Introduction

Breast cancer (BC) is the most common cancer among women in western countries, including Finland. Inherited BC risk is known to be associated with rare, highly penetrant variants, mainly single nucleotide polymorphisms (SNPs) and small insertions and deletions (indels) in BRCA1 and BRCA2, which account for nearly 20% of hereditary breast and/or ovarian cancer (HBOC) cases in Finland [1]–[3]. Additionally, variants in other BRCA1/2 interacting genes, including CHEK2, PALB2, RAD51C, and Abraxas, are known to account for a low proportion of HBOC susceptibility in the Finnish population [4]–[7].

In addition to SNPs and small indels, copy number variations (CNVs) contribute to susceptibility to complex diseases and disorders [8]. A CNV is a segment of DNA (1 kb or larger) that presents an altered copy number compared with the reference genome [9]. Depending on the location, CNVs may affect target gene expression through a dosage effect or by disrupting gene regulatory elements [10]. CNVs were initially associated with neurological disorders, but studies have demonstrated the role of CNVs also in other diseases, including several cancers [11]–[15].

Despite the fact that several heritable risk factors for breast and ovarian cancer have been recognised, in the majority (up to 80%) of HBOC families, inherited risk is likely explained by yet unknown factors, which makes genetic counselling and clinical surveillance challenging. The contribution of rare germline CNVs to breast and ovarian cancer susceptibility has also been established in the Finnish population, but their role is mostly unexplored [16], [17]. Therefore, new information regarding germline CNVs and their role in HBOC predisposition is needed to identify CNVs that may be used clinically to facilitate the genetic counselling of HBOC families.

To determine additional genetic factors contributing to HBOC susceptibility in the Finnish population and gain new information for genetic counselling, we analysed germline CNVs in a cohort of 84 well-characterised HBOC BRCA1/2-founder mutation-negative Finnish individuals who have been pre-screened for the most common high- and moderate-penetrant genes [18].

Materials and Methods

Results

We performed genome-wide CNV analysis with a SNP array targeting regions of known cytogenetic importance for individuals from 84 Finnish HBOC families and 36 healthy controls. After applying the quality control criteria, 81 HBOC individuals and 35 controls (n = 116) were included in the data analysis. The aim of this study was to identify germline CNVs contributing to HBOC susceptibility in Finnish families.

The PennCNV program was used to detect 545 autosomal CNVs at 273 different genomic regions in HBOC individuals and controls (n = 116). All of the identified CNVs are presented in detail in Table S1. A summary of the CNVs identified by PennCNV are shown in Table 1. The most important observations are that the average number of CNVs was slightly higher in HBOC individuals compared with controls, and deletions were more frequent in HBOC individuals. There was no statistically significant difference in the median size the CNVs between the HBOC individuals and controls (52.3 kb vs. 50.5 kb; P = 0.90). However, the median deletion size in HBOC individuals was smaller compared with the controls (39.2 kb vs. 56.8 kb; P = 0.07). In contrast, the median duplication size was significantly larger (P = 0.01) in HBOC individuals compared with controls (68.7 kb vs. 47.5 kb).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of the identified copy number variations (CNVs) by PennCNV in 81 hereditary breast and/or ovarian cancer (HBOC) individuals and 35 controls.

https://doi.org/10.1371/journal.pone.0071802.t001

Annotation of all of the 545 CNVs against the genes in the NCBI RefSeq database revealed 313 (57.4%) gene-affecting CNVs (Table 1). Most importantly, gene-affecting deletions were more common in HBOC individuals compared with controls (Table 1). The identified CNVs were compared with healthy control sample data collected in the Database of Genomic Variants. The main observation was that the proportion of novel deletions to all deletions in HBOC individuals was nearly three times larger compared with controls (Table 1). In contrast, novel duplications in HBOC individuals were observed less frequently compared with controls (Table 1).

