SNVs, indels and translocations in select introns from
formalin fixed, paraffin embedded (FFPE) extracted DNA. Limited access to patient
samples with a variety of orthogonally validated mutations including translocations
and large indels is a consistent obstacle for assessing test accuracy. Utilization of
synthetic DNA mutations overcomes this limitation and accelerates the development
and validation of DNA based NGS panels by facilitating robust training of a sensitive
bioinformatics pipeline and rigorous evaluation of assay limits of detection (LOD).
Methods: Multiple double stranded oligonucleotides harboring clinically relevant
translocations in ALK, NTRK1, RET and ROS1 fused to a variety of known partners
were generated. Numerous synthetic indels were generated with insertions ranging
from 1 to 88 bp or deletions ranging from 1 to 2829 bp across multiple genes
including large deletions in MET exon 14 and various indels within KIT exon 11. All
synthetic mutation containing oligonucleotides were individually diluted and spiked
into 50 ng of normal FFPE extracted DNA prior to shearing and NGS library
preparation. Digital droplet PCR (ddPCR) assays for all indels and 4 of the
translocations (one in each targeted gene) were developed to confirm samples at
specific variant allele frequencies used to optimize the bioinformatics pipeline and to
test assay LOD. For the remaining 40 synthetic translocations, one universal ddPCR
assay was developed to generate samples for accuracy testing. Sequencing was
performed using an Illumina NextSeq 500, 2x150bp paired-end reads. FASTQs
were aligned with BWA and variant calling was performed using Lofreq, Scalpel,
Manta and DELLY. Results: All 40 synthetic translocations used for accuracy testing
were detected. Four synthetic translocations were detected down to 2.5% variant
allele frequency. Seventeen of the synthetic indels sequenced to date were all
detected at 5% variant allele frequency including all MET exon 14 deletions up to 2.8
kb. Our calculated positive percent agreement for synthetic mutations is 100%
between our NGS results and orthogonal ddPCR data for all synthetic mutations
tested so far. Conclusions: Through the use of synthetic DNA mutations, we have
bolstered the development and validation of a focused NGS cancer panel designed
to detect translocations and large indels by targeted capture with high sensitivity. In
particular we have demonstrated the ability to detect large indels in MET which are
clinically relevant in lung cancer as well as large indels in KIT which are clinically
relevant in melanoma and GIST.
ST86. Characterization of Copy Number Alterations in Circulating Tumor Cells
from Metastatic Prostate Cancers Using a Novel Enrichment Platform and
Genome Wide Next-Generation Sequencing
G. Morrison1, C. Sims2, S. Koo3, A. Khurana2, M. Ting2, A. Ghosal3, C. Forcato4, A.
Ferrarini4, V. Sero5, G. Busson4, N. Manaresi4, M. Moore3, F. Bischoff5, P. Cotter3, S.
Gunn3, A. Goldkorn6
1University of Southern California Norris Cancer Center, Los Angeles,
CA; 2PacificDx, Irvine, CA; 3ResearchDx, Irvine, CA; 4Menarini Silicon Biosystems,
Bologna, Italy; 5Menarini Silicon Biosystems, San Diego, CA; 6University of Southern
California, Los Angeles, CA.
Introduction: Prostate cancer is the second most prevalent cancer (est.161,000
new cases in 2017) and third leading cause of cancer related deaths (est. 26,700 in
2017) in men. As the disease progresses on second line therapies it is paramount to
precision medicine’s success to monitor treatment driven genetic heterogeneity and
evolution of therapy resistant clones. Circulating Tumor Cells (CTCs) shed into the
bloodstream provide a means to interrogate this as both primary tumor and distal
metastatic sites contribute CTCs in advanced disease. Here we describe a method
for isolation of metastatic prostate CTCs from CellSearch (Menarini Silicon
Biosciences; MSB) enriched Veridex cartridges using the DEPArray system (MSB)
followed by single cell low pass genome wide copy number analysis. Methods:
Whole peripheral blood of metastatic prostate cancer patients was enriched for
CTCs using the CellSearch system under an IRB-approved protocol, and 5 samples
(>5 CTCs) were selected for further study. The DEPArray system was used to
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
identify and isolate single CTCs and paired white blood cells from enriched Veridex
cartridges per MSB’s standard protocol. In addition, single LNCaP prostate cancer
cells were micro-manipulated into individual tubes and used as positive assay
controls. Recovered single cells were whole-genome amplified with the Ampli1 WGA
kit (MSB) and QC was performed by gel-electrophoresis. Ampli1 LowPass kit (MSB)
was used to prepare NGS libraries for absolute CNA profiling by low-pass WGS on
Ion PGM System. Genome-wide copy number analysis was performed using FREEC
with ploidy set to 2. Results: For the patient sample cohort, an average of 18 CTCs
were isolated per patient sample after dielectric manipulation on the DEPArray. Copy
number alterations were observed in CTCs from 3 of 5 patient samples that were
characteristic of high-grade, advanced prostate carcinoma. These included focal
amplification of the androgen receptor on chromosome X, as well as characteristic
chromosomal gains and losses on chromosome 8p and 8q respectively.
Reassuringly, the LNCaP cell line’s known genomic heterogeneity was also
visualized by individual CTC genome characterization. Conclusions: We have
demonstrated the application of a unique sorting tool and low-pass genome wide
copy number determination in prostate cancer. This approach enhances the current
CTC enumeration methodology by incorporating downstream molecular (NGS)
analysis. Furthermore, this non-exhaustive approach allows abundant WGA product
remaining for deeper targeted analysis. The technique may be integrated into the
clinical management of patients with metastatic prostate cancer to inform treatment
decisions, monitor recurrence, and provide valuable prognostic information about a
patient’s disease.
ST87. Spectrum of Genetic Mutations in Colorectal Adenocarcinoma Among
Hispanics and Native Americans in New Mexico
C.J. Broehm, D. Chabot-Richards
University of New Mexico/Tricore Reference Labs, Albuquerque, NM.
