A commentary on the discrepancy between blood and tumorBRCA testing: an open question
Elisa De Paolis1, Claudia
Marchetti2;3, Paola Concolino1,
Giovanni Scambia2;3, Andrea
Urbani1,3, Anna Fagotti2;3, Angelo
Minucci1*
1Molecular and Genomic Diagnostics Unit, Fondazione
Policlinico Universitario A. Gemelli IRCCS, Rome, Italy;
2Division of Oncological Gynecology, Department of
Women’s and Children’s Health, Fondazione Policlinico Universitario A.
Gemelli IRCCS, Rome, Italy;
3Catholic University of the Sacred Heart, Rome, Italy.
Running title: Understanding discordant BRCA test cases.
*Corresponding author:
Angelo Minucci
Molecular and Genomic Diagnostics Unit, Fondazione Policlinico
Universitario A. Gemelli IRCCS, Rome, Italy.
angelo.minucci@policlinicogemelli.it
Molecular evaluation of BRCA1/2 (BRCA ) genes represents a
well-known example of precision oncology. The availability of Poly ADP
Ribose Polymerase inhibitors (PARPi) as target therapy option for
several BRCA mutated cancers types (e.g. ovarian, breast,
prostate, and pancreatic)1 changed the course ofBRCA testing over the last years. In this context, an emerging
path of molecular evaluation is represented by the BRCA testing
performed directly on tumor tissue (tBRCA ): this increased the
chance to identify more patients with higher likelihood of benefiting
from PARPi treatment. This approach leads to the simultaneous
identification of both constitutional and somatically acquired variants,
with a lower turnaround time: the identification of BRCApathogenic variants (PVs) could lead to a secondary “reflex” germlineBRCA (gBRCA) testing in order to assess personal and
familiar risks. In contrast, performing gBRCA as first molecular
test causes the loss of a relevant proportion of patients with tissue
acquired BRCA PVs, needing of a following tumor
test2;3.
However, challenges exist in tBRCA that may lead to inefficient
germline variant call. A recently published paper by Kordes et
al. reported a pancreatic adenocarcinoma patient with a germlinenovel BRCA2 c.516+4A>G variant
classified as deleterious by the authors based on in silico and
functional data4. Also the tumor tissue was sequenced
in order to achieve the enrolment criteria for a clinical trial of
Olaparib in combination with pembrolizumab (KEYLYNK-007). Unexpectedly,
the germline variant was not stated in the final report. The authors
took into account all the relevant basis of the experienced discrepancy,
without identifying a confident reason.
In our opinion it is crucial to investigate about the reliability of
tBRCA in the identification of both somatic and germline
variants. Inspired by the recently published commentary of
Gourley5 and taking into account that several
troubling cases of discrepancy between blood and tBRCA testing
have been reported in literature, we collected the recent relevant
studies covering the comparison between gBRCA and tBRCA to
give a critical opinion about some shared key points of the somatic
testing that could affect the final genotyping and reporting (Table 1).
Major reasons of discrepancies are related to: (1) differences in input
DNA quality, (2) type of BRCA gene alteration, (3) inherent
limitations of the Next Generation Sequencing (NGS), (4) bioinformatics
pipeline features (e.g. the ability to predict the occurrence of Copy
Number Alterations (CNAs) and the evaluation of the intron/exon
boundaries), and finally (5) the issues related to the BRCAvariants interpretation and classification.
To date, tBRCA testing is mainly performed on two sample types:
Fresh Frozen Tissue (FFT) and Formalin-Fixed Paraffin-Embedded (FFPE).