In this study, we focused on CNVs with the following characteristics: they were enriched in HBOC individuals compared with controls and 1) affected known or potential genes contributing to HBOC predisposition (3 CNVs); or 2) they were homozygous, and carriers presented with interesting clinical outcomes (1 CNV); or 3) they were not reported in the Database of Genomic Variants and affected genes related to BC (2 CNVs). CNVs of interest were confirmed by another program (QuantiSNP or cnvPartition). In total, six CNVs were selected for further validation by qPCR, and they were genotyped in additional cohort of index individuals from 20 HBOC families and five of the CNVs were genotyped in 299–869 additional healthy female controls (Table 2). The CNVs were correlated with clinical data from the HBOC individuals (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Validated copy number variations.

https://doi.org/10.1371/journal.pone.0071802.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. The clinical characteristics and family cancer history for HBOC individuals analysed in the SNP array with the six validated copy number variations.

https://doi.org/10.1371/journal.pone.0071802.t003

All six validated CNVs listed in Table 2 are located in genomic regions related to BC. CNVs in the intronic regions of ERBB4 and EPHA3 were enriched in HBOC individuals compared with controls (Table 2). EPHA3 and ERBB4 encode proteins that are involved in important signalling pathways. A homozygous deletion in the 5q15 locus was identified in one BC patient (1 out of the 101, 1.0%) with drastic clinical characteristics (Tables 2 and 3). This homozygous deletion was observed only in 1 out of the 899 (0.1%) healthy controls (Table 2). Deletions affecting the intronic region of the CSMD1 tumor suppressor gene and exonic regions of the highly penetrant BRCA1 were observed only in 1 out of the 101 (1.0%) HBOC individuals and CSMD1 deletion was identified in 1 out of the 436 (0.2%) controls (Table 2). Because large deletions in BRCA1 are known to predispose to HBOC, there was no need to screen for the deletion in additional controls. A duplication affecting the coding region of the ERVV-2 gene, which belongs to endogenous retroviruses, was more commonly homozygous in HBOC individuals compared with controls (Table 2).

The clinical characteristics and family cancer history for individuals with HBOC with the six validated CNVs are presented in Table 3 (only CNVs identified in our original cohort of 81 HBOC individuals first analysed in the SNP array are presented). Most importantly, 2 out of the 5 individuals with HBOC with a novel deletion at the 2q34 ERBB4 locus had bilateral BC that was diagnosed at ≤43 years of age (Table 3; families 221 and 212). We were able to analyse the segregation of the 2q34 deletion in family 249 (Table 3) in which a deleterious BRCA1 variant was previously identified in three individuals (Figure 1) [18]. The 2q34 deletion was identified in the index’s mother (homozygous) and two paternal cousins (heterozygous) (Figure 1). However, the index’s daughter did not carry the deletion (Figure 1). A common feature for all of the 3p11.1 deletion (at EPHA3 locus) carriers was ductal BC diagnosed at ≤50 years and positive hormone receptor status (6 out of the 8 carriers) in the cohort of 81 HBOC individuals (Table 3). In the second cohort of 20 additional HBOC individuals, 3p11.1 deletion was identified in two BC patients, one ovarian cancer patient and a patient who had both breast and ovarian cancer (File S2). Interestingly, all three patients with BC presented ductal form of the cancer and estrogen and progesterone receptor positive status (File S2). Intergenic deletion in the 5q15 region was of great interest because it was found as a homozygous deletion in a BC patient who was diagnosed at age 29 years and died of BC at the same age (Table 3, family 123). Additionally, one heterozygous 5q15 deletion carrier had BC diagnosed at an early age (24 years) and the other had thyroid and cervical cancers in addition to BC diagnosed before age 40 years (Table 3; families 250 and 246). A novel deletion of high interest at 8p23.2, which affects the CSMD1 intronic region, was identified in a patient with ductal grade 2, hormone receptor positive BC diagnosed at a relatively early age (36 years) with a paternal family history of BC (Table 3, family 128 and Figure 2). A deletion affecting BRCA1, NBR1, and NBR2 at 17q21.31 was identified in a patient with hormone receptor-negative BC with a family history of ovarian cancer (Table 3, family 252).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Family 249 pedigree.