Introduction: The pattern of prevalence and types of genetic mutations in colorectal
adenocarcinoma (CRC) among Hispanics and Native Americans is not well
understood, but is critical for optimizing care for these historically underserved
populations. We studied the prevalence of genetic mutations and microsatellite
instability (MSI) in CRC in White (W), Hispanic (H), and Native American (NA)
patients treated at the University of New Mexico Hospital. Methods: All cases of
primary and metastatic CRC diagnosed from January 2006 to January 2017 were
identified in our archives, as well as type and results of any genetic testing
performed on representative formalin-fixed paraffin-embedded (FFPE) tumor tissue.
Genetic tests performed included MSI by capillary electrophoresis and fragment
length analysis with reflex to MLH1 hypermethylation, KRAS (codons 12 and 13)
reflex BRAF V600E by pyrosequencing, a colorectal panel by next-generation
sequencing (NGS) consisting of KRAS and NRAS (exons 2 to 4), BRAF(exons 11
and 15), and PIK3CA (exons 6, 7, 9, 13, 19, 20, exon 6-7 intronic region) and the
NGS 50 Gene Cancer Hotspot Panel v2 (Life Technologies). Demographic data
including age, sex, and self-identified ancestry (White, Hispanic, Native American)
were recorded from the medical record. Results: Two hundred seventy-two patients
with CRC were identified (119 W [57M/62F, Age 30 to 86y], 116 H [59M/57F, Age 22
to 90y], 37 NA [22M/15F, Age 38 to 82y]). One hundred fifty-three cases had at least
one genetic test performed (71 W [35M/36F], 66 H [34M/32F], 16 NA [9M/7F]),
including 96 MSI (43 W, 41 H, 12 NA), 88 KRAS reflex BRAF (42 W, 39 H, 7 NA), 10
colorectal panel (8 W, 2 H), and two 50 gene panel (2 W). Overall, 60% of White,
57% of Hispanic, and 43% of Native American CRC had at least one test performed.
Microsatellite unstable cases included 9 Whites (21%, 5 hypermethylated), 2
Hispanics (5%, 1 hypermethylated), and 1 Native American (8%, hypermethylation
not performed). KRAS mutations were identified in 20 Whites (38%, codon 12=16,
codon 13=4), 20 Hispanics (49%, codon 12=16, codon 13=3, p.A146P=1), and 3
Native Americans (43%, codon 12=2, codon 13=1). Two
pathogenic PIK3CA mutations were identified (1 W, 1H). Pathogenic
BRAF (n=3), NRAS (n=1), and TP53 (n=1) mutations were identified in Whites only.
Significant sex and age differences were not noted. Conclusions: CRC among
Native Americans and Hispanics are often KRAS mutated. BRAF mutations and
microsatellite instability are rare in CRC among Hispanics and Native Americans.
Definitive conclusions are limited by the small number of Native Americans in the
sample and overall less frequent genetic testing, particularly KRAS testing, of tumors
from Native American individuals.
ST88. Major Factors Affecting NGS Failure in a Tertiary Care Hospital: The
Emory Experience
V. Avadhani1, O. Rouhi1, R.L. Geller1, Z.J. Dureau1, T. Schneider2, G.H. Smith1, C.
Hill1, M.R. Rossi1
1Emory University Hospital, Atlanta, GA; 2EGL Genetics, Tucker, GA.
Introduction: Targeted next-generation sequencing (NGS) of solid tumors has
become an integral part of clinical care of cancer patients. The Illumina TruSight
Tumor 26 (TST-26) assay is a 26 gene amplicon-based NGS panel used at our
institution as a laboratory developed test for melanoma, lung and colon
adenocarcinomas. The inability to fully process FFPE tumor specimens for this and
other NGS assays can impact patient care and increase cost. We evaluated a recent
cohort of 800 specimens to determine our rate of specimen QNS and its potential
causes. Methods: We collected synoptic information on 800 solid tumor tested from
2016-2017 using TST-26 using our CoPath clinical reporting system. This
information included: final results, specimen type (cytology versus surgical),
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
procedure (FNA, EBUS, excision, resection), fixative (cytolyt versus formalin), the
block age and the institution (the Emory health system versus outside). We matched
these data elements with nucleic acid, library preparation and sequencing quality
metrics. Results: The most frequent cause of specimen failure was at the level of
DNA QC, specifically high delta CT values representing poor DNA quality and
yield. The overall success rate for all 800 specimens was 73% (585/800), with 75%
(463/614) for surgical specimens and 65% (122/186) for cytology specimens. We
observed a significant difference in the success rate of the in-house cases compared
to the outside cases (p<0.001). The success rate of the in-house surgical specimen
was 85% (330/388), in contrast to 58% (133/226) for outside cases. The success
rate for the melanoma was the lowest among the solid tumors [In-house: Lung (84%239/286), Melanoma (68%-82/121), colon cancer (86%-43/50), Outside: Lung (56%47/84), Melanoma (55%-44/81), Colon (74%-30/41). Conclusion: Melanoma
specimens are generally the most challenging specimens for the TST-26 assay. To
our surprise, we observed a significantly lower success rate with specimens obtained
from outside the Emory University system. This may be related to multiple factors
that are technical (e.g. specimen size, different fixation protocols, receiving
unstained slides versus block, etc.) and/or pathologist specific (e.g. poor block
selection criteria). Further evaluation of factors impacting NGS failure rates from
specimens obtained outside of our institution are required to optimize laboratory
utilization and improve patient care.
ST89. Performance Evaluation of the Ion Torrent S5 XL for Targeted NextGeneration Sequencing (NGS) for Clinical Oncology
F. Sabato, C. Yang, V.S. Williamson, C.I. Dumur, A. Popa, P. Anderson, P.
Morris, E. Tully, A. Ferreira-Gonzalez
Virginia Commonwealth University Health, Richmond, VA.
