Aims Next Generation Sequencing (NGS)-based BRCA tumour tissue testing poses several challenges. As a first step of its implementation within a regional health service network, an in-house validation study was compared with published recommendations.
Methods Epithelial ovarian cancer (EOC) formalin-fixed paraffin-embedded specimens stored in the archives of the eight regional pathology units were selected from a consecutive series of patients with known BRCA germline status. Two expert pathologists evaluated tumour cell content for manual macrodissection. DNA extraction, library preparation and NGS analyses were performed blinded to the germinal status. Parameters used in the study were confronted with guidelines for the validation of NGS-based oncology panels and for BRCA tumour tissue testing.
Results NGS analyses were successful in 66 of 67 EOC specimens, with good quality metrics and high reproducibility among different runs. In all, 19 BRCA pathogenic variants were identified: 12 were germline and 7 were somatic. A 100% concordance with blood tests was detected for germline variants. A BRCA1 variant showed a controversial classification. In different areas of two early stage EOCs showing somatic variants, intratumour heterogeneity not relevant for test results (variant allele frequency >5%) was observed. Compared with expert recommendations, main limitations of the study were absence of controls with known somatic BRCA status and exclusion from the validation of BRCA copy number variations (CNV).
Conclusions A close collaboration between pathology and genetics units provides advantages in the implementation of BRCA tumour tissue testing. The development of tools for designing and interpreting complex testing in-house validation could improve process quality.
- ovarian neoplasms
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information. There are no additional unpublished data from the study.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Epithelial ovarian cancer (EOC) is the eighth cause of cancer death among women worldwide.1 EOC is a heterogeneous disease with different risk factors, morphologies, bio-molecular features and clinical consequences.2 According to 11 WHO classification, EOCs are classified into the following histotypes: high grade serous carcinoma (HGSOC), low grade serous carcinoma, mucinous carcinoma, endometrioid carcinoma (EC), clear cell carcinoma (CCC), Brenner tumours, seromucinous tumours and undifferentiated carcinoma. Among these, HGSOC are the most prevalent group (70%–80%), while EC and CCC account both for 10% and the others for a minority of cases.
The BRCA1 and BRCA2 (BRCA) genes are tumour suppressor genes involved in many cellular pathways.3 They were discovered nearly 25 years ago because some germline BRCA variants are associated with a high risk of breast and ovarian cancers (HBOC).4 5 Since then, the identification of BRCA pathogenic variants (PVs) has been introduced in clinical practice to improve early diagnosis and/or to reduce cancer risk.6 Several Next Generation Sequencing (NGS) approaches were developed and validated for BRCA testing on blood-derived DNA.7–10 Apart from inherent bioinformatics NGS limitations, the detection of germline mutations is not a technical challenge because it is expected that 50% of the DNA should carry the variant, if present. Today, the complexity of this test is mainly related to variant classification as the ability to discriminate the risk of cancer associated with each particular BRCA variant is limited by the wide mutational spectrum and the sequence variability within the normal population of the BRCA genes.11 It is recognised that the interpretation of germline BRCA testing requires an ad hoc expertise since the use of available classification tools and databases is still complex.12 13
BRCA PVs were found in a large fraction of HGSOC (20%–40%) and also in other histotypes (EC 10%; CCC 5%), with the marginal exclusion of the mucinous subtype.14 15 Cancer cells with BRCA loss of function are sensitive to Poly [ADP-ribose] polymerase inhibitors (PARPi) and, recently, the results of several clinical trials showed an impressive improvement of disease-free intervals in patients with platinum-sensitive EOC assuming PARPi as maintenance therapy, both after recurrence or first line medical therapy.16 17 Although most of BRCA PVs are germline, nearly a third of them are somatic.18–20 In addition, it was demonstrated that reversion mutations can partially restore the BRCA function in the tumour, leading to secondary resistance to PARPi.21 Therefore, BRCA testing on different biological specimens (eg, needle biopsy, surgical specimen, circulating tumour cells/DNA) is presently introduced in the practice of many laboratories worldwide.
