Patients whose blood samples were collected between 2003 and 2010 were selected from the registry of the Genetic Counseling Unit, Warsaw Cancer Center-Institute of Oncology. The personal and familial cancer histories were acquired by in-depth interviews. Healthy individuals with no known personal or familial history of malignancy, normal results of screening colonoscopy, and normal mammography or PSA levels were recruited primarily from the National Colorectal Cancer Screening Program. All patients and control subjects were Polish Caucasians recruited from the Masovian voivodeship population. Informed consent for hereditary cancer genetic testing was obtained from all of the patients. The permission for genetic testing was granted by the Ethical Committee at Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology on 9 May 2002 (No. 28/2002) and further extended to include testing with NGS on 11 June 2013 (No. 28/2002/1/2013). The study did not require ethical approval.

Genomic DNA was extracted from whole blood as described previously [7]. DNA concentration was determined on a Qubit 2.0 Fluorometer using the Qubit dsDNA BR Assay Kit (Life Technologies). For library preparation, a set of reagents from the Ion AmpliSeq™ Library Kit 2.0 (Life Technologies) and Ion AmpliSeq BRCA1 and BRCA2 Panel, which comprises 167 primer pairs spanning 16.25 kb, were used to amplify the coding regions of the BRCA1 and BRCA2 genes. Amplification for each patient was performed using three separate primer pools containing 1.4 μl of 5× Ion AmpliSeq™ HiFi Master Mix, 3.5 μl of 2× Ion AmpliSeq Primer Pool, and 3 ng of DNA in a total reaction volume of 7 μl. Cycling conditions were as follows: 99°C for 2 min, followed by 19 cycles of 99°C for 15 sec and 60°C for 4 min, ending with a hold at 10°C. At this point, three separate reactions for a given patient were combined, and amplicons were partially digested at primer sequences with 2 μl of FuPa reagent under conditions provided by the manufacturer in The Ion AmpliSeq DNA and RNA Library Preparation protocol (Revision B.0). Subsequent steps of library preparation were performed according to the above protocol. Briefly, following digestion, samples were subjected to sequencing adapter ligation with either 16 or 32 indexing barcodes, and purified with the AMPure XP reagent (Beckman Coulter). The DNA library was eluted from AMPure XP beads with 52 μl of PCR reagent solution consisting of 50 μl of Platinum®PCR Super Mix HiFi and 2 μl of Library Amplification Primer Mix, and amplified under the following conditions: 98°C for 2 min, followed by five cycles of 98°C for 15 sec and 64°C for 1 min, ending with a hold at 10°C. Next, the amplified DNA library was subjected to a two-round purification process using AMPure XP beads with 0.5× and 1.2× bead-to-sample volume ratios, respectively. Library concentrations were determined with the Qubit dsDNA HS Assay Kit (Life Technologies), and the respective size distributions were determined with the High Sensitivity DNA Analysis Kit on Bioanalyzer 2100 (Agilent).

Library molarity was determined, and either 16 or 32 libraries were pooled in equimolar concentrations and used for automatic template preparation on Ion Chef using reagents from the Ion PGM IC 200 Kit and Ion 316 Chip Kit v2 BC. Sequencing was performed on the Ion Torrent PGM sequencer using 500-flow runs. Data from the PGM runs were processed on the Ion Torrent server using a platform-specific pipeline incorporated in Torrent Suite v4.2.1 (Life Technologies) to obtain sequence reads, trim adapter sequences, filter and remove poor signal reads, and assign the reads to a given barcode. The reads were mapped to the hg19 (Homo sapiens) reference genome and adjusted to the specific amplicon target regions deposited in the “BRCA1_2.20131001.designed” BED file available in Torrent Suite. The coverageAnalysis (v4.2.1.4) and variantCaller (v4.2.1.0) plug-ins with a set of default parameters optimized for the BRCA1/BRCA2 panel were obtained from www.Ampliseq.com, and were ran for each sequencing dataset.

Variants were classified either as single nucleotide variants (SNVs) or insertion/deletion (indels) variants using a python script. Variants were annotated using data from BIC for the BRCA1 and BRCA2 genes. The identified variants were matched to existing variants in BIC according to four parameters as follows: chromosome number (chromosome 13 and chromosome 17 for BRCA2 and BRCA1, respectively); position of the variant in the human genome assembly hg19 (GRCh37), which was retrieved from the BIC database field HGVS; genomic and type of reference; and alternate base. The possibility of different alternate and reference base was taken into account as well as the possibility of variant on the reverse strand.

All the SNVs were analyzed using Variant Effect Predictor [17] (VEP) for the human genome assembly hg19 (GRCh37). The occurrence of variants was determined using an R script. Variants with unknown clinical significance in BIC and new changes were subjected to in silico analysis of their deleteriousness, and screened for evidence of pathogenicity in the literature.

