Mutations in BRIP1/FANCJ confer high risk of ovarian cancer Thorunn Rafnar 1*, Daniel F Gudbjartsson 1*, Patrick Sulem 1*, Aslaug Jonasdottir 1, Asgeir Sigurdsson 1, Adalbjorg Jonasdottir 1, Soren Besenbacher 1, Pär Lundin 1, Simon N Stacey 1, Julius Gudmundsson 1, Olafur Magnusson 1, Louise le Roux 1, Gudbjorg Orlygsdottir 1, Hafdis T Helgadottir 1, Hrefna Johannsdottir 1, Arnaldur Gylfason 1, Laufey Tryggvadottir 2,3 , Jon G Jonasson 2,3 , Ana De Juan 4, Eugenia Ortega 5, Jose M Ramon-Cajal 6, Maria D García -Prats 7, Carlos Mayordomo 8, Angeles Panadero 9, Fernando Rivera 4, Katja KH Aben 10,11 , Anne M van Altena 12 , Leon FAG Massuger 12 , Mervi Aavikko 13 , Paula M Kujala 14 , Synnöve Staff 15, 16 , Lauri A Aaltonen 13 , Kristrun Olafsdottir 17 , Johannes Bjornsson 17 , Augustine Kong 1, Anna Salvarsdottir 18 , Hafsteinn Saemundsson 18 , Karl Olafsson 18 , Kristrun R Benediktsdottir 3,17 , Jeffrey Gulcher 1, Gisli Masson 1, Lambertus A Kiemeney 10,11,19 , Jose I Mayordomo 20 , Unnur Thorsteinsdottir 1,3 , Kari Stefansson 1,3 1deCODE Genetics, Sturlugata 8, 101 Reykjavik, Iceland 2Icelandic Cancer Registry, Skógarhlíð 8, 105 Reykjavik 3Faculty of Medicine, University of Iceland, 101 Reykjavik, Iceland 4Division of Medical Oncolology, Marques de Valdecilla University Hospital, 39008 Santander, Spain 5Division of Medical Oncolology, Arnau de Vilanova University Hospital, Lérida, Spain 6Division of Gynecology, San Jorge General Hospital, 22004 Huesca, Spain 7Division of Pathology, San Jorge General Hospital, 22004 Huesca, Spain 8University of Zaragoza, 50009 Zaragoza, Spain 1 9Division of Medical Oncolology, Ciudad de Coria Hospital, 10800 Coria, Spain 10 Comprehensive Cancer Center The Netherlands, Location Nijmegen, 6501 BG Nijmegen, The Netherlands 11 Department of Epidemiology, Biostatistics and HTA, Radboud University Nijmegen Medical Centre, 6500 HB Nijmegen, The Netherlands 12 Department of Gynaecology, Radboud University Nijmegen Medical Centre, 6500 HB Nijmegen, The Netherlands 13 Department of Medical Genetics, Genome-Scale Biology Research Program, University of Helsinki, Helsinki, Finland 14 Department of Pathology, Centre for Laboratory Medicine, University of Tampere and Tampere University Hospital, Tampere, Finland 15 Laboratory of Cancer Biology, Institute of Biomedical Technology, University of Tampere, Finland 16 Department of Obstetrics and Gynecology, Tampere University Hospital, Tampere, Finland 17 Dept. Of Pathology, Landspitali University Hospital 101, Reykjavik, Iceland 18 Department of Obstetrics and Gynecology, Landspitali-University Hospital, 101 Reykjavik, Iceland 19 Department of Urology, Radboud University Nijmegen Medical Centre, 6500 HB Nijmegen, The Netherlands. 20 Division of Medical Oncology. University Hospital, 50009, Zaragoza, Spain * Authors with an equal contribution Correspondence to: Thorunn Rafnar, e-mail: [email protected] or Kari Stefansson, e-mail: [email protected] 2 Ovarian cancer causes more deaths than any other gynecologic malignancy in developed countries. Sixteen million sequence variants, identified through whole-genome sequencing of 457 Icelanders, were imputed into 41,675 SNP chip-typed Icelanders and their relatives and tested for association with ovarian cancer (N cases=656). We discovered a rare (0.41% allelic frequency) frame-shift mutation, c.2040_2041insTT, in the BRIP1/FANCJ gene that confers an increase in ovarian cancer risk (OR=8.13, P=2.8×10 -14 ). The mutation also associated with increased risk of cancer in general and reduced lifespan by 3.6 years. Another frameshift mutation in BRIP1 , c.1702_1703del, was seen in two out of 144 Spanish ovarian cancer cases and one out of 1780 Spanish controls (P=0.016) and also associated with breast cancer (seen in 6/927 cases, P=0.0079). Ovarian tumors from heterozygous carriers of the Icelandic mutation display a loss of the wild-type allele, indicating that BRIP1 serves as a classical tumor suppressor gene in ovarian cancer. 3 Worldwide, there were estimated to be 225,000 new cases of ovarian cancer and 140,000 deaths from the disease in 2008, accounting for around 4% of all cancers deaths in women 1. Family history of ovarian cancer is a strong risk factor for the disease as women with a single first-degree relative with ovarian cancer have a three to fourfold higher risk of the disease than average 2. Two familial cancer syndromes include high risk of ovarian cancer, the hereditary breast/ovarian cancer syndrome (HBOC) that is mostly attributed to mutations in the breast cancer genes BRCA1 or BRCA2, and Lynch syndrome caused by mutations in DNA mismatch repair (MMR) genes. In population-based studies, mutations in BRCA1 or BRCA2 are found in 5-15% of all ovarian cancer cases 3,4 whereas mutations in the MMR genes are detected in only 2% of the cases 5,6 . Recently, a genome-wide association study of 1,768 ovarian cancer cases and 2,354 controls, using DNA chips with common SNPs, yielded sequence variants at 9p22, 2q31, 8q24 and 19p13 that associate with