Background

The identification of BRCA1 and BRCA2 germline mutations as predictors of cancer susceptibility significantly improved the diagnosis and prevention of hereditary breast and ovarian cancers (HBOC). Recent advances in genetic testing have enabled the discovery of novel genes that increase the risk of cancer in patients with familial predisposition. Multiple research laboratories have evaluated these cancer-associated mutations in patients who are negative for BRCA1/2 mutations, but still have a high risk of HBOC. These efforts have identified mutations in moderate-risk genes, such as ATM, BRIP1, CHEK2, BARD1, MRE11A, NBN, RAD50, RAD51, and XRCC2, as well as those in high-penetrance genes, including TP53, PTEN, STK11, CDH1, and PALB2, have been reported across diverse ethnic populations [1].

Next generation sequencing (NGS) can provide detailed genetic information via multi-gene panel assays [2]. However, the application of NGS multigene panel test in a clinical setting represents a considerable challenge. It is necessary to not only validate this novel technique, but also to select candidate susceptibility genes. Furthermore, mutations indicative of cancer susceptibility vary across ethnicities; therefore, it is important to understand the clinical and genetic characteristics of multiple susceptibility genes identified by NGS multigene panels in each ethnic population.

In this study, we used comprehensive multigene panels that included 35 known or suspected cancer susceptibility genes to examine BRCA1/2 mutation-negative Korean patients who had clinical features indicative of hereditary breast cancer. We also investigated the feasibility of multigene panel testing for Korean patients, and evaluated potential clinicopathological risk factors related to germline mutations other than BRCA1/2.

Methods

Study population

The study population included 182 Korean BRCA1/2 mutation-negative breast cancer patients with a familial predisposition who were referred to the Cancer Prevention Center, Yonsei Cancer Center, Seoul, Korea between March 1, 2015 and November 11, 2016. Sixty-two patients opted to not participate. Finally, a total of 120 patients were enrolled in the study. Suspected clinical features of hereditary breast cancer were defined as follows: (1) at least one case of breast or ovarian cancer in first- or second-degree relatives; (2) a first diagnosis of breast cancer before age 40; (3) bilateral breast cancer; and (4) co-diagnosis of breast and ovarian cancers in the same patient.

Panel-based mutation analysis

Germline DNA was extracted from the participants’ peripheral blood samples. We used a customized targeted capture sequencing panel (OncoRisk®, Celemics, Seoul, Korea) which included all coding sequences and intron-exon boundaries of the coding exon from 35 cancer predisposition genes (BRCA1, BRCA2, PALB2, BARD1, BRIP1, RAD51C, RAD51D, RAD50, NBN, MRE11A, ATM, CHEK2, TP53, PTEN, APC, BLM, BMPR1A, CDH1, CDK4, CDKN2A, EPCAM, MEN1, MLH1, MSH2, MSH6, MUTYH, PMS2, POLE, PRSS1, RET, SLX4, SMAD4, STK11, VLH, and WT1). Products with each capture reaction were sequenced by 100 base pair paired-end reads on a MiSeq platform (Illumina, San Diego, CA). High-quality sequencing data with an average depth of 500−1000 folds were obtained.

We identified all single base pair substitutions, insertion-deletions, and copy number variants (CNVs) in each gene. Split-read-based detection of large insertions and deletions was conducted using the Pindel and Manta algorithms. CNVs detected by ExomeDepth software [3] were further crosschecked with our custom pipelines, which retrieved base-level depth of coverage for each binary alignment map (BAM) file using SAMtools software (http://samtools.sourceforge.net) and normalized the depths in the same batch (Additional file 1: Figure S1). All likely deleterious mutations were validated by Sanger sequencing, and all possible large rearrangements were confirmed by the multiplex ligation-dependent probe amplification (MLPA) method (Additional file 1: Figure S2).

Genetic variants were classified using a five-tier system following guidelines from the American College of Medical Genetics and Genomics (ACMG) as follows: pathogenic, likely pathogenic, variants of unknown significance (VUS), likely benign, or benign/polymorphism [4]. We used the Sorting Intolerant From Tolerant (SIFT, http://sift.bii.a-star.edu.sg/) and Polymorphism Phenotyping-2 (PolyPhen-2, http://genetics.bwh.harvard.edu/pph2) to generate in silico predictions of several of the identified nonsynonymous variants. Using large rearrangements of exons, pathogenic and likely pathogenic variants were considered as mutations, for consistency with previous studies [5].

