Patients cohort

Among consecutive patients who underwent BRCA tumour testing through ION Torrent-based sequencing between August 2017 and February 2022, we retrospectively selected 222 high-grade ovarian cancer (HGOC) patients with the following histological subtypes: 203 serous (HGSOC), seven endometrioid, five clear-cell and seven with mixed histotypes.

Since NGS BRCA1/2 tumour testing was not available before 2017 in our Institution, 19 of 222 subjects underwent germline testing before tumour sequencing based on personal (very early age at diagnosis/previous breast cancer) or family history. According to the workflow used by our Molecular Tumour Board (MTB), in 73 out of 203 patients with upfront tumour testing subsequently received genetic counselling, either for targeted germline sequencing of a tumour-detected PV or for large genomic rearrangement analysis14.

The Ethics committee of Fondazione IRCCS Istituto Nazionale dei Tumori of Milan approved the use of both clinical and molecular data collected by the MTB for clinical studies and granted exemption from requiring written consent for tumour genetic testing from the patients, as these analyses were carried out in the context of a diagnostic and care setting (Approval Number INT 227/20). All the probands who underwent germline testing were aged over 18 and provided signed informed consent for the use of their biological samples and data for both diagnostic and research purposes. All methods were carried out in accordance with relevant guidelines and with the ethical principles of the Declaration of Helsinki.

The BRCA1 and BRCA2 genes were assessed by in-house NGS testing using the Oncomine BRCA Research Assay (Thermo Fisher Scientific, Inc). This assay provided a 100% coverage of all BRCA1 and BRCA2 exons, with an average of 64 bases of intronic flanking sequences upstream and downstream of each exon. Five m sections from formalinfixed paraffin-embedded (FFPE) samples were manually microdissected to isolate the highest percentage of neoplastic cells. Genomic DNA was extracted with protease K (incubation ON at 55C) and quantified with Qubit dsDNA BR kit (Thermo Fisher Scientific, Inc). The libraries were prepared with the IonAmpliSeq Library kit 2.0 (Thermo Fisher Scientific, Inc) and quantified with Qubit dsDNA HS kit (Thermo Fisher Scientific, Inc) following the manufacturers instructions. The libraries are diluted to 25pm, pooled and loaded on the Ion Chef to perform emulsion PCR and chip loading on 318 v2 chips. Sequencing was performed on ION PGM, using the HI-Q view Chef kit, according to the manufacturers instructions. Data were processed using the Torrent Suite 5.12.3 (TS). The quality of sequencing output was first evaluated through the plugin Coverage Analysis on the TS. Only samples whose librarys uniformity and on-target values were at least 80% and with a medium coverage of 1500X were considered valid. SNV analysis was performed in duplicate: the first variant calling was generated by the Variant Caller plugin from the TS and the resulting VCF file was loaded in the Variant Effect Predictor Tool (Ensembl, Version GRCh37) for the variants annotation. To eliminate erroneous base calling, we set each variant coverage>40X, a variant frequency on each sample>2% and a quality value>30. Variants within homopolymer (HP) longer than eight bases and with strand bias80% were not reported. In the second analysis, the BAM files were automatically uploaded from the TS to the Ion Reporter Software (IR, version 5.6 to 5.16) and the variant calling was integrated into the analysis pipeline Oncomine BRCA Research Somatic318. The results of both analyses were manually compared. Each variant was displayed on IGV (ver. 2.3.97). Synonymous variants were filtered out, while the remaining variants were classified into pathogenicity classes according to the Evidence-based Network for the Interpretation of Mutant Alleles (ENIGMA) consortium guidelines (https://enigmaconsortium.org/). Our assay could not reliably detect large intragenic rearrangements.