Here, we focused on tBRCA performed on FFPE being the most common
tissue type used for clinical diagnostic purpose. Pre-analytical
procedures regarding fixation step, tissue section size, and neoplastic
cell content assessment, are well-known crucial aspects of the
tBRCA testing reliability. In fact, sub-optimal DNA quality
represents a relevant reason of inaccuracy of tBRCA and also it
is the cause of around 5% of FFPE tBRCA NGS testing fails, with
the consequent need of additional new samples9. In
Bekos et al. only the retesting of a newly extracted tumor DNA
solved two cases of discrepancies with gBRCA : the BRCA1c.1881_1884del variant was not recognized due to poor NGS
quality data related to the input materials, as well as for theBRCA2 c.8537_8538del variant2. Also in Careet al. the test failure rate was related to fixation methods or
storage of FFPE material8. Ad hocrecommendations for the “ideal” starting tissue material are
available9;14.
Furthermore, different approaches should be used in the analytical step
for the BRCA genes amplification and sequencing, with several
types of sequencing chemistries (e.g. amplicon-based, capture-based),
platforms (e.g. Illumina, IonTorrent) and data analysis pipelines (e.g.
full-coding regions or hot spot analysis, different size
of splice site region analysed, CNAs detection). Each one of these could
be characterized by specific pitfalls that affect the downstream
bioinformatics variants filtering and calling. For example, in
amplicon-based approaches, a reason leading to the missing of a variant
detection may be related to the experimental design of the primers
distribution along the genomic region of interest. Variants located at
the 3’ or 5’ ends of overlapping amplicons could be covered by only one
read and could be consequently identified with a “strand bias” flag
and filtered out at the bioinformatics quality check3.
Also the use of different bioinformatics pipeline for the NGS data
analysis derived from the germline and the somatic tests of the same
patient could be the cause of apparent inconsistent results. For
example, in a large cohort of patients affected by several types of
malignancies and analysed for the evaluation of the utility of germline
test following tumor test, Lincoln et al. identified several
cases of discrepancies between the two tests (n=4)15.
Among these, the germline BRCA2 c.8967_8973del variant
was not detected in tumor sequencing due to the characteristic of
somatic panel (hot spot type), not comparable to the germline
one. Moreover, in case of discrepancy involving splice site variants
could be useful to check the concordance of the splice site region size
included in the germline and somatic bioinformatics
pipelines3. Regarding data analysis, it should be
acknowledged that some tumor testing platforms filter out germline
variants in the final reports in order to improve the accuracy of
somatic variant calling.
A well-known cause of gBRCA /tBRCA non-concordance resulted
from the challenge in the bioinformatics calling of CNAs in tissue
samples2;3;15. NGS sensitivity in CNAs detection
mostly depends on DNA quality, tumor heterogeneity, library preparation,
type of algorithms, and size of rearrangement. As a consequence, the
somatic bioinformatics pipeline must require computational algorithms
developed ad hoc and specific characteristics of sequencing raw
data (e.g. maximum amount, coverage uniformity and sufficient reads
depth)1. Even if the majority of methods are optimized
for somatic CNAs identification6;8, attention should
be given in the comparison of blood and tissue tests
results13. As an example, Bekos et al. failed
to identify in the tumor sample a verified pathogenic germline deletion
of BRCA1 exon 20. Only a careful re-evaluation of the
bioinformatics variant calls finally revealed the deletion and leaded to
the correction of the report2.
Relevant role in the evaluation of non-concordant results is played by
the post-analytical step involving the BRCA variants
interpretation. Complex issues underlying the classification ofBRCA variants exist. The American Collage of Medical Genetics
(ACMG) and the Association for Molecular Pathology (AMP) have
established the best practice for germline variant interpretation
providing the well-known classification using a five-tier
system16. Conversely, the interpretation of somatic
variants should be focused on their impact on clinical care.