Introduction: Detection of actionable somatic variants in solid tumors and
hematological malignancies by targeted NGS is becoming standard of practice in the
management of oncology patients. Our laboratory uses the Ion AmpliSeq Cancer
Hotspot Panel v2.0 (CHP2) on the PGM for clinical purposes. We evaluated the
newly released Ion S5 XL system which offers advantages compared to the PGM
system due to a fluidics mechanism and single-use reagent cartridges that avoids
the need for external gas and water supplies. It also provides higher capacity chips,
and faster run time and data analysis which increases throughput and efficiency.
Here, we present the validation of the S5XL system for routine clinical utilization with
the CHP2 in oncology samples. Methods: Amplicon libraries prepared with CHP2
and previously tested in our laboratory on the PGM were run with Ion 520 OT2 or
Chef kits and sequenced on S5XL. Optimization of library input concentration
and Torrent Variant Caller v5.2 parameter settings were performed on a diluted
SeraCare reference sample containing 26 variants (22 SNVs, 4 indels) at ~5% (AF5)
allele frequencies (VAF). Accuracy was assessed with SeraSeq reference materials
at 20% (AF20) and 10% (AF10) VAF. Seven replicates of AF5 were tested for lower
limit of detection (LOD) and effect of sample barcoding. A positive control was tested
in 4 runs for reproducibility. Accuracy to detect insertions was performed with
mixtures of cell lines containing a 30 bp insertion in FLT3 ITD, 4 bp insertion
in NPM1 and 36 bp insertion in NOTCH1 at 5% VAF. Agreement and correlation
were evaluated in 40 positive and 8 negative clinical samples. Results: Library
loading concentration of 50 pM for Ion Chef and 5 pM for OT2 generated up to 6
million sequencing reads. The SeraSeq 26 expected variants were accurately
detected with quality scores >600 in AF20 and >400 in AF10. LOD was 5% with
variants quality score of ≥100 when up to 16 libraries were multiplexed, and average
coverage >1000X when TVC v5.2 default somatic-low stringency settings were
applied. High reproducibility was observed across variants in the positive
control (CV≤7%). The S5XL detected different size insertions
in FLT3, NPM1 and NOTCH1 with accuracy. Both systems agreed at calling SNVs
and indels in all clinical samples. However, the S5XL called 3 variants that were no
call by the PGM. Reanalyses of the PGM data confirmed the presence of those
variants with low quality score and low allele frequencies. All patients’ variants were
called with highly correlated frequencies (r > 0.99) between the systems.
Conclusions: S5XL is an accurate, reproducible and sensitive system that can
replace the PGM system for sequencing of cancer hot-spot variants providing higher
throughput compared to PGM.
ST90. Analysis of Immune Response Gene Expression and Tumor Associated
Macrophages in Triple Negative Breast Carcinoma
K. Walker, D. Jebakumar
Scott and White Medical Center- Temple, Temple, TX.
Introduction: Triple negative breast cancers (TNBC) are aggressive heterogeneous
tumors which need to be refined to identify therapeutic subsets. Very limited data is
available on predictive immune response markers in different ethnic groups. Our
study aims to investigate the expression of immune cell infiltration markers and their
prognostic significance in TNBCs among Caucasian (CA) and African American (AA)
ethnic groups. Methods: Baylor, Scott & White database was searched, 42 AA and
42 CA TNBC cases from 2001 to 2012 were identified. Total mRNA was extracted
from paraffin embedded tissue blocks (Qiagen, MD) and mRNA expressions of 770
breast cancer-related genes were quantitated by high throughput multiplexed
methodology and analyzed using nSolver Analysis Software (V2.5). Results: The
mean age at diagnosis in the AA group was 51.4 years and the CA group 58.4 years
(p-value: 0.02). The tumor stage at presentation in the AA group showed stage I – 18
1029
AMP Abstracts
(43%), stage II - 18 (43%), stage III - 4 (10%) and stage IV – 2 (5%). The tumor
stage at presentation in the CA group showed stage I - 21 (50%), stage II – 16
(38%), stage III – 4 (10%) and stage 1V - 1 (2%). Gene expression profile shows
upregulation of STAT1, CXCL9, MX1, VEGFA and IFITM1 in both groups. CXCL10,
FOXP3, MX1, IFIT1, IFITM1, LAG3, CTLA-4, CD40, CD 40LG, CD163, IDO1,
VCAM1, VEGFC and IFN-γ showed statistically significant difference in expression
with p-values <0.1 as well as correlated with presence of metastasis (p value: 0.04)
and recurrence (p value: 0.003) in AA subset. CD 68 IHC showed presence of
increased Tumor Associated macrophages (TAMs) as a component of tumor
infiltrating lymphocytes (p value: 0.0017) and in the stroma (p value: 0.0005) of AA
subset. Conclusions: Our study identified younger age at presentation in the AA
subset with a statistically significant p- value 0.02. The tumor stage is matched
between both ethnic groups with statistically significant differences in outcomes in
relation to recurrence and metastasis in the AA subset. There is significant
differential expression of immune markers as well as increased presence of TAMs in
the AA subset which correlates with significantly unfavorable outcome. Of interest is
the inverse correlation of IFN-γ and VEGFC expression with outcome in the AA
subset of TNBCs, possibly linked to increased angiogenesis and higher rates of
metastasis in this sub-group.
ST91. Specimen Identification and Tracking from DNA Extraction to NGS
Results Through the Addition of Barcoded Synthetic DNA
R. Bastien1, S. Dames2, A. Bolia2, K. Simmon2, P. Rindler2, J. Raney2, J. Nelson2, A.
Hall2, G. Deftereos2, A. Matynia2, K. Gligorich2, M. Bronner2, H. Best2, B.
