Screening for BRCA1, BRCA2, CHEK2, PALB2, BRIP1, RAD50, and CDH1 mutations in high-risk Finnish BRCA1/2-founder mutation-negative breast and/or ovarian cancer individuals

Introduction Two major high-penetrance breast cancer genes, BRCA1 and BRCA2, are responsible for approximately 20% of hereditary breast cancer (HBC) cases in Finland. Additionally, rare mutations in several other genes that interact with BRCA1 and BRCA2 increase the risk of HBC. Still, a majority of HBC cases remain unexplained which is challenging for genetic counseling. We aimed to analyze additional mutations in HBC-associated genes and to define the sensitivity of our current BRCA1/2 mutation analysis protocol used in genetic counseling. Methods Eighty-two well-characterized, high-risk hereditary breast and/or ovarian cancer (HBOC) BRCA1/2-founder mutation-negative Finnish individuals, were screened for germline alterations in seven breast cancer susceptibility genes, BRCA1, BRCA2, CHEK2, PALB2, BRIP1, RAD50, and CDH1. BRCA1/2 were analyzed by multiplex ligation-dependent probe amplification (MLPA) and direct sequencing. CHEK2 was analyzed by the high resolution melt (HRM) method and PALB2, RAD50, BRIP1 and CDH1 were analyzed by direct sequencing. Carrier frequencies between 82 (HBOC) BRCA1/2-founder mutation-negative Finnish individuals and 384 healthy Finnish population controls were compared by using Fisher's exact test. In silico prediction for novel missense variants effects was carried out by using Pathogenic-Or-Not -Pipeline (PON-P). Results Three previously reported breast cancer-associated variants, BRCA1 c.5095C > T, CHEK2 c.470T > C, and CHEK2 c.1100delC, were observed in eleven (13.4%) individuals. Ten of these individuals (12.2%) had CHEK2 variants, c.470T > C and/or c.1100delC. Fourteen novel sequence alterations and nine individuals with more than one non-synonymous variant were identified. One of the novel variants, BRCA2 c.72A > T (Leu24Phe) was predicted to be likely pathogenic in silico. No large genomic rearrangements were detected in BRCA1/2 by multiplex ligation-dependent probe amplification (MLPA). Conclusions In this study, mutations in previously known breast cancer susceptibility genes can explain 13.4% of the analyzed high-risk BRCA1/2-negative HBOC individuals. CHEK2 mutations, c.470T > C and c.1100delC, make a considerable contribution (12.2%) to these high-risk individuals but further segregation analysis is needed to evaluate the clinical significance of these mutations before applying them in clinical use. Additionally, we identified novel variants that warrant additional studies. Our current genetic testing protocol for 28 Finnish BRCA1/2-founder mutations and protein truncation test (PTT) of the largest exons is sensitive enough for clinical use as a primary screening tool.


Introduction
Breast cancer (BrCa) is the most common cancer among women in Finland, with about 4,000 cases diagnosed yearly (Finnish Cancer Registry). It has been estimated that a monogenic trait accounts for 5 to 10% of all BrCa cases [1]. The two major high-penetrance BrCa genes, BRCA1 (breast cancer 1) and BRCA2 (breast cancer 2), are responsible for 30% of hereditary breast cancer (HBC) cases worldwide, but only for about 20% in Finland [2][3][4]. BRCA2 mutations have been found to be more common in the Finnish population than BRCA1 [5]. In addition to BRCA1 and BRCA2 mutations, there are certain hereditary cancer syndromes, such as Li-Fraumeni, Cowden, Peutz-Jeghers and diffuse gastric cancer syndromes, associated with a high risk of BrCa [6][7][8][9]. However, these syndromes very seldom explain HBC.
BRCA1 and BRCA2 have many DNA damage response functions in the cell [10]. Therefore, it has been hypothesized that genes coding for proteins that interact with BRCA1 or BRCA2 or act in the same DNA repair pathway would be likely candidate genes for HBC susceptibility. As expected, CHEK2 (checkpoint kinase 2), PALB2 (partner and localizer of BRCA2), BRIP1 (BRCA1-interacting protein 1), and RAD50 (human homolog of Saccharomyces cerevisiae RAD50) have been shown to have rare, moderate-risk BrCa-associated variants, which have also been studied in the Finnish population [11][12][13][14]. In addition, BrCa-associated variants have been reported in the CDH1 (cadherin-1) [15].
Although mutations in many genes have been found to predispose an individual to BrCa, approximately 75 to 80% of HBC cases remain unexplained [16]. It is likely that additional BrCa susceptibility gene mutations remain unidentified, especially in the category of moderate-to low-penetrance gene variants that individually confer only minimal risk but that, through multiplicative and/or cumulative effects, can cause relatively high risk for the carriers [17]. Genome-wide association studies (GWAs) have revealed multiple low penetrance, single nucleotide polymorphisms (SNPs) in many genes and chromosomal loci with increased risk of BrCa. For example, SNPs in the fibroblast growth factor receptor 2 (FGFR2) gene have shown significant association with increased risk among BrCa cases with strong family history [18].
To address the problem of heterogeneous HBC in genetic counseling, we wanted to investigate possible additional mutations in HBC-associated genes. The aim of this study was to screen seven known BrCa susceptibility genes for additional mutations in 82 well-characterized, Finnish, high-risk hereditary breast and/or ovarian cancer (HBOC) individuals tested to be BRCA1/ 2-founder mutation negative. In addition, the sensitivity of our current BRCA1/2 mutation analysis protocol was defined for genetic counseling purposes.