SNVs and indels were analyzed using Condel [18] and SIFT Indel [19], respectively. Missense variants were classified as pathogenic when reported in BIC as pathogenic, and/or predicted as pathogenic by a Condel score, and identified through a literature search [20-29]. Frameshift mutations were classified as pathogenic when reported as pathogenic by BIC and/or predicted as pathogenic by SIFT Indel.

Pathogenic mutations detected by NGS were confirmed by Sanger sequencing. In addition, five selected variants, namely, Q356R, N3124I, Q563X, c.7249delCA, and c.9118-2G > A, were genotyped using Custom TaqMan SNP Genotyping Assays (Life Technologies, USA) and a 7900HT Real-Time PCR system (Life Technologies) as described previously [7].

The incidence of SNVs and indels within 167 amplicons covering the BRCA1 and BRCA2 exons was tested using the Ion Torrent PGM sequencer in 31 (a training set) and 512 women with breast cancer or ovarian cancer newly diagnosed before the age of 50 years who were positive and negative for BRCA1/2 mutations, respectively, as determined by targeted genotyping. The genotyping comprised 11 mutations in BRCA1, namely c.66_67delAG, C61R, c.3700_3704del5, c.3756delGTCT, c.3777delT, c.4035delA, c.4041delAG, c.4065delTCAA, c.5263delC, R1738E and R1751X, and nine mutations in BRCA2, namely E394X, c.5239insT, c.5946delT, c.5964delAT, c.6447delTA, c.7910del5, c.8924delT, R3128X and c.9402delT [7,30]. Of the 512 women, 317 had familial breast and/or ovarian cancer, and 195 had only early onset cancer and/or contralateral breast and ovarian cancers; the median age of women with breast cancer at diagnosis was 43 years.

The average number of bases with ≥ Q17/≥Q20 across all 21 PGM sequencing runs was 343/324 Mbp, constituting 96/90% of total output. The average coverage across all samples was 507, whereas the mean mapping rate reached 94%.

The NGS-based procedure confirmed the pathogenic mutations in all 31 DNA samples of a training set (not shown). Among 512 DNA samples analyzed, we identified 146 SNVs (64 in BRCA1 and 82 in BRCA2) and 32 indels (20 in BRCA1 and 12 in BRCA2). Sixteen indels were found in coding regions (nine in BRCA1 and seven in BRCA2). The results of the detailed analysis of SNVs and indels are summarized in Additional file 1.

The types and number of SNVs are summarized in Table 1. Of these, eight SNVs (five in BRCA1 and three in BRCA2) were already reported in BIC as pathogenic. Three additional SNVs were reported as variants of uncertain significance (VUS); variants co-localized with these SNVs were reported as pathogenic in ClinVar. In addition, we identified three novel nonsense variants (Table 2). Of 41 missense variants that were either reported as VUS in BIC or had not been reported previously, 24 had at least one transcript in which the Condel score was reported as deleterious (Table 3).Table 1

Types and number of single nucleotide variants in the BRCA1 and BRCA2 genes

Of the indels identified in the present study, five in BRCA1 and two in BRCA2 had already been reported in BIC as pathogenic. One additional indel was reported as unknown in BIC, although its co-located variants in dbSNP were reported in ClinVar as pathogenic (Table 4). We also found nine novel indels in coding regions (Table 4).Table 4

Pathogenic indels in BIC, ClinVar, or SIFT indel prediction

Altogether, among 512 patients in whom the previous targeted genotyping did not reveal a pathogenic BRCA1/2 mutation massive amplicon sequencing identified 52 patients with 31 deleterious mutations, of which sixteen were frameshift, eight were nonsense, three were missense, and four affected splicing. Of note, one patient carried two pathogenic mutations, namely Q563X and N3124I. All pathogenic variants were confirmed by Sanger direct sequencing. Five other variants were likely pathogenic, and 17 variants were VUS. All of them were missense mutations (Table 3).

Using the TaqMan SNP Genotyping Assay, the incidence of five selected variants, one common polymorphism (Q356R) and four mutations (N3124I, Q563X, c.7249delCA, and c.9118-2G > A) were tested in additional group of 445 cancer patients diagnosed under the age of 50 and 1300 healthy individuals (856 women and 444 men with a median age of 59 years).

Although Q356R was described as pathogenic in ClinVar, its prevalence was similar among patients (15.1%) and healthy controls (16.9%). Therefore, the clinical significance of this polymorphic variant was not confirmed. By contrast, Q563X was detected in 5 (1.1%) out of 445 cancer patients, respectively, but only in one healthy individual. The three other mutations (N3124I, c.7249delCA, and c.9118-2G > A) were detected by TaqMan genotyping only in DNA samples in which they were identified by NGS.