Kennedy2, E. Gee2, L. Furtado2
1Associated Regional and University Pathologists, Incorporated, Salt Lake City,
UT; 2ARUP Laboratories, Salt Lake City, UT.
Introduction: With the dramatic increase in the use of next generation sequencing
(NGS) as a diagnostic platform, confidence in accurate sample tracking and
reporting is increasingly important given the high sample multiplexing capacity in
NGS testing. Laboratories combine many unique patient samples and panels on
single robotic NGS library preparations and sequencing runs. This high level of
sample multiplexing increases the chance of sample switching and inaccurate
reporting with the potential to impact patient management and care. The use of
synthetic DNA oligonucleotides to track sample chain of custody can improve patient
care by reducing errors in improper result reporting. Methods: Synthetic
oligonucleotides were created from a 170bp platypus (Ornithorhynchus anatinus)
genomic sequence with a 10bp unique specimen index sequence inserted into the
middle. The sequences of the molecular indices were designed to have a minimum
Hamming distance of 4, no homopolymers greater than 2bp, and a GC content
between 40 to 60%. These oligonucleotides were either spiked in at the DNA
extraction step or into previously extracted DNA samples. A probe bridging the 10bp
molecular barcode designed to capture the platypus oligonucleotide was also added
to an existing NGS probe pool and the captured samples were sequenced.
Sequencing was performed using an Illumina NextSeq 500, 2x150bp paired-end
reads. A custom Kraken classification database was built from the 170bp platypus
sequence. All sequencing reads were analyzed with Kraken to identify any read
containing platypus sequence. The unique specimen index was found within these
reads using the location of known platypus 21-mers. Results: The synthetic
oligonucleotides added to both the DNA extraction step and extracted DNA were
successfully captured and sequenced. Of the platypus-positive reads, the intended
10bp unique specimen index barcode sequence constituted approximately 90% of
the recovered specimen index sequence. About 2% of the alternate specimen
indices, comprising the next largest population of identified specimen indices, have a
Hamming distance of one from the intended specimen index and are likely due to
sequencing error. Conclusions: Barcoded synthetic DNA oligonucleotides provide a
methodology to track samples through a NGS workflow by using benchtop methods
and tools in the downstream bioinformatics pipeline to confirm specimen
provenance. This technology can help reduce the potential of sample switching and
errors in patient reporting by tracing a sample through the end-to-end workflow
spanning extraction to reporting. Efforts are underway to continue development of
these specimen indices to improve sample tracking performance.
ST92. Plasma Mutation Spectrum Matches Known Tumor Mutations in Active
Cancer Patients
N.D. Montgomery, A.M. Tanner, J.P. Zevallos, A.L. Mazul, N.L. Ferguson, J.T.
Auman, S. Elmore, M.L. Gulley
University of North Carolina at Chapel Hill, Chapel Hill, NC.
Introduction: Modern sequencing technologies facilitate measuring cancer-related
mutations in plasma DNA to 1) track tumor burden, 2) identify emerging subclones
signifying aggressiveness or drug resistance, & 3) distinguish germline from somatic
alterations. We evaluated a commercial test system that preferentially amplifies
fragmented DNA to reduce impact of ex vivo release of cellular DNA,
that adds “unique molecular identifiers” to natural DNA fragments, and that
generates variable amplicon lengths. These innovations permit judging diversity of
input DNA as a quality indicator, and promote precise quantification of natural DNA
variants even at low allele fractions. Methods: Total nucleic acid was extracted from
51 plasma specimens from 47 cancer patients having known mutations based on
previous massive parallel sequencing of tissue representing melanoma (n=7),
oropharyngeal squamous (n= 7) or adenocarcinoma (lung, colon, bladder, pancreas;
1030
n=37). Nucleic acid extracted from 1-3mL EDTA plasma (Promega Maxwell RSC)
was sequenced across hotspots in 28 human cancer genes including all exons
of TP53 using commercial library preparation reagents and software (Archer Reveal
ctDNA 28 library reagents; Illumina NextSeq with Mid v2 sequencing reagents;
Archer Analysis software v5.0). Noise reduction algorithms and population
frequency data supported interpretation of variants confirmed in both strands and
cataloged along with variant allele fractions. Data were analyzed in concert with
clinical status at the time of plasma collection. Results: Control plasma (Horizon,
50ng DNA) revealed 8/8 variants at ~1%, and 6/8 at ~0.1% allele fraction. Four
plasmas sequenced in replicate yielded equivalent results. At least one known
tumor-related mutation was detectable in 40/51 (78%) active cancer patient plasmas
at variant allele fractions from <0.1% to 74% (average 20%). Variants in plasma that
were not seen in matched cancer tissue were generally <0.3% allele fraction. In
some patients, only a proportion of known cancer variants were detectable in
plasma, reinforcing the value of testing multiple tumor markers. Two treated lung
cancer patients had EGFR T790M implying outgrowth of a drug resistant clone.
Conclusions: Non-invasive detection of cancer-related mutations is feasible using
off-the-shelf reagents and software that support inter-laboratory harmonization. Error
correction and noise reduction algorithms resulted in high analytic sensitivity and
specificity. Multiple cancer types were amenable to analysis, especially late stage
melanoma and adenocarcinoma for which there was excellent correlation with active
cancer status. The findings demonstrate feasibility to track tumor burden and to
characterize emerging subclones in a wide range of cancer patients.
ST93. Comparison of Highly-Multiplexed Genomic Analysis System to
Targeted Re-Sequencing for Detection of Clinically-Relevant Mutations in
Colorectal Cancer.
T. Raz, F. Zhuang, V. Chellappa, P. Mary, M. Griesbach, D. Goldfarb, V. Dunn, C.
Ortega, H. Nguyen3, N. Brown, S. Powers, L. Hu, N. Walsh, L. Jensen-Long
Bio-Rad Digital Biology Center, Cambridge, MA.