In the last 5 years, several studies demonstrated that BRCA testing on formalin-fixed paraffin-embedded (FFPE) EOC specimen is reliable,22–24 and some authors suggested that a protocol starting from universal tumour tissue testing may be more efficient even for the diagnosis of HBOC compared with the genetic counselling/testing strategy.25 However, the implementation of NGS-based tumour tissue analysis poses several additional difficulties as compared with germline testing due to preanalytical limitations that may severely affect test results.26 Each laboratory has to perform an internal validation before introducing NGS-based BRCA tumour tissue testing in its current practice. Guidelines for validation of NGS-based oncology panels27 and specific technical recommendations for BRCA tumour tissue testing were published.28 29
In the Liguria region, HBOC genetic counselling/testing were centralised at the Hereditary Cancer Unit (HCU) of the IRCCS Ospedale Policlinico San Martino. As a first step of the implementation of a regional network for BRCA tumour tissue testing, a validation protocol was designed on the basis of current literature recommendations and a study was conducted to evaluate the performance of the preanalytical standards at the eight regional pathology units and of NGS analyses at the reference laboratory.
Materials and methods
Patients with EOC who underwent germline BRCA testing at the HCU between September 2014 and January 2020, and gave written informed consent to research studies, were retrospectively identified. Germline testing was performed by NGS and Mutation Ligation-dependent Probe Amplification (MLPA) analyses. Patients with a histological diagnosis of non-borderline and non-mucinous primary tubo-ovarian carcinoma performed in one of the eight Ligurian pathology units were considered eligible. Surgical specimens obtained from primary debulking surgery, post neoadjuvant chemotherapy surgery, diagnostic laparoscopy or needle biopsy were selected.
This retrospective series of EOC specimens was prepared according to standard protocols. In brief: after the surgical excision, all the specimens were sent unfixed to the pathology units where they have been fixed in 10% buffered formalin (12–18 hours); after grossing, the samples have been routinely processed and paraffin embedded to obtain histological slides stained in H&E. The paraffin blocks were kept in dedicated archives, at room temperature, in cardboard boxes kept away from dust, light and heat sources.
The most significant paraffin block was selected for molecular analysis according to the following criteria: optimal fixation/storage, high representativeness of the entire neoplasia, high tumour cellularity, low percentage of stroma cell, fibrosis and necrosis. Histological subtype and tumour, node, metastases staging were reviewed by two pathologists’ expert in gynaecological pathology (VGV and MP). Additional areas of EOCs showing a somatic BRCA PV were selected to estimate the threshold of mutation detection of our NGS analysis protocol with respect to low tumour cellularity (<<50%) and intratumour heterogeneity. From each selected sample, manual macrodissection was performed and sections (three sections of 10 µm thickness) were obtained for molecular analyses.
Sample preparation and NGS analysis
DNA was isolated using the DNA FFPE Tissue Kit with an automated device (QIAsymphony, Qiagen). For samples with low starting material, the DNA was manually extracted using the same kit (MinElute Columns, Qiagen). DNA concentration was assessed by the Qubit 3.0 Fluorometer (ThermoFisher Scientific) and DNA quality was assessed by the Agilent TapeStation 2200 (Applied Biosystems). A modification of the initial DNA extraction protocol (GeneRead DNA FFPE Treatment Kit) was introduced to reduce formalin-fixation induced artefacts by adding the enzyme Uracil-N-Glycosylase (UNG).30
BRCA tumour tissue analysis was performed using the Oncomine BRCA Research Assay (ThermoFisher Scientific). Both library preparation and chip loading were automatically performed on the Ion Chef System and sequencing was run on the Ion S5 System. Analyses of sequencing raw data were performed with Torrent server software V.5.10.1 and processed using the Ion Reporter V.5.10 software. Parameters for analysis excluded variants with: variant allele frequency (VAF) <5%, coverage <500X, quality score (PHRED) <30 and total variants>40. Copy number variations (CNV) have not been evaluated in these analyses. All variants identified by NGS were verified via visual inspection of .bam alignment files and confirmed by Sanger sequencing. For variants with ambiguous quality bioinformatic data, Sanger sequences were analysed with MinorVariantFinder software (ThermoFisher Scientific) in order to identify minor variants at a detection level as low as 5%. Tissue samples NGS data were analysed blindly to the germline/somatic status.