Patients and controls
Index individuals of 82 high-risk Finnish HBOC families were screened for germline alterations in BrCa-associated genes. All individuals had been detected to be founder mutation-negative by minisequencing of the previously known 28 Finnish BRCA1/2 mutations and by protein truncation test (PTT) of exon 11 for BRCA1 and exons 10 and 11 for BRCA2. Study material had been collected from the individuals, who visited the Tampere University Hospital Genetics Outpatient Clinic between January 1997 and May 2008. The hospital district, in the area of Pirkanmaa, consists of over 20% (1.23 million) of the Finnish population. Individuals were chosen to be included in this study according to the following criteria of high-risk HBC: (a) the individual or her first-degree relative (only female family members were included when defining first-degree relatives) had BrCa or ovarian cancer (OvCa) at younger than 30 years of age; or (b) two first-degree relatives in the family had BrCa and/or OvCa and at least one of the cancers had been diagnosed at younger than 40 years of age; or (c) three first-degree relatives in the family had BrCa and/or OvCa and at least one of the cancers had been diagnosed at younger than 50 years of age; or (d) four or more first-degree relatives had BrCa and/or OvCa at any age; or (e) the same individual had BrCa and OvCa. Patient with bilateral BrCa was considered to have two separate cancers. According to these criteria, our study material also included 11 non-affected females in addition to 71 BrCa and/or OvCa patients. We were also able to get blood samples from two affected relatives in 2 out of 11 separate families with healthy index. These relatives with BrCa were screened for the same variant as that identified in the index. The clinical data of the studied individuals are presented in Table 1. As controls, 384 blood samples from anonymous healthy females, collected from the Finnish Red Cross, were used. All individuals have been informed of the analyses, and they have given written consent to use their already existing DNA samples. Permission for the research project has been received from the Ethical Committee of Tampere University Hospital and the National Authority for Medicolegal Affairs.

Statistical analyses
Carrier frequencies between 82 studied individuals and 384 population controls were compared by using Fisher's exact test [21]. All P-values were two sided. Odds ratios (OR) were generated by two-by-two table.
In silico prediction of novel missense variants effects The effects of five novel-coding missense variants, BRCA2 c.72A > T (Leu24Phe), CHEK2 c.1363G > A (Val455Ile), PALB2 c.814G > A (Glu272Lys), PALB2 c.1000T > G (Tyr334Asp), and RAD50 c.1544A > G (Asp515Gly), were predicted with a number of tools by using Pathogenic-Or-Not-Pipeline (PON-P) [22]. The predictions included those for amino acid tolerance (programs PolyPhen version 2, Sift, PhD-SNP, SNAP) and protein stability (I-Mutant version 3). PON-P allows simultaneous submission of a number of variations and proteins to selected predictors. PON-P utilizes machine learning to combine results from several individual predictions.

MicroRNA database and BLAST search for novel variants
MicroRNA (miRNA) target site search was performed for the novel variant genomic positions from the micro-RNA database (miRBase) [23]. Also BLAST search [24] was performed for the novel human variant genomic positions to see if these sites are conserved among different organisms including mouse, rat, cow, and chicken.