Results

Baseline characteristics of the patients are presented in Additional file 2: Table S1. A total of 7.5% (9/120) of patients were found to carry at least one pathogenic or likely pathogenic variant. A total of ten gene variants (Fig. 1a) were identified in nine patients: TP53 in two patients, PALB2 in three patients, BARD1 in two patients, BRIP1 in two patients, and MRE11A in one patient. We detected a large deletion from exon 2−9 in the TP53 gene, and the other pathogenic variants identified were as follows: PALB2 (c.3267_3268delGT, p.Phe1090SerfsTer6, rs587781890; c.2257C > T, p.Arg753Ter, rs180177110; and c.695delC, p.Gly232ValfsTer6); BARD1 (c.1345C > T, p.Gln449Ter); BRIP1 (c.1066C > T, p.Arg356Ter, rs730881633; and exon 5–6 deletion); and MRE11A (c.1773_1774delAA, p.Gly593LysfsTer4). Likely pathogenic variants were found in TP53 (c.733G > A, p.Gly245Ser, rs28934575). Pathogenic variants in PALB2 and MRE11A were identified in a 34-year-old patient who was co-diagnosed with breast and gastric cancer (Table 1). Three of the pathogenic variants identified in this study were not reported previously.

Fig. 1
figure 1

a Percentage of patients with pathogenic or likely pathogenic mutations corresponding with each gene. b Number of patients with variants of uncertain significance (VUS) for each gene (n = 120 patients total)

Full size image
Table 1 Characteristics of patients with pathogenic or likely pathogenic variants
Full size table

A total of 87 patients (72.5%) had at least one VUS (median, 1; range, 0–3). A total of 139 VUS were identified in 30 cancer susceptibility genes, including SLX4 (n = 11), BLM (n = 10), POLE (n = 10), ATM (n = 9), CDH1 (n = 9), CHEK2 (n = 9), BRCA2 (n = 8), RAD50 (n = 7), BRIP1 (n = 6), EPCAM (n = 5), PALB2 (n = 5), PRSS1 (n = 5), TP53 (n = 5), APC (n = 4), MLH1 (n = 4), RET (n = 4), MRE11A (n = 3), MSH2 (n = 3), MSH6 (n = 3), MUTYH (n = 3), RAD51D (n = 3), STK11 (n = 3), BMPR1A (n = 2), BRCA1 (n = 2), CDKN2A (n = 1), MEN1 (n = 1), NBN (n = 1), PMS2 (n = 1), VHL (n = 1), and WT1 (n = 1) (Fig. 1b).

First diagnosis of breast cancer at a relatively young age (<35 years) was correlated with pathogenic or likely-pathogenic variants in high-penetrance cancer susceptibility genes. Pathogenic variants in high-penetrance genes were detected in 21.1% (4/19) of these patients, which was significantly higher than that for patients who were first diagnosed with breast cancer at age ≥ 35 years (1/101, 1.0%, p = 0.003) (Table 2).

Table 2 Association between the clinicopathological features of suspected hereditary breast cancer and the pathogenic or likely pathogenic variants of non-BRCA cancer predisposition genes (n = 120 patients)
Full size table

Discussion

Previous studies using multigene panel tests identified cancer susceptibility genes in 2.1−16.8% of BRCA1/2 mutation-negative patients [5,6,7,8,9,10,11]. Our tests of high-penetrance genes identified a large exon deletion in TP53, and pathogenic and likely pathogenic variants in TP53 and PALB2 (Table 1). We also identified a frameshift mutation of MRE11A c.1773_1774delAA (p.Gly593LysfsTer4) in a patient with a PALB2 mutation. The MRE11 protein functions in non-homologous end-joining and homologous recombination, which occur during the repair of double-stranded DNA breaks [12]. Therefore, the risk for patients with concurrent dysfunction in PALB2 and MRE11A is unclear and should be assessed in future studies. Because the two frameshift variants in PALB2 (c.3267_3268delGT, p.Phe1090SerfsTer6, rs587781890; and c.695delG, p.Gly232ValfsTer6) were not found in the control group, the variants met the criteria to be likely pathogenic according to the ACMG guideline (PVS1 and PM2) (Table 1) [4]. One nonsense variant in PALB2 (c.2257C > T p.Arg753Ter, rs180177110) had a higher prevalence in affected patients compared to the control group [odds ratio (OR), 127.0; 95% confidence interval (CI), 14.1–1140.1; p < 0.0001]. Therefore, this variant conformed to the criteria to be classified as pathogenic according to ACMG guidelines (PVS1 and PS4) (Table 1) [4]. In addition, a missense variant in TP53, c.733G > A (p.Gly245Ser, rs28934575) was classified as a pathogenic or likely pathogenic variant in the ClinVar database (http://www.ncbi.nlm.nih.gov/clinvar/), and met the criteria for a likely pathogenic variant according to the ACMG guidelines (PM2, PM5, PP2, PP3, and PP5) (Additional file 2: Table S2) [4].