Variants were annotated according to Human Genome Variation Society nomenclature. Variants were classified as pathogenic or likely pathogenic (collectively termed pathogenic) according to the Evidence-based Network for the Interpretation of Germline Mutant Alleles (ENIGMA) criteria31 and to the American College of Medical Genetics and Genomics (ACMG) recommendations.32
Samples, preanalytical and analytical phases
The clinical and pathological characteristics of the 67 EOC specimens were summarised in table 1. The median archiviation time was 30 months (range 1 month–5 years). The majority of samples (64 of 67) were ≤3 years old. In 43 cases (64.2%) samples were stored in the archive of the Anatomic Pathology University Unit of the IRCCS Ospedale Policlinico San Martino, whereas 24 (35.8%) in the eight regional pathology units (Ligurian BRCA Working Group).
In 20 of 67 (29.8%) samples, DNA fragmentation assay was performed and confirmed good DNA quality (online supplementary table S1). In the initial sequencing runs, three samples did not meet the criteria for NGS analysis as a conspicuous number (>40) of C:G to T:A artefacts were observed. Quality criteria were met after additional tumour areas were tested (OCS7, OCS8) and after employing the UNG enzyme DNA extraction protocol (OCS12 and all subsequent specimens) (online supplementary table S1).
A total of 18 sequencing runs were performed; the first 16 samples were run in duplicate and the subsequent sample sets were run one time (online supplementary table S2). A mean depth of 3800 (range 1807–5382) with 99% of target base coverage at 500X was observed.
An overall tumour test success rate of 98.5% (66 of 67) was reached. The only sequencing failure was observed for the DNA extracted from a very small macrodissected area (tumour cell content 90%) of a sample derived from a diagnostic laparoscopy for tubal HGSOC showing a histological measure of cancer cells of mm 2×1 (OCS67). The other 66 specimens showed a mean tumour cell content of 54% (range 15%–90%).
Result of BRCA testing according to tumour characteristics
The results of BRCA tumour tissue analysis were reported in table 2. Nineteen of 66 tumours (28.7%) harboured a BRCA PV: 12 (63.2%) were germline and 7 (36.8%) were somatic (table 3). Variant of uncertain significance (VUS) were detected in 5 of 66 EOCs (7.5%): all of them were germline variants. HGSOC presented 18 of the 19 observed BRCA1 PVs, with a detection rate of 30.5% (18 of 59). The tumour tissue analysis identified all germline variants previously identified in the blood DNA (concordance 100%).
All the seven somatic BRCA PVs were identified at high VAF (range 51%–91%). This mutant allele fraction was comparable with the tumour cell content (50%–60%) (table 3). From two of these cases (OCS6 and OCS8), additional tumour areas (n=8) with a tumour cell content below 50% (range<10%–35%) were analysed: the BRCA PVs were detectable in all these specimens (table 4). Different spatial areas of the neoplastic mass were also analysed in the same cases: areas with a tumour cell content between 40% and 60% were sampled (n=13) and a VAF ranging from 24% to 71% was detected (table 5).
Tables 6–8 were prepared as a framework for designing and interpreting an in-house validation study.
Samples and preanalytical phase
When planning an in-house validation study, the first decision to be taken is how to select the samples for validation. Guidelines suggest the use of a sufficient number of in-house cases and of reference materials as controls. According to Jennings et al,27 a suitable number of cases for an NGS validation study is (a minimum of) 59 samples. It is also very important to select sample representative of the types of specimen that will be routinely analysed. Cases selected in this study included a high prevalence of advanced HGSOC, similarly to consecutive series of EOC cases. Also, the 28% detection rate of BRCA PVs was in agreement with the literature, including the distributions of germline (63%) and somatic mutations (37%).18–20 Therefore, retrospective series of cases analysed for germline BRCA testing may be a valuable source of material for the validation of BRCA tumour tissue analysis. In fact, this type of study allows a direct comparison of germline and somatic test results: their concordance is an important part of the validation process, particularly if universal tumour tissue testing will be proposed, because a false negative tumour test may preclude testing of at risk family members (table 6).