Results
Index individuals of 82 high-risk HBOC families were screened for germline alterations in BRCA1, BRCA2, CHEK2, PALB2, BRIP1, RAD50, and CDH1 genes. Detailed clinical information of analyzed individuals is shown in Table 1. All of the identified 54 sequence variants with their observed genotype frequencies and rsnumbers are presented in Supplementary Table S1 in Additional file 1. All of the identified non-synonymous and novel sequence alterations are summarized in Table 2. Table 2 variants are presented in Table 3 with index individual and family cancer history. In addition, as our study material also included healthy index individuals from 11 families, we made an effort to get blood samples from two affected relatives in 2 out of 11 separate families. These relatives with BrCa were screened for the same variant as that identified in the healthy index. Analysis was performed for the new cases in family 112 (CHEK2 c.470T > C and PALB2 c.1676A > G variants) and family 231 (BRCA1 c.4883T > C variant; Table 3). In family 112, the case proved to have the same PALB2 c.1676A > G variant as the index individual but in family 231, the affected relative did not carry the BRCA1 c.4883T > C variant (data not shown). To further evaluate the impact of these 11 healthy index cases, we recalculated the frequencies without these 11 individuals for those variants accepted to be meaningful for BrCa risk. Supplementary Table S2 in Additional file 2 shows these re-calculated frequencies for BRCA1 c.5095C > T, CHEK2 c.470T > C, and CHEK2 c.1100delC variants. No statistically significant effect was seen for exclusion of the 11 cases.

BRCA1 and BRCA2 mutation analysis
Analysis of BRCA1 and BRCA2 revealed altogether 16 different sequence variants, seven in BRCA1 and nine in BRCA2 [see Supplementary Table S1 in Additional file 1]. All but two of the identified variants in BRCA1, c.4883T > C and c.5095C > T, have been reported to be neutral in the databases. Heterozygous c.4883T > C variant was observed in 4 of 82 (4.9%) women of which three had BrCa and one had a family history of breast, cervix and skin cancers (Tables 2 and 3). In population controls, the frequency of the c.4883T > C variant was 6 of 367 (1.6%). The c.5095C > T variant has been classified as a deleterious mutation in the Breast Cancer Information Core (BIC) database. The heterozygous c.5095C > T mutation was observed in 1 of 82 (1.2%) women. The mutation carrying woman had BrCa diagnosed at the age of 42 years and a strong family history of cancer (Tables 2 and 3, Figure 1, Family 249). Additional mutation analysis also revealed two other affected women carrying the c.5095C > T mutation in the same family. In BRCA2, three of the nine identified variants were novel, c.68-80insT, c.72A > T, and c.793 + 34T > G (Tables 2 and 3.). The heterozygous missense variant c.72A > T (Leu24Phe), was observed in 1 of 82 (1.2%) women but not in population controls. The c.72A > T variant carrying woman had BrCa diagnosed at the age of 53 years. She had also two affected first-degree relatives (mother and sister). Protein predictions by PON-P suggested that substitution of leucine by phenylalanine in position 24 changes significantly the properties of the side chain and the substitution would not be tolerated. All the other identified variants in BRCA2 have been reported previously and they are either neutral or the clinical significance of the variants is yet uncertain especially with the three missense variants, c.8182G > A, c.9976A > T and c.10234A > G (Tables 2 and 3 [46]. c The NCBI dbSNP [46]. d The Breast Cancer Information Core database [47]. e Reported in the Finnish population by Vahteristo et al. [11]. f Heterozygous deletion or insertion. g Due to the high frequency of the variant observed in analyzed individuals, variant is not presented in Table 3.   were identified either in BRCA1 or BRCA2 by multiplex ligation-dependent probe amplification (MLPA).

CHEK2 mutation analysis
In CHEK2, two previously reported BrCa-associated variants in the Finnish population, c.470T > C and c.1100delC, were identified in 10 of 82 (12.1%) individuals (Tables 2 and 3). The heterozygous c.470T > C variant was observed in eight women of which three were healthy. Two of the c.470T > C variant carriers had bilateral BrCa and they carried also PALB2 missense variants (an example of the family pedigree of the index individual carrying the both variants is presented in Figure 2, Family 129).
The heterozygous c.1100delC variant was detected in three women (Tables 2 and 3 (Tables 2 and 3). Previously reported PALB2 missense variants, c.1010T > C, c.1676A > G, c.2794G > A, and c.2993G > A were identified here with frequencies from 1.2% to 12.2% in analyzed individuals (Tables 2 and 3) but the variants have not been associated with BrCa risk (an example of the family pedigree of the index individual carrying the c.1676A > G variant in addition to the BRIP1 c.584T > C variant is presented in Figure 7, Family 131).