Introduction: Precision medicine is being driven in large part by molecular
technologies that have proven to be effective in detecting clinically-relevant
mutations by a variety of methods including PCR and NGS. However, to date, no
technology has proven optimal in meeting all laboratory requirements: precise
detection and analysis of key driver mutations, fast turn-around time, simple
workflow, low sample input, and cost effectiveness. This study evaluates the
capabilities of a new, highly-multiplexed, sample-in-report-out molecular diagnostic
platform - MonoRAyL g1 Genomic System - in comparison to targeted re-sequencing
to evaluate these requirements in the analysis of variants in colorectal cancer tissue
samples. Methods: One sample consisting of a combined set of 6 Horizon
Discovery Reference Standard samples and 8 samples extracted from FFPE
colorectal tumor tissue samples harboring KRAS, NRAS, BRAF and EGFR
mutations were analyzed. 25-50 ng of input genomic DNA was used for the
assessment of mutational status of 66 actionable genetic variants using the
MonoRAyL g1 Hot Spot Assay and instrument. Concordance was determined
against the Illumina TruSight Tumor 15 sequencing panel on the MiSeq platform.
Additionally, platform robustness was demonstrated by studies composed of
Repeatability and Reproducibility and Limit of Detection (LOD) via dilution series and
minor allele frequency (MAF) distribution. Results: Preliminary results from 8 FFPE
samples showed >99% analytical assay specificity and >97% analytical assay
sensitivity. Reproducibility was shown to be 100% in detecting 6 mutations in a
single sample over multiple (n=5) runs, 2 instruments and 3 lab operators. This
multiplexed genotyping system demonstrated >95% concordance for mutation calls
versus NGS targeted re-sequencing. All variants at >8% MAF were detected.
Conclusions: Preliminary data from Bio-Rad’s MonoRAyL g1 Genomics System
and assay provided precise analysis of the important driver mutations in colorectal
cancer samples analyzed in this study. The assay workflow involved < 20 min handson cartridge preparation, followed by ~4h of on-instrument processing including
automated variant analysis, in a true walk-away system. Additional studies, both
within US and outside are planned to show concordance to this original work with
FFPE tissue and other sample types.
ST94. An Integrated Genomic and Proteomic Analysis of Human Tumors
Enables Epitope Prediction for Cancer Immunotherapy.
M. Davis1, B. Bulik-Sullivan1, J. Busby1, A. Clark1, T. Murphy1, M. Busby1, F.
Duke1, C. Palmer1, C. Couter1, A. Yang1, K. Jooss2, M. Skoberne1, A. Derti1, J.
Francis1, R. Yelensky1
1Gritstone Oncology, Cambridge, MA; 2Gritstone Oncology, Emeryville, CA.
Introduction: Immunotherapy has become essential in the treatment of cancer,
although clinical benefits have been limited to a minority of patients. The discovery of
tumor-specific neoantigens (TSNA) as principal targets of T cells under immune
checkpoint blockade opens the possibility to improve responses with personalized
immunotherapies such as TSNA vaccines. Methods: We examined key questions in
clinical neoantigen identification for TSNA immunotherapy by performing deep
tumor/normal exome sequencing (median 571x/119x), tumor transcriptome
sequencing (median 111M reads), and mass-spec HLA class I peptide sequencing
(median 3,557 peptides) on fresh-frozen tumor specimens from 22 NSCLC
patients. Results: We assessed the impact of sequencing depth and transcriptome
measurement on neoantigen detection, quantified capabilities of current algorithms
to identify antigens, and developed a novel deep learning model for neoantigen
jmd.amjpathol.org ■ The Journal of Molecular Diagnostics
AMP Abstracts
prediction. A total of 6,233 nonsynonymous somatic mutations were identified and
62% of mutations were transcribed in mRNA. Down-sampling tumor DNA
sequencing data to a typical median unique coverage of ~150x and RNA to ~50M
reads revealed a 20% loss of candidate neoantigens, demonstrating the importance
of sequencing depth. Gene expression strongly predicted HLA presentation, with
peptides from genes detected in RNA found presented >20X more often than
peptides from non-detected genes. With this and public datasets we trained a model
to predict HLA presentation and compared it to standard HLA binding affinity
prediction at the strong-binder threshold for well-characterized alleles. Whereas
binding affinity gave a positive predictive value (PPV) of only 3% for HLA presented
antigens, our deep learning approach gave a PPV of 30% at equivalent recall, a tenfold gain. HLA peptide sequencing proved scalable, with >50 additional lung tissues
profiled successfully. Initial results indicated that HLA presented TSNA can prime T
cells in vitro. Conclusions: This study evaluated requirements for neoantigen
identification, characterized key correlates of HLA presentation, and enables
neoantigen identification for personalized cancer immunotherapy.
coverage filter was employed for filtering variants. A limit-of-detection study was performed
using serially diluted SW480 cell line DNA positive for TP53 p.R273H and TP53p.P309S
mutations. Assay performance characteristics such as analytical sensitivity, specificity and
reproducibility were established for clinical implementation. Results: We achieved an
average cluster density of 1,100K/mm2 with coverage of 6,000X per sample using Miseq for
this TP53 NGS assay. A total of 31 expected variants (24 SNVs and 7 indels) were detected.
Three additional SNVs were detected by TP53 NGS including 2 SNVs that were in genomic
regions not covered by the Ampliseq panel. There was a good correlation between the mutant
allelic frequencies of TP53 variants detected between the 2 panels (R2: 0.91; p<0.0001). The
NGSTP53 panel required less hands-on time (~3 hrs) compared with the Ampliseq panel (~6
hrs). Sensitivity studies showed a 5% limit of detection with high inter- and intra-run
reproducibility. Conclusions: This TP53 single gene NGS assay, with all exons covered, is a
highly sensitive, less labor intensive, and cost-effective method for TP53 mutation detection
as compared with conventional sequencing technologies and multiplex targeted NGS hotspot
panels.