Index individual carries a novel 59.0 kb deletion in the 2q34 locus. The deletion affects intronic region of the ERBB4 gene, which encodes a receptor tyrosine kinase family member that plays an important role in several cellular signalling pathways. The deletion was also identified in index’s mother and two paternal cousins. Mother carried homozygous deletion (indicated with an asterisk). Index’s daughter was tested to be negative for the deletion. Additionally, deleterious BRCA1 c.5095C>T variant has been previously identified in three individuals in the family. Females are marked with circles and males are marked with squares. Index individual is marked with an arrow. Breast and ovarian cancers are marked with black circles with the age at diagnosis. Other cancers are marked with grey and specified with the age at diagnosis (Br: breast, Co: colon, Kid: kidney, Mel: melanoma, Ov: ovarian, To: tongue, Ute: uterus). Deceased individuals are marked with a slash. Current age of index’s healthy sister is indicated. Generations are marked with the Roman numerals on the left. The pedigree figure has been modified from Kuusisto et al, 2011 [18].

https://doi.org/10.1371/journal.pone.0071802.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Family 128 pedigree.

Index individual carries a novel 10.8 kb deletion in the 8p23.2. The deletion affects intronic region of the CSMD1 tumor suppressor gene. Females are marked with circles and males are marked with squares. Number in circle or squares indicates descendants. Index individual is marked with an arrow. Breast cancers are marked with black circles with the age at diagnosis. Other cancers are marked with grey and specified with the age at diagnosis (Br: breast, GI: gastrointestinal, Mel: melanoma). Deceased individuals are marked with a slash. Current ages of healthy females are presented in the paternal side of the family. In addition, the current age of index’s healthy daughter is indicated. Generations are marked with the Roman numerals on the left.

https://doi.org/10.1371/journal.pone.0071802.g002

Enrichment analysis was performed for the CNV-affecting genes to identify common functions of the gene products. EPHA3, ERBB4 and BRCA1 were identified in several GO term categories and pathways that were significantly overrepresented (P<0.05) (presented in detail in Table S2). Both EPHA3 and ERBB4 were identified to have molecular functions related to receptor activity, transmembrane receptor activity, molecular transducer activity, and signal transducer activity. In contrast, BRCA1 was identified in several pathways related to DNA double-strand breaks and repair. In addition, Online Mendelian Inheritance in Man (OMIM) and The Genetic Association database searches revealed the role of CSMD1 in BC.

Discussion

In this study, we aimed to identify CNVs contributing to HBOC susceptibility in Finland and obtain new information for the genetic counselling of HBOC families. We utilised a cohort of well-characterised BRCA1/2-founder mutation-negative individuals from 84 Finnish hereditary breast and/or ovarian cancer families who had been previously screened for variations in seven known BC genes [18].

Here, we identified more gene-disrupting deletions in HBOC individuals compared with controls suggesting that altered function of their protein products, particularly in critical pathways, could explain pathogenic events in HBOC individuals. Additionally, a proportion of novel gene-affecting deletions, which were not reported in healthy controls in the database, was higher in HBOC individuals compared with controls, suggesting that these novel CNVs are more likely to be disease -related.

We focused on CNVs that were enriched in HBOC individuals compared with controls and affected genes that likely play a role in HBOC predisposition. In addition, one intergenic deletion was also included for further validation based on the homozygous form of the aberration and notably poor clinical characteristics of the carrier. Thus, six CNVs were considered to be the most relevant for further validation. Because our sample number in the SNP array was limited, we also genotyped the six CNVs in a cohort of 20 additional HBOC individuals. Furthermore, five of the CNVs were genotyped in 299–869 additional healthy controls. Because clinical characteristics of the additional cohort of 20 HBOC individuals were comparable to our original cohort of 81 HBOC individuals, we combined the observed frequencies of the CNVs in both cohorts in Table 2. Additionally, we performed segregation analysis of one family to determine how the CNV co-segregated with the disease and another BC-associated variant. The CNVs were compared with the clinical data of the HBOC individuals.