BRIP1, RAD50, and CDH1 mutation analysis
In BRIP1, two silent [see Supplementary Table S1 in Additional file 1] and two missense variants (Tables 2  and 3) were identified. All of the identified variants have been reported previously and they are likely to be neutral. In RAD50, altogether seven sequence alterations were observed [see Supplementary Table S1 in Additional file 1] and three of these were novel ( Table 2). The novel missense variant, c.1544A > G (Asp515Gly), was observed in 1 of 82 (1.2%) women and in 4 of 384 (1.1%) population controls. The c.1544A > G variant carrying woman had BrCa diagnosed at the age of 39 years and one affected first-degree relative (Table 3). According to protein predictions, c.1544A > G variant is likely to be neutral. Two other novel variants, c.2398-32A > G and c.3475 + 33C > G, were both observed with the frequency of 1 of 82 (1.2%) in analyzed individuals (Tables 2 and 3). In CDH1, 10 different sequence alterations were identified [see Supplementary Table S1 in Additional file 1]. All of the variants have been reported previously and they are likely neutral. c.1544A > G variant position shared similarities with mouse, rat, cow and chicken. Three novel variant positions in CHEK2 exon 11 and the RAD50 c.3475 + 33C > G variant shared sequence similarity between mouse, rat and cow. Variants that occur in the genomic regions that are conserved across species may indicate a pathogenic role.