The failure of NGS analysis is mainly related to the quality of DNAs that can be derived from archival samples. In the context of an organised healthcare network, it is important to verify the suitability of FFPE specimen from each pathology unit referring samples to the reference laboratory. Around 5% of FFPE BRCA NGS analyses in universal EOC testing fail and need to be repeated on additional samples, with no test result available in nearly 3% of cases.25 In this study, the only NGS failure was in an EOC case with a poor quantity of available tumour tissue, suggesting that the processing of such type of samples is a present limit of our protocol. After the introduction of UNG treatment, no critical aspects related to the preanalytical phase emerged in the 66 samples, indicating that the EOC standard procedures of the Ligurian pathology units are of overall good quality. However, the number of samples for each regional pathology unit and from post-therapy or relapsed tumour tissue was limited. Indeed, in designing a validation study, it is important to take into account the many factors that can irreparably affect the outcome of an FFPE DNA extraction. Challenges can potentially arise from different sampling types. In primary debulking, an extensive amount of neoplastic tissue is often available but more attention is required in fixation to avoid cold ischaemia artefacts. In postchemotherapy debulking surgery, scant damaged cancer cells may be present within a large amount of reactive tissue. In fine-needle biopsy, the size and quality of the tissue may be scarce. Finally, the tumour cell content has to be defined and reported to the laboratory as it affects the sensitivity of NGS analysis. In our study, samples were reviewed by two pathologists. Formal training may represent an important step in implementing this protocol if tumour areas for NGS analysis are selected at referring pathology units.
Analytical phase: NGS performance
In the present study, the NGS protocol followed manufacturer’s instructions. The basic performance characteristics we obtained confirmed a high reliability of target enrichment and subsequent sequencing. Also, a high reproducibility was observed when running the same samples in different runs. These results are in agreement with the core quality metrics reported in the literature, supporting a good analytical performance of our protocol (table 7).
The NGS technique is still affected by several difficulties regarding the bioinformatics analysis that need to be standardised when entering in the diagnostics phase.33 In our study, a commercial pipeline was used and, therefore, validation of an in-house bioinformatics pipeline is outside the scope of our validation protocol. Using a validated pipeline is probably common in many diagnostic laboratories. However, it is important to enrich the bioinformatics competence of laboratory professionals to address issues related to NGS analysis interpretation.34
The identification of somatic large deletion/duplication (CNV) is challenging. At the time of this study, reliable bioinformatics pipelines were not available for amplicon-based NGS tumour analysis. MLPA analysis may represent an additional technique for CNV detection but this quantitative method must be in itself validated.23 Therefore, CNV identification was outside the scope of the present study and it represents a limit of our validation protocol.
Variant analysis and classification
A given number of samples with known results must be present in the validation process to estimate analytical sensitivity and specificity. In our study, the presence/absence of BRCA germline variants was known due to the previous blood test and a 100% concordance was observed, confirming that tumour tissue testing can identify germline variants with high sensitivity. Conversely, we could not estimate somatic false negatives as control samples for somatic mutations were not available. Finding somatic BRCA control specimen may not be easy for each local laboratory: to improve in-house validation studies, in the context of a public health service, national NGS oncology panel testing validation networks could be implemented to make characterised sample available. Participation to external quality control schemes may be part of a validation process and it is recommended to monitor continuing performance (table 8).
The identification of somatic BRCA PVs is complicated by the fact that EOCs generally become more heterogeneous during the course of disease and this heterogeneity might result in a non-uniform distribution of genetically distinct cell subpopulations across and within tumour sites, leading to a significant intratumour heterogeneity. We took advantage of specimens derived from areas with various tumour cell content and from different spatial regions of two early stage EOCs with somatic BRCA PVs to estimate the lower limit of detection and to evaluate intratumour heterogeneity, respectively. All the analysed areas showed a VAF well above 5% and, although a difference in VAF was observed, intratumour heterogeneity was not so high to justify multi-sampling in routine BRCA tumour tissue testing. However, this finding needs to be confirmed in a large sample set including advanced stage cancers.