Specifically, evidence-based categorization of somatic variants released
by the AMP, the American Society of Clinical Oncology (ASCO), and the
College of American Pathologists (CAP) includes a four-tier system: (1)
variants of strong clinical significance (level A and B of evidence);
(2) variants of potential clinical significance (level C and D of
evidence); (3) variants of unknown clinical significance; (4) benign of
likely benign variants17. To date, with the
publication of an increasing number of large-scale tumor sequencing
projects, a plenty of information is being collected into several public
databases useful for the querying about the significance of aBRCA variant. Cancer-specific variant databases are available as:
BRCAexchange, OncoKB, Catalog of Somatic Mutations in Cancer, My Cancer
Genome, cBioPortal, Memorial Sloan Kettering Cancer Center,
International Cancer Genome Consortium, and VARSOME. Likewise,
constitutional variant databases available are mainly: ClinVar, Human
Gene Mutation Database, ENIGMA, Leiden Open Variation Database, and
VARSOME. Differences in the germline- and somatic-based annotation may
exist between the abovementioned tools. Consequently, the risk of
non-concordant annotations of a BRCA variant could occur. This is
crucial in the comparison between the same molecular test performed by
different labs and it is exacerbated in the case of tBRCA and
gBRCA concordance evaluation: variants that met germline
guidelines16 to be considered pathogenic may not meet
the criteria17 to be considered oncogenic in the
somatic test. This situation could more likely affect the missense
Variants of Unknown Significance (VUSs)15. As reported
by Bekos et al. , after the inclusion of BRCA VUSs in the
secondary data analyses, the concordance rate of tumor testing compared
to germline one decreased, mainly due to VUSs
classification2. Moreover, in a large study
investigating the differences in variant interpretation between germline
and somatic variants accounted in several cancer-related genes, Moodyet al. highlighted a relevant percentage of discrepancies in
variants classification. Among these, the authors reported fourBRCA2 variants with discordant somatic/germline
annotations15.
In a retrospective cohort of 57 subjects tested for both germline and
somatic BRCA status, Kim et al. highlighted one case of a
germline variant not identified in the tissue
evaluation10. This discrepancy derived from a true
reversion of the germline BRCA1 variant accounted via restoration
of the wild-type allele in the tissue cells. Finally, tBRCAshould follow specific criteria that maximize molecular information,
improving the clinical relevance of the test and giving a more
comprehensive interpretation of each variant. With these purposes,
peculiar role is played by the “naturally occurring” BRCAsplicing isoforms. As we recently described for the BRCA1c.788G >T variant, complex considerations
should be done for rare variants that not only are different germline
and somatic annotations, but also are characterized by variability in
final effect and annotation in the context of all gene relevant
transcripts19.
In conclusion, we underline as the systematic and careful checking of
tumor tissue suitability could prevent and solve non-concordance cases.
Moreover, the robust identification of BRCA variants in FFPE
sample correlates with the confidence of the bioinformatics pipeline
adopted for the variant filtering and calling, especially for the CNAs
detection. In addition, translation of variant calls into clinical
decisions relies on proper annotations and discrepancies in
classifications of specific variants between tumor and germline contexts
could represent a relevant pitfall.
We argue that only harmonized guidelines encompassing the abovementioned
methodological and post-analytical steps could solve the BRCAgermline and somatic testing bias. In our laboratory, BRCAgenetic testing is routinely performed on blood, FFT and FFPE
samples1. In many cases, we routinely analyze matched
blood and tissue samples belonging from the same patient, in order to
perform an efficient BRCA test comprehensive of both germline and
somatic evaluation. This approach pointed out also the relevance of
multi-disciplinary and skilled resources for a solid molecular
characterization of the tumor. Together with the need of
standardization, we suggest as performing BRCA molecular test at
both germline and somatic levels in the same laboratory could improve
the reliability of the entire molecular path taken by the patient and
his clinicians.
Disclosure of interests: none declared.
Contribution to Authorship: E.D.P. and A.M. conceived of the
presented paper and wrote the manuscript with support from P.C. and
C.M.. G.S., A.U. and A.F. supervised the project. All authors discussed,
edited and contributed to the final manuscript.
Details of Ethics Approval: not applicable.
Funding: no funding was received for this commentary.