Pathogenic or likely pathogenic variants also were detected in BRCA1-associated RING domain 1 (BARD1) and BRCA1-interacting protein C-terminal helicase 1 (BRIP1). BARD1 and BRIP1 encode proteins that interact with the BRCA1 protein during the repair of DNA double- stranded break, and pathogenic variants of these genes have been investigated [13]. However, there is a controversy as to whether these rare variants are clinically associated with a risk of breast cancer [11, 14]. In a previous study that screened for BRIP1 mutations among 235 Korean patients with BRCA1/2 mutation-negative high-risk breast cancers using fluorescent-conformation sensitive gel electrophoresis (F-CSGE), there was no case of a protein-truncating BRIP1 mutation, which suggests that the prevalence of BRIP1 mutations is likely to be low in the Korean population [15].

Cell cycle checkpoint kinase 2 (CHEK2) is a well-established moderate-penetrance breast cancer gene. Several studies have shown that essentially no case of CHEK2 (c.1100delC) was observed in Asian populations, in contrast to the observed prevalence in European populations [16,17,18,19]. Liu and colleagues reported that the CHEK2 c.1111C > T (p.His371Tyr, rs531398630) variant was observed in 4.24% (5/118) of Chinese familial breast cancer cases without BRCA1/2 mutations, and was associated with dysfunctional phosphorylation of T68 in the SQ/TQ rich domain, which is an activation point following DNA damage [18]. We also identified CHEK2 c.1111C > T variants in 2.5% (3/120) of Korean breast cancer patients without BRCA1/2 mutations (Additional file 2: Table S2). Population-based investigations are required to establish the prevalence of this variant, especially in Asian patients. We identified the CHEK2 c.908 + 2delT variant in one patient, and it was classified as likely pathogenic according to the ACMG guideline (Additional file 2: Table S2). However, we did not classify this variant as a positive result because the experimental study was not sufficient.

In the current study, clinically important likely pathogenic or pathogenic variants of high-penetrance genes were identified in only five (4.2%) patients (TP53 in two patients, and PALB2 in three patients). These variants were identified in 4 of 19 patients (21.1%) with early-onset breast cancer (< 35 years old at onset) (Table 2). A previous study identified cancer susceptibility mutations in 11% of BRCA1/2-negative patients with early-onset breast cancer (diagnosed at <40 years of age) [20]. Considering the frequency of pathogenic variants of high-penetrance genes in patients with early-onset cancer, clinicians should be encouraged to consider performing multigene panel tests for these patients if their conventional BRCA1/2 tests are negative.

This study has several limitations. The primary limitation is the small number of patients (n = 120), which provides only limited data for cancer susceptibility genes in Korean patients with breast cancer. A large-scale cohort study will be required to establish the accurate prevalence and spectrum of pathogenic variants in these patients. The majority of patients (87 of the 120, 72.5%) had VUS. A functional and population-based study will be necessary to clarify the clinical meaning of these VUS. Despite these limitations, to the best of our knowledge, this is the first prospective study to apply customized multigene panels to BRCA1/2 mutation-negative Korean patients with a high risk for HBOC. A recent study conducted by Couch et al. assessed the commercial multigene panel test results of 65,057 patients with breast cancer; however, the frequency, phenotypic association, and cancer risks related to each variant were analyzed among Caucasian women only [11]. Regarding diversity of prevalence of the genetic variants, more prospective studies will be required among diverse ethnic populations.

Conclusions

Wider application of multigene panel tests that include high-penetrance cancer susceptibility genes, so-called “beyond BRCA1/2 genes”, will likely provide clinically relevant information for some patients with high risk for hereditary cancer [1, 13, 21]. However, these panels can produce abundant and conflicting results in clinical practice. To efficiently utilize these data, clinical databases should be established with respect to ethnic backgrounds, and genetic results should be carefully applied for high-risk patients.