ST95. Development of Real-Time PCR Assay for Relative Expression of Total
EGFR mRNA and Detection of EGFRvIII mRNA in Glioblastoma Multiforme
Tumors
R. Kular, H. Alexander, J. Wuitschick, K. Nelson, E. Pabich, J. Rhoads, S. Huang, C.
Mullen
Abbott Molecular, Des Plaines, IL.
Introduction: Glioblastoma Multiforme (GBM), the highest grade glioma (grade 4), is
the most common and most aggressive malignant primary brain tumor in adults.
GBM tumors frequently have increased epidermal growth factor receptor (EGFR)
gene amplification and over-expression of EGFR mRNA. In addition, deletion of
exons 2 to 7 within the EGFR gene results in a deletion mutant described as
“EGFRdel2-7” or “EGFRvIII”, which occurs mostly in EGFR amplified tumors.
EGFRvIII expression is associated with a worse prognosis in GBM patients. Abbott
Molecular designed a real time multiplex PCR assay to detect EGFRvIII mRNA and
determine the relative expression of total EGFR (tEGFR) mRNA in GBM FFPE
specimens. The assay is currently being used in clinical trials of an investigational
therapeutic agent for GBM patients. Methods: Precision for the EGFR assay was
evaluated by testing a 9-member panel using 3 lots of EGFR amplification reagents,
3 m2000rt instruments and 3 operators. 5 assay runs were performed for each
reagent lot/instrument/operator combination to assess precision. Analytical linearity
was evaluated by testing 5 independent dilution panels prepared with RNA eluate
extracted from GBM FFPE tumor tissue. Specificity of the RealTime EGFR assay for
the EGFRvIII target was assessed by evaluating 20 FFPE brain tissue specimens
from normal donors. In addition, the distribution of EGFRvIII, tEGFR and Actin Ct
values from 179 GBM FFPE tumor specimens was evaluated. Results: The results
of the precision study showed a total %CV of less than 3% for EGFRvIII, tEGFR and
Actin CN values. For all 5 panels evaluated, the EGFRvIII CN (for EGFRvIII positive
specimens), tEGFR CN, ACT CN and tEGFR ∆CN values were linear between the
least diluted panel member (PM 01) and the most diluted panel member (PM 15).
The linear range of the specimen dilution panels spanned 6.6000 µL to 0.0004 µL of
RNA eluate per reaction. The 20 brain tissue specimens from normal donors did not
generate a positive EGFRvIII assay result; thus an EGFRvIII negativity rate of
100.0% was attained. Real time PCR results for EGFR mRNA expression were
correlated with gene amplification using FISH (Fluorescence in-situ Hybridization)
technology and EGFRvIII detection was correlated with Next generation sequencing
results. Conclusion: The new Abbott RealTime EGFR assay is a sensitive, specific
and reliable assay which can determine relative quantification of tEGFR mRNA and
detect deletion mutant EGFRvIII mRNA in patients with GBM tumors.
ST97. Molecular Profiling of Gallbladder Cancer Tumors of New Mexico
Populations
R. Gullapalli1, E. LaBauve2, J. Pomo2, D. Martin2
1University of New Mexico Health Sciences Center, Albuquerque, NM; 2University of
New Mexico, Albuquerque, NM.
Introduction: Gallbladder Cancer (GBC) is a relatively rare cancer. It is the sixth
most common malignancy of the gastrointestinal tract. Despite its relative rarity, it
has a distinctly poor prognosis with less than 8% 5-year survival in patients of Stage
III and above. Epidemiologically, GBC has a distinct incidence pattern across the
world with high incidences in Chile, India, Poland and the state of New Mexico within
USA. The Native American Population of New Mexico has been shown to have a 6-8
times higher incidence of GBC compared to Caucasians for currently unknown
reasons. We examined the molecular profile of a cohort of GBC patients here in New
Mexico. Methods: We collected GBC FFPE tissues after obtaining prior IRB
approval and the final analysis included 24 patient samples which were eligible for
the study. The cohort comprised of 9 Caucasian, 9 Hispanic and 6 Native American
Ethnicities. We used the Ion Proton platform to run the 409 gene comprehensive
cancer panel on the DNA samples of these patients. The downstream bioinformatics
analysis was performed using the Torrent Variant Caller (v.5.0.4) in the preliminary
step. This was followed by the upload of BAM files to the cloud to perform variant
annotation using the Ion Reporter software (v.5.0). In addition to next generation
sequencing, we also performed IHC and FISH based analysis to assess and confirm
various biomarkers of likely clinical significance in GBC signaling pathways.
Results: We successfully completed the sequencing and analysis of the 24 sample
GBC patient cohort of patients in New Mexico. We identified a spectrum of mutations
in the various samples which have been reported previously in the context of GBC.
These include the mutations of genes such as TP53, ERBB pathway genes and
KRAS to name a few. Two out of 24 patients showed a high mutational burden which
is in line with the current estimates of the prevalence of MSI in GBC. We also
identified ERBB2 (HER2) amplification in ~25% of the patients sequenced here in
New Mexico (n=6/24). We confirmed these findings using HER2 by
immunohistochemistry as well as FISH based testing. HER2 amplification has been
reported previously in GBC and we confirm those findings here in GBC patients of
New Mexico. Conclusions: We have performed a high-throughput next generation
sequencing analysis of paired tumor-normal samples of gallbladder cancer patients
here in New Mexico. We have identified HER2 gene amplification in roughly ~25% of
the patients here in New Mexico. In combination with the ability to detect high
tumoral burden of mutations, we are currently exploring options to enable a
personalized, investigator-initiated trial (IIT) for gallbladder cancer patients here in
New Mexico.
ST96. Validation of a Single-Gene Next-Generation Sequencing Assay for TP53
Mutation Detection in Solid Tumor FFPE Samples in CLIA Laboratory Using
Illumina MiSeq
B.A. Barkoh, B.A. Barkoh, M. Mehrotra, R.R. Singh, L. Yubitza, D. Vidal, S. Zalles, B.