In this study, the most frequently observed aberration in HBOC individuals was a deletion disrupting the EPHA3 intronic region (Table 2). EPHA3 belongs to the ephrin receptor subfamily of the receptor tyrosine kinase (RTK) family, which plays an important role in normal cell physiology and disease pathogenesis [24]. Ephrin receptor signalling together with ephrin-ligands is known to regulate both tumour growth and suppression in several different cancers including BC [25]. According to recent studies, altered EPHA3 expression is associated with gastric and colorectal cancers, and CNVs in the EPHA3 region have been found to be associated with haematologic malignancies [26]–[28]. However, haematologic malignancies were not observed in EPHA3 deletion carriers in this study. Our data suggest that an intronic deletion may disrupt the EPHA3 regulatory elements, thus leading to altered protein function and pathogenic BC events. Thus, considering the important role of EPHA3 in signalling pathways, the segregation of the intronic deletion should be studied in the families and the deletion should be further screened in a larger sample set.

The intergenic 5q15 deletion, particularly as a homozygous deletion, is highly interesting from a clinical perspective. This deletion was identified in a patient who had been diagnosed with BC at age 29 and died of the disease at the same age. Homozygous deletion of the 5q15 locus was extremely rare in healthy controls (1 out of the 899, 0.1%) (Table 2), which emphasises the importance of the variation. Moreover, it is possible that a fraction of the anonymous controls may develop breast or ovarian cancer later in life although they were healthy at the time of the blood draw. The 5q15 deletion may affect the transcriptional control of target gene expression. Regulatory elements of the target gene can extend to long distances outside of the transcription unit [29], which makes gene expression regulation a complex process. Interestingly, aberrant expression of the nearest neighbouring gene (1.0 Mb distance), RGMB, has been implicated in BC [30]. Additional analysis is needed to determine whether RGMB regulatory elements exist in the 5q15 deletion locus. Moreover, a previous copy number study of breast tumours-associated aberrations in the 5q15–5q21 locus with p53 status and patient survival suggests that the 5q15 region may be important in BC predisposition [31]. Furthermore, to reveal possible functional elements located in the deletion region, the Encyclopedia of DNA Elements (ENCODE) (http://genome.ucsc.edu/ENCODE/) was utilised. Preliminary analysis revealed enhancer and promotor-associated histone mark (H3K4Me1) activity and DNase hypersensitivity, which indicate that regulatory elements are active in this genomic region. Thus, the 5q15 homozygous deletion requires special attention because it may have clinical significance for screening families with BC with early disease onset. Interestingly, two heterozygous 5q15 loss carriers with lobular BC (Table 3, families 264 and 129) were previously found to carry BC-associated CHEK2 variants [18].

The novel 8p23.2 deletion affects an intronic region in the CSMD1 tumour suppressor gene. CSMD1 has mainly been associated with head and neck squamous cell carcinoma, but CSMD1 losses is also reported to contribute to the tumourigenesis of several other epithelial cancers, including BC [32]. In addition, CSMD1 deletions and aberrant splicing have been shown to contribute to altered CSMD1 function in vivo [32]. Moreover, decreased CSMD1 expression has been associated with high tumour grade and the poor survival of invasive ductal breast carcinoma, and the role of CSMD1 expression as a potential BC prognostic marker has been suggested [33]. In this study, the CSMD1-affecting intronic deletion was identified in the index individual for one BC family (1 out of the 101, 1.0%) (family 128, Figure 2 and Table 3). In this family, the index patient and her paternal aunt and grandmother had been diagnosed with BC at ages 36, 45, and 58 years, respectively (Figure 2). In addition, gastrointestinal cancer was diagnosed on the paternal side of the family (father) (Figure 2). Interestingly, the CSMD1-affecting deletion was observed only in 1 out of the 436 (0.2%) healthy controls, suggesting that this rare variant likely predisposes individuals to BC. We are currently seeking DNA samples from the other family members (family 128) to determine whether the variation co-segregates with BC in the family. In addition, although the deletion should be screened for in larger sample set, the CSMD1 gene is a potential candidate for the further study of HBOC susceptibility in Finnish families.