Two EDTA tubes of peripheral blood samples were collected from each patient who performed genetic counselling and was eligible for germline testing, either for targeted sequencing of tumour-detected pathogenic/likely pathogenic variants or for the analysis of large genomic rearrangements in patients with no actionable variants detected at tumour testing. Whole blood DNA was isolated through the MagCore Super automatic workstation with the MagCore Genomic DNA Whole Blood Kit (Diatech LabLine SRL, Jesi, Italy). Targeted Sanger sequencing of tumour-detected BRCA1/2 PVs was performed on purified PCR products by using BigDye Terminator v.3.1 Cycle Sequencing kit (Thermo Fisher Scientific, Inc.) and run on 3730Xl DNA Analyzer (Applied Biosystems; Thermo Fisher Scientific, Inc.), after purification with Agencourt CleanSeq-Beckman Coulter. Sequences were analysed by Mutation Surveyor Software (v5.0.1; SoftGenetics, LLC., State College, PA, USA). Targeted sequencing results were confirmed on both blood aliquots collected from each patient. Variants of uncertain clinical significance identified at tumour testing were not systematically investigated at the germline level. Eligible probands, who resulted negative at tumour testing with the Oncomine BRCA assay, were analysed for large deletions and duplications of BRCA1 and BRCA2 on blood DNA with the SALSA MLPA kits P045 BRCA2/CHEK2 and P002 BRCA1 probe mix (MRC-Holland, Amsterdam, the Netherlands), following the manufacturers instructions. MLPA products were run on the 3730Xl DNA Analyzer (Applied Biosystems; Thermo Fisher Scientific, Inc.) with the Gene Mapper Module (Applied Biosystems; Thermo Fisher Scientific, Inc.). The results were analysed through the Gene Marker Software v2.7.0 (SoftGenetics, LLC, State College, PA, USA).

Based on previous observations on the performance at homopolymers of ion semiconductor sequencing techniques, we focussed our analysis on stretches of six or more identical bases since the calling accuracy has been consistently shown to dramatically drop beyond this length15,17,25,28,30. We thus selected all 29 homopolymeric regions exceeding five repetitions to be analysed within the coding regions of both genes, including five in BRCA1 and 24 in BRCA2. Since truncating variants beyond codon 3326 of BRCA2 are not classified as high-risk variants, homopolymers downstream of the residue c.9976 of BRCA2 were not included in the analysis (Table 2).

To overcome the limitations of the ION reporter software, which filters out most indels at homopolymeric regions, we manually visualised the BAM alignment files of the 222 patients at the 29 regions with the IGV software (ver. 2.3.97). The median depth of coverage of the regions of interest ranged from 1045 to 6989X. Each sample showed a variable frequency of sequence alterations (both insertions and deletions) at each region. We estimated the variant allele frequency (VAF) of insertions and deletions (indels) by calculating the ratios of the maximum inserted or deleted reads over the total reads at each homopolymer (Suppl. Table 1).

Since the VAF of indels at homopolymeric regions has, in general, a left-skewed distribution, we employed a modified version of the Cancer Outlier Profile Analysis (COPA) approach31, which consists in scaling the above-cited to a normal distribution and subsequently calculating the outliers that exceeded the mean+3 median-adjusted deviations (+3) threshold.

To validate the outliers, for each of the 29 regions, we further defined a threshold based on the normalized distributions (either for percentage of insertion or deletion) of a control population. Since both in ovarian and other BRCA-associated cancers the predominant second hit is most often represented by loss of heterozygosity (LOH), while a second point mutation is an extremely rare event32,33,34, we used as control population a cohort of 46 patients in which a non-homopolymeric PV (either somatic or germline) had been already identified.

To avoid potential selection bias, which would affect the estimated frequency of pathogenic variants occurring at homopolymeric regions in our cohort, we excluded from the analysis the 19 patients who underwent germline testing before tumour testing. This group also included two patients with germline-confirmed homopolymeric PVs who resulted negative at tumour testing.

Therefore, we applied the (+3) thresholds estimated on the control population to the normalized distributions of the study cohort, composed of 157 individuals with no evidence of pathogenic variants at tumour testing with ION Torrent. In addition, according with filtering criteria used in a previous study, which focused on germline variants29, we considered only homopolymeric indels with an absolute VAF above 15% and in any case higher than the maximum value of the control population at each homopolymeric region. Lastly, regions with a total read count of less than 100 were excluded from the analysis.

Targeted Sanger sequencing was performed on tumour DNA to confirm the occurrence of outlier homopolymeric indels selected by using the defined thresholds.

More:
Prevalence of BRCA homopolymeric indels in an ION Torrent-based ... - Nature.com