Hannigan, A. Bolivar, D. Dzifa, W. Chen, M.A. Harmon, K.P. Patel, M. Routbort, R.
Kanagal Shamanna, I. Wistuba, J. Medeiros, R. Luthra
University of Texas MD Anderson Cancer Center, Houston, TX.
Introduction: The tumor suppressor gene TP53 is mutated frequently in many cancers
andmutations span the entire coding regions of the gene. TP53 single gene testing is
performed currently using Sanger sequencing. Cost, labor, turnaround time and limited
detection sensitivity (~20%) are some of the limitations of using conventional Sanger
sequencing, whereas most NGS panels to assess TP53 have a higher sensitivity, but cover
only hotspots. Here we tested a comprehensive single gene NGS assay for detection
of TP53 mutations and compared its performance, analytical sensitivity and specificity to the
targeted Ion Torrent Ampliseq Cancer Hotspot NGS panel. Methods: Thirty-seven FFPE
solid tumor samples with known TP53 mutation status (24 SNV, 7 indel and 8 wild type) as
detected previously by the targeted NGS Ampliseq Cancer Hotspot Panel v2 that includes 50
cancer-related genes formed the study group. The TP53 NGS assay consisted of 17
amplicons in 2 pools covering the entire TP53 coding regions (size 1.8 kb). 20ng genomic
DNA was used for library amplification according to the manufacturer’s instructions
(Multiplicom, Belgium). Quantified barcoded libraries were pooled and sequenced (24
samples/run) on a Miseq using a 300v2 nano flow cell. Miseq reporter was used for variant
calling at 5% allelic frequency (AF), sequencing depths of 250X coverage and a 25X variant
ST98. IDH1 and IDH2 Mutations in Gliomas, AML, and Intrahepatic
Cholangiocarcinoma
M.B. Wachsmann, K.J. Hatanpaa, W. Chen, S. Hammer, R. Ram, D. Oliver
The University of Texas Southwestern Medical Center at Dallas, Dallas, TX.
Introduction: Mutations of isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) are
common in gliomas, acute myeloid leukemia (AML) and intrahepatic
cholangiocarcinomas (IHCC). Mutant IDH1/2 promotes carcinogenesis by
generating the oncometabolite 2-hydroxyglutarate, which inhibits
normal TET2 function with loss of cytosine hydroxymethylation. Previous studies
have found IDH mutations in 85% of grade II-III glioma, 14% to 35% of AML, and
10% to 30% of IHCC cases. To better understand the carcinogenic role of IDH in this
variety of tumor types, we sought to characterize the spectrum of IDH1 and 2
mutations found in clinical specimens from these 3 sites. Methods: A total of 645
FFPE, blood, and bone marrow cases were included in the study and analyzed by
one of 3 methods that all queried IDH1 R132, IDH2 R140, and IDH2 R172. DNA
was extracted using either Qiagen QIAamp DNA Blood or FFPE Tissue kits. Sanger
sequencing used BigDye Terminator v3.1 chemistry with lab developed primers.
Sequenom MassARRAY mass spec analysis was done using software-designed
primers and iPLEX chemistry. Next generation sequencing testing was performed
with the 50-gene Ion AmpliSeq Cancer Hotspot Panel v2 on an Ion Torrent PGM
using a 316 chip. Results: In our series, IDH was mutated in 32% of glial
neoplasms, 25% of AML, and 14% of IHCC at initial diagnosis. The glioma positive
rate was low due a high number of glioblastoma cases and in some instances
The Journal of Molecular Diagnostics ■ jmd.amjpathol.org
1031
AMP Abstracts
withholding cases positive for R132H by immunohistochemistry. The most common
mutations in gliomas were IDH1 c.395G>A (42%) and IDH1 c.394C>T (20%). In AML
the most common were IDH2 c.49G>A (48%) and IDH1 c.394C>T (27%). In IHCC
cases the most common mutation was IDH1 c.394C>T (86%). The positivity rate was
consistent across all 3 analysis methods, indicating all are adequate for clinical
testing. Conclusions: In our patient population, gliomas harbored IDH1 more
frequently than IDH2 mutations, while in AML the frequency of IDH1 versus IDH2
mutations is more balanced. Strikingly, while the IDH1 c.394C>T makes up a small
percent of the mutations in glioma and AML, it makes up the vast majority (nearly
90%) of mutations in IHCC cases. This may be a result of carcinogen exposure,
potential tumorigenicity of a defined mutation, or other causes. Specific mutations
may predict response to targeted therapies in clinical trials.
ST99. Integration of HER2 Overexpression/Amplification with Molecular
Mutation Profile in Urothelial Carcinoma
J. Zhao1, X. Lu2, M. Zhang3, C. Guo3, L. Huo3, A. Yemelyanova3, N. Navai3, R.
Broaddus3, H. Chen3
1Hunan Cancer Hospital, Changsha Hunan, China; 2Northwestern University,
Chicago, IL; 3MD Anderson Cancer Center, Houston, TX.
Introduction: HER2 overexpression and amplification have been reported in
urothelial carcinoma (UC) of bladder with micropapillary features. Mutations
in TP53 and FGFR3 have been described in bladder UC with distinct morphologies.