a modest increase in risk of invasive epithelial ovarian cancer (all ORs 1.22) 7-9. However, these variants explain only a small fraction of the genetic risk for ovarian cancer (all s ). In order to search for rare and low-frequency variants with greater impact on the risk of disease than the common variants measured by DNA chips, the whole genomes of a large number of Icelanders are currently being sequenced. The SNPs identified through sequencing are then imputed into 41,675 chip typed Icelanders using long range phasing information 10 . The nation- wide Icelandic genealogical database then allows the propagation of this genotype information and creation of in silico genotypes for close relatives of the chip typed individuals who have not been genotyped (see Methods for a detailed description). This approach has recently led to the 4 discovery of two missense mutations that have a large effect on the risk of developing sick sinus syndrome 11 and gout 12 , respectively. To search for sequence variants that associate with invasive ovarian cancer, we performed genome-wide association scans of 15,957,390 autosomal SNPs identified through whole genome sequencing of 457 Icelanders and imputed into Icelandic cancer cases and controls. A total of 68 Icelandic ovarian cancer cases had been chip typed with the Illumina HumanHap300 or CNV370 chips and an additional 572 ovarian cancer cases without chip genotype information were assigned in silico genotypes. The genotypes of the cases were compared to the genotypes of controls matched on genotype informativeness, including 41,607 chip typed controls. Excluding markers that are correlated with the known BRCA2 999del5 founder mutation in Iceland 13 , the strongest association with ovarian cancer was with 9 highly correlated SNPs on chromosome 17q23.2 (Supplementary table 1). The most significant association was with the C allele of rs34289250 (OR=7.95, P=5.65×10 -13 , allelic frequency of 0.89% in controls). This SNP is in an intron of the BRCA1 interacting protein C-terminal helicase 1 (BRIP1/BACH1/FANCJ ) gene that is required for the normal double-strand break repair function of BRCA1 , located 20Mb telomeric of the BRCA1 gene 14 (Figure 1). Bi-allelic mutations in BRIP1 result in the chromosome instability disorder Fanconi anemia complementation group J 15-17 and heterozygous inactivating mutations in BRIP1 have been reported to have a modest effect on breast cancer risk 14,18 . Using whole-genome sequence data from 10 carriers of rs34289250, we scrutinized genes in the region for exonic mutations that might explain the strong association with ovarian cancer. A novel two basepair insert in exon 14 of the BRIP1 gene (c.2040_2041insTT, BRIP1 NM_032043), was detected in 4 of the 10 SNP carriers but in none of the 447 non-carriers 5 analyzed. The mutation causes a frameshift (p.L680fs) and premature termination of protein translation at amino acid 687 out of the 1,249 residues in the full-length protein (Supplementary figure 1). Using Sanger sequencing and microsatellite genotyping, we genotyped the insert directly in 11,741 Icelandic cancer cases and 3,913 controls, including 318 ovarian cancer cases. Using these data, the insert was imputed into all available cancer cases and controls, both those who had been chip genotyped and un-genotyped relatives of chip-typed individuals. Combining results from directly genotyped and imputed cancer cases, the association between ovarian cancer and the insert became even more significant (OR=8.13 (95% CI 4.74, 13.95), P=2.8×10 -14 ) (Table 1). Mutations in the genes shown to increase risk of ovarian cancer, i.e. BRCA1 , BRCA2 and the MMR genes, also carry an increased risk of other cancer types, mainly breast and pancreatic cancers in the case of BRCA1 and BRCA2 and colorectal and endometrial cancers in the case of the MMR genes. We examined whether c.2040_2041insTT is associated with an increase in the risk of other cancer types, using information from the nation-wide Icelandic Cancer Registry and genotype information from 14,331 directly genotyped or chip typed and imputed cancer cases and 16,873 in silico genotyped cancer cases (representing 23 different cancer types) (Supplementary table 2). When all the cancers were analyzed together, carriers of the insert had an increased risk of being diagnosed with cancer in general (OR=1.50, P=1.5×10 -6) (Table 1, Supplementary table 2). Given the large effect of the insert on ovarian cancer and the fact that ovarian cancer can co-occur with other cancer types, we repeated the analysis excluding the ovarian cancer cases (Supplementary table 3). Even without the ovarian cancer cases, the risk of any cancer remained significant (OR=1.36, P=9.2×10 -4). Although no cancer type was individually significantly associated with the insert after correcting for the number of tests (Table 6 1 and Supplementary table 2), two cancer types showed nominally significant association: pancreatic cancer (OR=2.71, P=0.0069) and rectal cancer (OR=2.25, P=0.025).