In this study, a BRCA1 donor splice-site variant (c.4096+1G>A) was identified and classified as (likely) pathogenic (table 3). At present, the classification of this variant differs among the ENIGMA (VUS) and ACMG (pathogenic) guidelines as ENIGMA takes into account provisional clinical and RNA research data that question its pathogenicity.31 In the past, ±1–2 canonical splice site BRCA variants were classified as pathogenic because they were assumed to lead to the formation of aberrant messenger (m)RNAs. More recently, they are classified by ENIGMA as likely-pathogenic (95% probability of pathogenicity) and the mRNA analysis is required to demonstrate a splicing defect as exceptions to this role were observed. In fact, the BRCA genes contain several naturally occurring RNA isoforms35 36 and it was demonstrated that some BRCA spliceogenic variants are not pathogenic.37 38 This example highlights the importance that the interpretation of BRCA testing is performed by personnel appropriately trained to handle the complex issues underlying the classification of various types of BRCA variants. In addition, specific criteria for somatic BRCA variant classification need to be developed as BRCA variants conferring a high risk of cancer may not necessarily confer a therapeutic response.39 40 To rapidly transfer information from research to clinical setting, ideally the same laboratory should perform both germline and tumour tissue testing and should be part of national/international networks. In this context, a strict collaboration of geneticists, pathologists and clinicians is required not only for a proper classification of the detected variants but also for the proper clinical interpretation of negative results (ie, to discuss the appropriateness of additional germline CNV testing on blood DNA).
The implementation of NGS-based BRCA tumour tissue analysis encloses numerous complexities and gaps of knowledge. A close collaboration of regional pathology unit networks with the genetics unit provides advantages in all phases of the process. The development of practical tools to design and interpret results of in-house validation studies could improve the process quality.
Take home messages
Next Generation Sequencing-based BRCA tumour tissue analysis has been proposed as universal testing for epithelial ovarian cancer (EOC) due to its implications in EOC medical therapy but its standards are still under development.
The design and interpretation of in-house validation studies of BRCA tumour tissue analysis need cautious attention given the present complexities and gaps of knowledge.
Specimen processing and selection can severely affect the results and interpretation of BRCA tumour tissue analysis.
Interpretation of the results of BRCA tumour tissue analysis is improved by the presence of a strict collaboration with genetics units.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information. There are no additional unpublished data from the study.
The study has been approved by the Ligurian Ethical Committee (472REG2015).
The authors would like to thank the Ligurian BRCA Working Group for their contributions to the study. They are also thankful to laboratory technicians of the Hereditary Cancer Unit and of the Anatomic Pathology University Unit for their technical support.
Handling editor Runjan Chetty.
DR, MP, VGV and LV contributed equally.
Collaborators Ligurian BRCA Working Group: Paola Baccini (Anatomic Pathology Unit, ASL 4, Sestri Levante, Italy); Rodolfo Brizio (Anatomic Pathology Unit, ASL 1, Sanremo, Italy); Paolo Dessanti (Anatomic Pathology Unit, ASL 5, La Spezia, Italy); Ezio Fulcheri (Fetal, Placental & Gynecologic Pathology Unit, IRCCS Istituto Giannina Gaslini, Genoa, Italy); Marina Gualco (Anatomic Pathology Unit, ASL 3, Genoa, Italy); Eugenio Marinaro (Anatomic Pathology Unit, Ospedali Galliera, Genoa, Italy); Mariangela Rutigliani (Anatomic Pathology Unit, Ospedali Galliera, Genoa, Italy); Ezio Venturino (Anatomic Pathology Unit, ASL 2, Savona, Italy).
Contributors LV and VGV conceived and designed the study. GA contributed to the technical part relating to the preanalytical phase. DR, MP and VG performed the experiments and analysed the data. All authors wrote the paper and approved the final version.
Funding This project was supported by AstraZeneca under liberal donation to the University of Genoa (VGV).
Disclaimer The funding organisation had no influence on the conduct of the project and the evaluation of the data.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.