A novel deletion at 2q34 affects the intronic region of the ERBB4 gene, which is known to play a role in BC [34]. ERBB4 encodes an epidermal growth factor RTK subfamily member that regulates several cellular processes and plays an important role in cancer [35]. We found that the aberration in ERBB4 is 1.5 times more common in HBOC individuals compared with controls suggesting that it may be a disease-related low-risk variant (Table 2). In addition, the clinical features of the ERBB4 deletion carriers were interesting because two of the HBOC individuals had bilateral BC diagnosed at a relatively early age (Table 3). To further analyse the deletion, we were able to perform a segregation analysis in one family in which a deleterious BRCA1 c.5095C>T variant was previously recognised (Figure 1) [18]. Thus, three BC cases in the family (index, index’s daughter and paternal cousin) are explained by the paternally inherited high-penetrant BRCA1 variant. The ERBB4 deletion was observed on the maternal and paternal sides of the family (Figure 1). However, in the mother, who had BC diagnosed at an older age, the ERBB4 deletion was homozygous, suggesting that the deletion could contribute to BC development at an older age, particularly in its homozygous form. Thus, it would be interesting to screen for the deletion in other BC cases diagnosed at an older age on the mother’s side of the family as well. Additionally, an ovarian cancer patient who was negative for the highly -penetrant BRCA1 variant was found to carry a heterozygous form of the 2q34 deletion, suggesting that the deletion may also contribute to ovarian cancer risk to some extent (Figure 1).

BRCA1 deletions are known to predispose to breast/ovarian cancer [36]. In this study, a large deletion overlapping exons 1A-13 of BRCA1 was observed in one individual with BC diagnosed at age 46 years and with ovarian cancers diagnosed in her mother and half-sister (Table 3, family 252). In our previous analysis, the sample was excluded from the MLPA analysis due to a low sample quality value [18]. The BRCA1 deletion encompassing exons 1A-13 has been reported in a Finnish breast/ovarian cancer family [37]. Here, the deletion was found to affect also the neighbouring genes NBR1 (entire gene) and NBR2 (exons 1–10) according to the PennCNV, QuantiSNP and cnvPartition programs. Similar findings have been reported worldwide in a few studies [38], [39]. Because the BRCA1 deletion is known to be clinically relevant, MLPA analysis was performed to validate the BRCA1 deletion (Figure S1). Genetic counselling was offered for the deletion carrier patient.

The duplication identified at 19q13.41 affects exon 1 of the ERVV-2 gene. ERVV-2 belongs to the human endogenous retrovirus (ERV) family and the involvement of ERVs in the pathogenesis of human cancer has been suggested but their roles in biological disease processes are poorly understood [40]. Because 19q13 genomic region has been previously associated with BC [41], this prompted us to further examine the duplication affecting the ERVV-2 coding region. Screening for the duplication in additional controls revealed that it was as common in controls compared with HBOC individuals (Table 2). However, the homozygous form of the variation was 4.4 times more common in HBOC individuals compared with controls (Table 2), suggesting that the aberration may contribute to breast and ovarian cancer risk to some extent, but further studies are needed to confirm the findings. Of interest, one of the homozygous duplication carriers (Table 3, family 240) had been reported to carry a novel BRCA2 variant predicted to be pathogenic [18].

In conclusion, this study is a continuation of our previous work with the aim of elucidating genetic factors contributing to HBOC susceptibility in Finland. We have identified several potential CNVs that likely increase the risk of HBOC susceptibility that may thus explain a fraction of breast and ovarian cancer cases. The aberrations at 3p11.1, 5q15, and 8p23.2 regions require special attention because they may be utilised for the genetic counselling of HBOC families, but more studies are needed to confirm the preliminary findings.

Supporting Information

Acknowledgments

We thank all of the patients and their family members for participation in this study. We also thank the personnel of the Tampere University Hospital Genetics Outpatient Clinic and JS’s research group for all of their help. In addition, we greatly acknowledge the Finnish Institute for Molecular Medicine Technology Centre for genomic services.