However HER2 status and mutations involving other genes in UC originating from
other genitourinary sites are not well characterized. In this report, we
described clinicopathological and molecular features of UC originating from bladder
and other genitourinary sites. Methods: We retrospectively reviewed cases of highgrade UC that were found to have HER2 amplification assessed by fluorescence in
situ hybridization (FISH) and/or HER2 overexpression by immunohistochemistry
(IHC) at MD Anderson Cancer Center from 2014-2016 in accordance with
ASCO/CAP 2013 recommendations. We reviewed clinically reported molecular
profiles generated by next generation sequencing using the Ion Torrent PGM 50gene hotspot panel (N=63) and Oncomine 134-gene panel (N=4). Results: This
study included 77 patients with median age of 66 years that were predominantly
males (80.5%). Primary tumor sites were bladder (66), renal pelvis (6), ureter (4) and
urethra (1). Histologic subtypes included 32 papillary UCs (of which 7 cases with
micropapillary features) and 45 non-papillary UCs (of which 15 cases with
micropapillary features). HER2 overexpression was observed in 8/16 (50%)
of HER2-amplified tumors, but was not in any of the tumors
without HER2 amplification (0/55; p<0.01). HER2 amplification was more frequent in
cases with micropapillary features (10/22, 45.5%) than those without micropapillary
features (8/55, 14.5%; p<0.01). Gene mutations were identified in 54 of 67 (80.6%)
of the tested cases. The most frequently mutated genes were TP53 (36/67, 53.7%)
followed by PIK3CA (6/67, 9.0%), FGFR3 (6/67, 9.0%) and RB1 (5/67, 7.5%).
Patients with non-papillary UC had more prevalent TP53 mutation than papillary UC
(26/40, 65.0% vs 10/27, 37.0%, p=0.04), and less common PIK3CA (1/40, 2.5%
versus 5/27, 18.5%, p=0.02) and FGFR3 (0, 0.0% versus 6/27, 22.2%, p<0.01)
mutations, respectively. UC originating from bladder more often
harbored TP53 mutation than UC from other genitourinary sites (35/57, 61.4%
versus 1/10, 10.0%, p<0.01), and less commonly FGFR3 mutation (2/57, 3.5%
versus 4/10, 40.0%, p<0.01). No association between TP53 mutation
and HER2 amplification was observed. Conclusions: HER2 amplification by FISH
was highly concordant with HER2 expression by IHC. The frequencies of genetic
alternations in HER2, TP53 and FGFR3 were dependent on histologic type and
primary tumor site. Accurate assessments of these genetic alterations are of clinical
importance in determining patient eligibility for targeted therapy.
ST100. Utilizing a Comprehensive Next-Generation Sequencing Panel to
Improve Clinical Outcomes in Patients with Non-Small Cell Lung Cancer
S. Springborn, K. Bartow, D. Abbott, A.C. Mackinnon
Medical College of Wisconsin, Milwaukee, WI.
Introduction: Lung cancer remains the leading cause of cancer-related deaths
worldwide. Non-small cell lung cancer (NSCLC) accounts for more than 85% of
cases. Recurrent NSCLC mutations include single nucleotide variants (SNV), indels,
copy number variants (CNV), and gene fusions. These biomarkers have both
prognostic and therapeutic value. Differences in treatment response rates and
progression-free survival have been documented in patients with unique mutational
profiles. Gene profiling therefore significantly improves patient care. Current mutation
analysis involves multiple diagnostic platforms often performed at independent
laboratories. Next generation sequencing (NGS) allows for massively paralleled
profiling across large numbers of genes and mutation types. A comprehensive NGS
assay has the potential to provide in-depth data while improving the delivery time of
genetic information to clinicians. To further investigate the added utility of an
extensive lung panel, we performed NGS on a cohort of NSCLC specimens using
ArcherDx’s Comprehensive Thyroid and Lung (CTL) Panel. Methods: The control
cohort consisted of 9 gDNA and FFPE standards with known variant allele
frequencies. FFPE DNA and RNA were isolated from 2 additional patient cohorts: a)
18 NSCLC specimens with known EGFR (n=4), ALK (n=12), and ROS (n=2)
mutations and b) 12 NSCLC specimens with no identifiable “driver” mutation by
targeted NGS and/or FISH analysis. Archer’s VariantPlex and FusionPlex CLT
Panels were used to prepare sequencing libraries. Sequencing was performed on
1032
Illumina’s MiniSeq Platform using High Output Sequencing Kits per manufacturer’s
guidelines. Results were evaluated with Archer Analysis 5 software. Results: The 9
control specimens demonstrated a 96.8% variant detection rate and a 100% fusion
detection rate with a false positive detection rate of 0%. The panel was unsuccessful
in detecting a known MET CNV in a single control specimen. The concordance rates
for the 18 known NSCLC specimens were 100% for variant detection and 85.7% for
fusion detection. In 3 of the 12 “driverless” NSCLC specimens, clinically significant
CNV (MDM2 amplification) and/or fusion (ALK) events were detected. Conclusions:
The observed concordance rates in both the control and known positive samples
demonstrates the utility of the CTL panel for both variant and fusion detection. The
CTL assay identified 2 ALK rearrangements that were previously FISH negative.
Sample quality likely affected the assay’s performance, and specimen requirements
should be evaluated for clinical testing. Additional investigation interrogating a larger
cohort is necessary to further examine the panel’s ability to reliably detect CNV.
ST101. MET Amplification Predicts Primary Resistance to EGFR-TKIs in
Advanced Non–Small Cell Lung Cancer Patients with Sensitive EGFR Mutation
L. Fang1, X. Lu2, S. Hu1, Z. Tang1, R. Luthra1, M. Routbort1, K. Patel1, R. Wang3, R.
Broaddus1, H. Chen1
1UT MD Anderson Cancer Center, Houston, TX; 2Northwestern University, Chicago,
IL; 3Jiangsu Province Hospital of TCM, Nanjing Jiangsu, China.
Introduction: MET amplification has been reported in approximately 20% of nonsmall cell lung cancer (NSCLC) patients with acquired resistance to EGFR-tyrosine
kinase inhibitors (TKIs) targeted therapy. Studies have indicated the presence of a
non-sensitive EGFR mutation contributing to subset of primary resistance to EGFRTKIs. In vitro study also suggested MET amplification might contribute to primary
resistance to EGFR-TKIs therapy; however the clinicopathological and molecular
features for NSCLC patients with primary resistance to EGFR-TKIs are not fully
underst