Discussion
In the present study, we screened BrCa susceptibility genes in 82 Finnish high-risk HBOC individuals with no known Finnish BRCA1/2-founder mutations. As genetic counseling and surveillance is greatly needed for these individuals and their families, we decided to study BRCA1/2 in more detail and also to analyze five additional genes that had previously been associated with BrCa risk. The majority of known BRCA1/2 alterations are small insertions and deletions or point mutations (BIC database). Also, large genomic rearrangements have been reported in both genes with varying frequencies in different populations [26]. In Finland, so far only Pylkäs et al. have reported a large deletion in BRCA1 identified in a Finnish OvCa family [27]. In our study, no deletions or duplications were found in either BRCA1 or BRCA2 by MLPA, which suggests the existence of more restricted alterations. A total of 16 different sequence variants were identified from these two genes [see Supplementary Table S1 in Additional file 1] and only one of the identified variants, c.5095C > T in BRCA1, has been classified as a clinically significant mutation in the BIC database. In line with this classification, our BrCa patient carrying this variant had a strong family history of cancer (Tables 2 and 3, Figure 1, Family 249) and two other variant carriers with BrCa were also observed in the same family. The c.5095C > T mutation thus can explain a fraction of the BrCa cases also in the Finnish population. The clinical significance of the BRCA1 c.4883T > C variant in BrCa predisposition is uncertain [28,29]. Our data support the idea that it is a low-penetrant risk allele, because the variant was observed to be three times more common in analyzed high-risk individuals than healthy population controls (Tables 2 and 3). Novel variant findings in BRCA2 (Tables 2 and 3) warrant additional studies, especially the novel missense variant, c.72A > T (Leu24Phe), which was shown not to be tolerated by protein prediction. Prediction indicated that the substitution decreases the stability of the produced protein and this might be the mechanism behind the disease for this variant. The amino acid position 24 is located near the N-terminal part of BRCA2. Amino acids 1 to 40 interact with PALB2, and sequence variants in this region have been shown to have effects on the PALB2 and BRCA2 interaction and thus are suspected to have a role in cancer predisposition [30]. The role of the three BRCA2 missense variants, c.8182G > A, c.9976A > T, and c.10234A > G, in HBOC risk, is uncertain [31][32][33]. All three heterozygous variants were observed in two healthy women with a history of BrCa, one carrying the c.9976A > T variant and the other both the c.8182G > A and c.10234A > G variants (Tables 2 and 3, Figure 8, Family 005). At this stage, because we only have samples from the index individuals, no segregation analyses of the variants have been performed, but these families clearly warrant additional studies. In recent risk models, it has been suggested that multiple low-risk variants within the same individual may actually cause a significantly elevated risk for the carrier [17]. The overall low frequency of new variants identified in BRCA1/2 genes suggests that the present protocol for testing 28 Finnish BRCA1/2-founder mutations and PTT of the largest exons is adequate for clinical use to detect the majority of harmful mutations in these two genes in the Finnish population.
Two of the CHEK2 variants, c.470T > C and c.1100delC, have been widely studied in BrCa predisposition in Finland and elsewhere. Previous studies have shown that the c.1100delC allele confers about a twofold elevated BrCa risk in women, whereas c.470T > C is a lower risk variant [34,35]. Both variants also associate with other cancers in the Finnish population [36][37][38]. In our study, two of the CHEK2 variants, c.470T > C and c.1100delC, were identified in 10 out of 82 analyzed individuals (12.2%) suggesting that the contribution of the two CHEK2 variants to BrCa risk is remarkable in the high-risk Finnish BRCA1/2-founder mutation-negative individuals. However, clinical screening of the CHEK2 variants has not yet been justified due to unclear clinical consequences related to incomplete segregation of the variants with BrCa in the high-risk BrCa families [39,40]. Based on the findings of this study, we agree that interpretation of the CHEK2 mutation analysis results is very difficult, because many other gene variants were also identified in individuals with either c.470T > C or c.1100delC variants and some of the variant carriers had not (yet) been diagnosed with BrCa. Thus profound segregation analysis of the c.470T > C and c.1100delC variants for BRCA1/2-founder mutation-negative families would be needed to further study clinical significance of these variants. Also the novel variants identified in CHEK2 should be further analyzed.
PALB2 has been associated with BrCa predisposition in Finland by Erkko et al. [12] and the c.1592delT variant was classified as a Finnish founder mutation. In this study the founder deletion was not found, which is probably explained by the limited number of analyzed high-risk HBOC individuals. We identified two novel PALB2 missense variants, c.814G > A (Glu272Lys) and c.1000T > G (Tyr334Asp), in affected individuals (Tables  2 and 3). Protein predictions suggested a non-pathogenic role of these substitutions but further studies are needed to confirm these findings. None of the four previously reported PALB2 missense variants, c.1010T > C, c.1676A > G, c.2794G > A, and c.2993G > A, have been associated with BrCa risk [12,41]. Interestingly, these variants were identified also together with other variants in analyzed individuals (Tables 2 and 3). One of the identified intronic variants, c.2586 + 58C > T, has been associated with an increase of BrCa risk in a Chinese population [25] but there is no evidence of that in the Finnish population.
BRIP1 and RAD50 genes have been shown to have rare BrCa associated variants in familial BrCa patients [14,42]. Here, BRIP1 mutation analysis revealed only previously reported likely neutral variants. Whereas analysis of RAD50 identified three novel sequence alterations, including one missense variant, c.1544A > G (Asp515Gly) ( Tables 2 and 3). To further study the role of these novel variants, additional analyses are needed. Germline mutations in CDH1 have been previously found to associate with hereditary diffuse gastric cancer syndrome, but mutations have been also identified in familial invasive lobular BrCa patients without hereditary diffuse gastric cancer [15,43]. Here, only neutral variants were identified, and all of them have been reported earlier [see Supplementary Table S1 in Additional file 1]. No clear results were found that any of the identified genetic variants in BRIP1, RAD50, or CDH1 would increase the BrCa/OvCa risk in the analyzed high-risk Finnish HBOC individuals.

Conclusions
In this study, 13.4% of the analyzed, high-risk BRCA1/2founder mutation-negative HBOC cases can be explained by previously reported mutations in BrCa susceptibility genes. CHEK2 mutations, c.470T > C and c.1100delC, make a considerable contribution (12.2%) to these high-risk individuals but further segregation analysis are needed to evaluate the clinical significance of these mutations before applying them in clinical use. Novel variant findings warrant additional studies with special interest in the novel missense variant, BRCA2 c.72A > T (Leu24Phe), which was predicted to bear untolerated mutations and to destabilize the protein.
The complex nature of HBOC addresses the need for genome-wide approaches to further study these individuals and to create new tools for genetic counseling. This study also confirmed that our current genetic testing protocol for the 28 Finnish BRCA1/2-founder mutations and PTT of the largest exons is sensitive enough for clinical use in the majority of Finnish HBC/HBOC individuals.

Additional material
Additional file 1: Supplementary Table S1. All of the identified 54 sequence alterations. Supplementary Table S1 include detailed information about all of the identified sequence alterations.